ProgrammableMarketing/assemblyai.com (opens in a new tab)
PDF

Views

  • Plan
  • Roadmap
  • Research

On this page

  • The brief
  • Start here
  • Idea portfolio
  • Effort vs impact
  • This week · 6
  • This quarter · 9
  • Big bets · 3
  • Wildcards · 3
  • How we can help
Generated Oct 18, 2026 2 views
  1. Programmable
  2. Playbooks
  3. AssemblyAI
AssemblyAI logo

growth playbook · Oct 18, 2026

assemblyai.com (opens in a new tab)

AssemblyAI growth playbook

Diagnosis

AssemblyAI is hiring to turn a product that wins evaluations into a sales machine: a Head of Sales for digital native companies, three AI-native AEs in San Francisco, New York and London, its founding Enterprise AE, a first Head of GTM in Europe, and two engineers who change how it finds customers: a GTM AI engineer to find "every company building voice AI before they contact us", and an Agent Experience engineer because developers' coding agents now do the integrating.

The playbook follows that: make every evaluation easier to start (EU residency that works on the page, one set of scale numbers, benchmarks on the buyer's own task), turn usage and agent-driven signups into pipeline, choose Europe's first countries from AssemblyAI's own data, and market to the coding agents that pick APIs.

The free sample idea fixes a broken EU residency page and puts the EU endpoint on the European language pages, the cheapest win for the new London team.

Growth Score

8/10

Biggest Opportunity

Europe and the coding agent: buyers who need EU residency and developers whose agents choose the API. AssemblyAI already has the EU endpoint, llms.txt and an MCP server; showing them where the decision is made could open both channels within a quarter.

Biggest Risk

Speech-to-text is crowded and getting cheaper: Deepgram, Speechmatics, Gladia, Soniox, ElevenLabs and the model labs all sell accuracy, languages and voice agents, and Deepgram's packages are downloaded about three times as often on npm. A sales team that grows faster than its proof and its targeting would just spend more per deal.

01

Start here

The three moves to make first. Each one is written up in full below.

  1. 1

    Put EU data residency on the pages European buyers land on, starting with the broken /deployments/eu-residency link

    Fix the 404 today: redirect /deployments/eu-residency to the docs page while the real page is written.

    High impact First result in 2 weeks
  2. 2

    One set of scale numbers, everywhere: 600M or 840M calls a month, 1M or 2M hours a day

    Ask Finance for this month's calls, hours and data volume.

    Medium impact First result in 1 week
  3. 3

    Send each target account a Voice Standard Bench result for its own product: "your post-processing on a cheaper model"

    Pick 20 accounts from the AE book and tag each with its main voice task.

    High impact First result in 2 weeks

02

Your idea portfolio

21 ideas, grouped by when to run them. Open any idea for how to run it: the steps, who to approach, copy you can use, and how to measure it.

  • 8 ideas are marked Unexpected: plays you are unlikely to have tried.

    1 of these are things Programmable can build and run for you.

Every idea placed by the effort it takes against the impact it could have. Quick wins sit top left; select a number to open that idea. The filters below narrow the list by status, timing, category and what Programmable can build.

Effort vs impactEach number is an idea. Select one to open it.
Impact →
Quick wins · 4Big bets · 10Fill-ins · 4Reconsider · 3
Effort →

Showing 21 of 21 ideas

This week 6

Quick wins with a clear first action.

  • 01

    AssemblyAI already has the answer Europe's buyers ask first: an EU endpoint that keeps audio and transcripts in the EU, for pre-recorded, streaming, Sync, Dictation and LLM Gateway. But the page the sitemap lists for it is a 404, and the German, French, Spanish and Dutch pages promise "fully compliant" without saying where the data goes.

    High impactLow effort First result in 2 weeks Head of GTM

    First action

    Fix the 404 today: redirect /deployments/eu-residency to the docs page while the real page is written.

    How to measure it

    Accounts making their first call to api.eu.assemblyai.com

    Now
    –
    Target
    +30% in a quarter
    By
    90 days

    Steps

    1. 1

      Stop the 404

      Redirect /deployments/eu-residency to the data-residency docs so the sitemap link works today.

      Marketing 30 minutes

    2. 2

      List what stays in the EU

      Per product: pre-recorded, streaming, Sync, Dictation, LLM Gateway, Voice Agent API; note anything that does not yet run in the EU.

      Head of GTM, Europe 1 day

    3. 3

      Write the page

      What stays in the EU, the one-line code change, the DPA, the sub-processor list, the biometric addendum and self-hosted options, with a "talk to the London team" form.

      Marketing 2 days

    4. 4

      Legal review

      Counsel checks every promise on the page against the DPA and the docs.

      Head of GTM, Europe 3 days

    5. 5

      Add the block to language pages

      A short EU residency block with the base URL on the German, French, Spanish, Dutch, Italian, Portuguese and Nordic language pages.

      Marketing 1 day

    6. 6

      Change the code sample

      Show base_url="https://api.eu.assemblyai.com" in the sample on European language pages.

      Software Engineer, Agent Experience 2 hours

    7. 7

      Link from compare pages

      Add the block to the Speechmatics and Gladia compare pages, where European buyers compare.

      Marketing 2 hours

    8. 8

      Measure

      Track first EU-endpoint calls by signup source each week and share them with the Head of GTM, Europe.

      Senior GTM AI Engineer 2 hours

    First result
    2 weeks
    Cost
    $0 plus legal review
    Owner
    Head of GTM, Europe

    Smallest test

    Ship the redirect and the block on the German page alone, and compare EU-endpoint first calls from German-page signups over four weeks.

    Why it works

    Rebuild /deployments/eu-residency as a real page (what stays in the EU, per product; the base URL; the DPA, sub-processors and biometric addendum; self-hosted options), then add a short "Your data stays in the EU" block with the endpoint line of code to every European language page, the enterprise page and the compare pages for Speechmatics and Gladia. Make the EU endpoint the default in the code sample on European language pages.

    The new Head of GTM, Europe is told to make compliance "an advantage, not a blocker" and to run evaluation-led deals that replace incumbents. Data residency is the first gate in those evals; a working page that answers it, linked from where the buyer already is, shortens every regional deal before the first hire in London starts.

    Ready to use

    CopyEU residency block for language pages

    Your audio and transcripts can stay in the EU
    Use the EU endpoint and your data never leaves the European Union: transcription, streaming, dictation and LLM Gateway.
    
        base_url = "https://api.eu.assemblyai.com"
    
    DPA, sub-processors and our biometric data addendum: [link]
    Talk to our European team: [link]

    Risks and how to handle them

    • A promise goes further than the product

      Write per product and have counsel review; say plainly where a product is US-only today.

    • Engineers copy the EU sample by mistake

      Label it clearly as the EU option and keep the default sample on non-European pages.

    Evidence

    • Your websiteThe sitemap lists /deployments/eu-residency, which returns a 404 page ("This page could not be found").Source
    • Your websiteLanguage pages for German, French, Spanish and Dutch (in English): the German page promises "From call centers in Berlin to media companies in Vienna" and "cloud or self-hosted, fully compliant", but none of the four mentions EU data residency, the EU endpoint or GDPR.Source
    • DocsDocs, "Cloud Endpoints and Data Residency": the EU endpoint api.eu.assemblyai.com "guarantees your data never leaves the European Union"; streaming is processed in AWS regions in the US and the EU; Sync STT, Dictation and LLM Gateway each have their own data-residency page; there is a self-serve BAA, and the legal pages include a biometric data addendum.Source
    • HiringHead of GTM, Europe (London): "Europe is already one of our fastest growing regions." Calling platforms, contact center software and voice agent companies are moving workloads to AssemblyAI "for multilingual accuracy, often replacing incumbents or self-hosted models after head-to-head evals". Today the region is "covered largely from the US"; this is the first senior leader on the ground. Duties: choose which countries come first (UK & Ireland, DACH, France, Benelux, Nordics, Southern Europe), build Sales, Applied AI and Forward Deployed Engineering, handle "GDPR, the EU AI Act, data residency, DPAs" so compliance "becomes an advantage", represent AssemblyAI at conferences and meetups, and report "which languages, accents, and deployment requirements are winning or losing deals".Source
  • 02

    The job posts say "600M+ inference calls monthly" and "1M+ hours of audio daily", then in the same overview "over 2 million hours of audio daily" and "more than 1 million API calls every day" (about 30M a month). The website says "840M+ API calls processed every month" and "40TB+ data processed every single day". An enterprise buyer, an analyst or an AI assistant summarizing AssemblyAI gets four different sizes.

    Medium impactLow effort First result in 1 week Marketing

    First action

    Ask Finance for this month's calls, hours and data volume.

    How to measure it

    Places with the current numbers and an "as of" date

    Now
    –
    Target
    100% of site, job posts and decks
    By
    30 days

    Steps

    1. 1

      Get the numbers

      Monthly API calls, hours a day and data a day, from the warehouse, with a date.

      Marketing 1 day

    2. 2

      Choose the wording

      One sentence for each, rounded down, with "as of" month.

      Marketing 1 hour

    3. 3

      Fix the job posts

      Replace the overview paragraph in all 15 Greenhouse posts.

      GTM Recruiter 1 hour

    4. 4

      Fix the site

      Enterprise page, startup program page and homepage stats.

      Marketing 2 hours

    5. 5

      Add to llms.txt

      A short "AssemblyAI at a glance" section with the same numbers.

      Software Engineer, Agent Experience 30 minutes

    6. 6

      Update decks and boilerplate

      Sales deck, press boilerplate, partner one-pagers.

      Marketing 1 day

    7. 7

      Set a refresh date

      A quarterly calendar reminder to update all of them together.

      Marketing 15 minutes

    First result
    1 week
    Cost
    $0
    Owner
    Marketing

    Smallest test

    Fix the job-post overview and the enterprise page first, the two places buyers and candidates read most.

    Why it works

    Pick the current numbers with Finance, put them in one "facts" source the website, job posts, decks and press boilerplate all read from, and add an "as of" date. Give AI assistants the same line in llms.txt.

    AssemblyAI is hiring a founding Enterprise AE whose deals go through procurement and security review, and a GTM Recruiter whose candidates read the job posts. Scale is the proof point both use, and numbers that disagree invite the question the rep least wants.

    Risks and how to handle them

    • Numbers go stale again

      Keep one source and a quarterly owner.

    • Publishing a lower figure

      Use a clearly defined measure (calls or hours) rather than the biggest one.

    Evidence

    • Job postsCompany overview in the job posts: "under 100 people generating roughly $600K ARR per employee"; models "serve 600M+ inference calls monthly, process 1M+ hours of audio daily"; later in the same overview, "processing over 2 million hours of audio daily and handling more than 1 million API calls every day". Named customers: Granola, Fireflies, Figure AI and CallRail.Source
    • Your websiteEnterprise page and Startup Program page: "94% Most word accurate model on the market", "840M+ API calls processed every month", "40TB+ Data processed every single day", "<1hr Average time to first API call". The enterprise page groups solutions by industry (contact center, healthcare, financial services, legal, education, media, retail).Source
    • HiringEnterprise Account Executive, described as "AssemblyAI's founding Enterprise Account Executive": own a book of strategic enterprise accounts, lead model evaluations and win enterprise engineering and product leaders, through procurement, legal and security review.Source
    • HiringGTM Recruiter: "our go-to-market team is growing fast"; hires Account Executives, BDRs, Customer Success, Forward Deployed Engineers and Marketers, 5-6 open roles at a time; "we're hiring a lot of people". Also open: a Finance Engineer for deal desk and sales-compensation systems, and inference engineers in the UK and North America. 15 open roles in all, 6 of them in sales.Source
  • 03

    The job posts ask AEs for "custom benchmarks ... that earn replies". AssemblyAI just published the raw material: 50+ LLMs scored on the voice tasks its customers run (summaries, extraction, cleanup) with cost and speed, and the finding that most of them do not need frontier prices.

    High impactLow effort First result in 2 weeks Digital Native AI Account Exe…

    First action

    Pick 20 accounts from the AE book and tag each with its main voice task.

    How to measure it

    Meetings from benchmark outreach

    Now
    –
    Target
    10% of accounts contacted
    By
    30 days

    Steps

    1. 1

      Choose 20 accounts

      From the digital native book, with an obvious voice task in their product.

      Digital Native AI Account Executive 2 hours

    2. 2

      Map the task

      Summary, extraction or text correction, and the likely transcript length.

      Digital Native AI Account Executive 2 hours

    3. 3

      Pull the bench rows

      The likely current model and the cheapest model that matched it, with cost and speed.

      Applied AI Engineering 1 day

    4. 4

      Estimate the saving

      At a stated, conservative volume; show the assumption.

      Digital Native AI Account Executive 2 hours

    5. 5

      Write the one-pager

      Task, two models, cost per hour of audio, saving, offer to test on their transcripts.

      Digital Native AI Account Executive 1 day

    6. 6

      Send and follow up

      Email to the CTO or VP of AI, with a LinkedIn note and a second touch at day 5.

      Digital Native AI Account Executive 2 days

    7. 7

      Run the real test

      For every yes, run their sample transcripts through the bench within 48 hours.

      Applied AI Engineering ongoing

    First result
    2 weeks
    Cost
    $0
    Owner
    Digital Native AI Account Executive

    Smallest test

    Send five one-pagers this week and count replies.

    Why it works

    For each AI-native target account, pick the voice task its product clearly runs (a notetaker's summary, a scribe's note, a dictation app's cleanup), and send a one-page result: the model it probably uses today, the cheapest model that matched it on the bench, and what that would save at its likely volume, with an offer to run the same test on its own transcripts.

    This is an opener about the buyer's cost, not AssemblyAI's accuracy, so it gets a reply from teams already happy with their speech-to-text. It also sells LLM Gateway, which gives AEs a second product to land with.

    Risks and how to handle them

    • Guessing the wrong model or volume

      Say it is an estimate and invite them to correct it.

    • Looks like a teardown of their stack

      Frame it as savings, never as a criticism of their choice.

    Evidence

    • Your websiteVoice Standard Bench (blog, October 13, 2026): LLM Gateway "now has 50+ models", on the same API key as speech-to-text, including low-cost open models hosted on AssemblyAI's own infrastructure; the bench scores LLMs on summarization, information extraction and text correction at four transcript lengths, with cost and speed side by side. "The main finding: most voice tasks don't need frontier models, or frontier-model prices."Source
    • HiringDigital Native AI Account Executive, posted three times (San Francisco, New York, London): own a book of AI-native startup accounts, generate pipeline through "personalized outreach, conferences, exclusive events, and finding warm introductions", send "custom benchmarks, competitive insights, use-case-specific demos that earn replies", and win CTOs, CPOs and VPs of AI. Expect "to be on the road 50%+ of each month".Source
    • HiringHead of Sales, Digital Native (San Francisco): scale "a small team of AEs and BDRs today to a larger sales org within the next year", own the land number for digital native and AI-native companies, hire the front-line managers, design territories and target account tiers, and drive "an outbound-first culture where pipeline is created, not waited for", with "custom benchmarks, competitive insight, and use-case-specific demos". The motion is "technical, product-led, and evaluation-driven": get builders to run an eval, then convert wins into production usage.Source
  • 04

    The sitemap still lists /waitlist/streaming, which says Universal-Streaming "is now live!" above a waitlist form and spells the company "AssesmblyAI", plus /humanx ("See you at booth 315"), /waitlist/slam-1 and /waitlist/reinvent. These pages rank for product names and greet developers with an expired event or a form for something they can already use.

    Medium impactLow effort First result in 1 week Software Engineer

    First action

    List every URL with "waitlist", "booth" or an event name in the sitemap.

    How to measure it

    Old campaign URLs still served as pages

    Now
    –
    Target
    0
    By
    2 weeks

    Steps

    1. 1

      Find them

      Search the sitemap for waitlist, event and report-year pages.

      Software Engineer, Agent Experience 1 hour

    2. 2

      Decide each

      Redirect to the live product, merge into an events hub, or retire.

      Marketing 1 hour

    3. 3

      Redirect

      Set 301s and remove the old URLs from the sitemap.

      Software Engineer, Agent Experience 2 hours

    4. 4

      Fix the typo

      Correct "AssesmblyAI" wherever it appears.

      Marketing 15 minutes

    5. 5

      Update llms.txt links

      Make sure no old page is linked from llms.txt or docs.

      Software Engineer, Agent Experience 30 minutes

    6. 6

      Add a check

      A quarterly script that flags pages with past dates or "coming soon".

      Software Engineer, Agent Experience 2 hours

    First result
    1 week
    Cost
    $0
    Owner
    Software Engineer, Agent Experience

    Smallest test

    Redirect /waitlist/streaming first and watch streaming signups from that URL.

    Why it works

    Redirect each old waitlist to its live product page, turn event pages into a single events hub, fix the typo, and drop retired pages from the sitemap. Add a quarterly check for pages with dates or "coming soon" wording.

    The Agent Experience role exists because agents and developers read the site literally; an agent that lands on a waitlist may tell the developer the product is not out yet.

    Risks and how to handle them

    • Losing a page that still converts

      Check signups per page for the last 90 days before retiring it.

    Evidence

    • Your websiteOld campaign pages still in the sitemap: /waitlist/streaming ("Universal-Streaming is now live!", a waitlist form, and the typo "AssesmblyAI"), /humanx ("See you at booth 315"), /waitlist/slam-1 and /waitlist/reinvent.Source
    • HiringSoftware Engineer, Agent Experience: "More and more developers don't read our docs or click through our dashboard. Their coding agent does." Owns the skills, MCP server, llms.txt, code samples and docs an agent reads; the north star is that "an agent should go from zero to a working, production-grade AssemblyAI integration on the first try". Builds evals that run real coding agents on real integration tasks and tracks "success rate, time to first working call, and activation from agent-driven signups".Source
    • Your websiteSitemap (1,655 URLs): 26 /compare pages (Deepgram, Speechmatics, Gladia, Soniox, ElevenLabs and Scribe v2, OpenAI Realtime, Whisper, AWS, Google and Azure, plus LLM gateways such as OpenRouter, LiteLLM, Portkey and the Vercel, Cloudflare and Kong AI gateways), 99 /languages pages, 31 /solutions, 27 /partners, 24 /customers, about 800 blog posts and 700+ docs URLs.Source
  • 05

    The first GTM AI engineer is asked to watch "product usage and tell an AE when a customer is about to expand or churn". The quickest useful version is a single SQL query on the BigQuery usage data, not a scoring model.

    High impactLow effort First result in 2 weeks Senior GTM AI Engineer

    First action

    Write the query for "weekly hours up 100% and over a spend line" and run it on last month.

    How to measure it

    Alerts that become a meeting within 14 days

    Now
    –
    Target
    25%
    By
    60 days

    Steps

    1. 1

      Write the query

      Weekly hours by account, week-over-week change, spend line.

      Senior GTM AI Engineer 1 day

    2. 2

      Add the company

      Map the account to a domain from the signup email; skip personal emails.

      Senior GTM AI Engineer 1 day

    3. 3

      Post daily

      A Slack post to the digital native channel at 9am with each account and an owner.

      Senior GTM AI Engineer 2 hours

    4. 4

      Agree the rule

      With the Head of Sales: who takes which alert, and within how long.

      Head of Sales, Digital Native 1 hour

    5. 5

      Log the outcome

      Meeting, no response or not a fit, in the CRM.

      Digital Native AI Account Executive ongoing

    6. 6

      Tune

      Adjust the thresholds after four weeks using the outcomes.

      Senior GTM AI Engineer 2 hours

    First result
    2 weeks
    Cost
    $0
    Owner
    Senior GTM AI Engineer

    Smallest test

    Run the query on last month and ask an AE to contact the top ten accounts.

    Why it works

    Each morning, list self-serve accounts whose weekly audio hours doubled or crossed a set spend line, with the company domain, products used and recent growth, and post them to the digital native sales channel with an owner. Track which alerts become meetings.

    Product-led sign-ups that grow fast are the cheapest deals the new digital native team will ever win. The job posts say customers choose AssemblyAI after running an eval; an account doubling its usage has already run one.

    Risks and how to handle them

    • AEs crowd out happy self-serve users

      Lead with help (rate limits, concurrency, pricing tiers), not a contract.

    • Privacy

      Use account-level usage only, never audio or transcript content.

    Evidence

    • HiringSenior GTM AI Engineer (the first): a software engineer inside the revenue org who builds "the pipelines, internal tools, and agents that decide which companies we go after": "Finding every company building voice AI before they contact us", account research, first-email drafts, "Watching product usage and telling an AE when a customer is about to expand or churn", and turning product usage data in BigQuery and Looker into usage alerts, expansion triggers and account health scores. Judged on whether pipeline and revenue move.Source
    • HiringHead of Sales, Digital Native (San Francisco): scale "a small team of AEs and BDRs today to a larger sales org within the next year", own the land number for digital native and AI-native companies, hire the front-line managers, design territories and target account tiers, and drive "an outbound-first culture where pipeline is created, not waited for", with "custom benchmarks, competitive insight, and use-case-specific demos". The motion is "technical, product-led, and evaluation-driven": get builders to run an eval, then convert wins into production usage.Source
    • Your websitePricing (read October 18, 2026): Universal-3.5 Pro $0.21/hr, "works across 18 languages, with native code switching"; Universal-2 $0.15/hr, "Supports 99 languages"; add-ons such as keyterms prompting $0.05/hr, speaker diarization $0.02/hr and Medical Mode $0.15/hr; "Start free, pay-as-you-go after that – no commitments required".Source
  • 06

    Real Chrome users on mobile wait 5.2 seconds at the 75th percentile for the main content, 39% slower than a year ago, on a homepage that leads with a live voice recorder and a code sample. For a company whose pitch is speed, the site is the first latency test a developer runs.

    Medium impactMedium effort First result in 1 month Software Engineer

    First action

    Run PageSpeed Insights on the homepage and pricing page and list the LCP element on each.

    How to measure it

    Mobile LCP p75

    Now
    –
    Target
    under 2.5 s
    By
    60 days

    Steps

    1. 1

      Measure

      PageSpeed Insights on the homepage, pricing and two language pages.

      Software Engineer, Agent Experience 1 hour

    2. 2

      Find the LCP element

      And what delays it: fonts, scripts, images or the recorder.

      Software Engineer, Agent Experience 2 hours

    3. 3

      Load the recorder later

      Render the headline first and load the recorder and code sample after.

      Software Engineer, Agent Experience 1 day

    4. 4

      Trim scripts

      Defer analytics and consent scripts that block rendering.

      Software Engineer, Agent Experience 1 day

    5. 5

      Test on a phone

      A mid-range Android on 4G.

      Software Engineer, Agent Experience 1 hour

    6. 6

      Re-measure

      Chrome UX Report after 28 days, and signups from mobile.

      Software Engineer, Agent Experience 1 hour

    First result
    1 month
    Cost
    $0
    Owner
    Software Engineer, Agent Experience

    Smallest test

    Lazy-load the recorder on the pricing page first and compare LCP.

    Why it works

    Measure the homepage and pricing page in PageSpeed Insights, find the largest element and what blocks it, lazy-load the recorder and code sample below the headline, and check the cookie and analytics scripts. Re-measure in the Chrome UX Report after 28 days.

    Signups start on these pages, and an AI-native founder on a phone is the buyer the new digital native team is chasing.

    Risks and how to handle them

    • Moving the recorder lowers demos tried

      A/B the position and watch recorder use and signups.

    Evidence

    • programmable.infoSite scan: Chrome UX Report mobile LCP p75 5,175 ms, 39% slower than a year earlier (INP 129 ms, CLS 0.03); the homepage runs a live "Try your own voice" recorder and code sample above the fold. The assemblyai npm package has about 289,000 weekly downloads (Deepgram's packages about 922,000); PyPI package last released September 24, 2026; 71 public GitHub repos.Source

This quarter 9

The core plays for the next three months.

  • 07

    The Agent Experience engineer will build evals that run real coding agents on real integration tasks. That is also a marketing asset nobody in speech-to-text has: a public, repeatable number for how easily an agent builds with each API.

    High impactMedium effort First result in 2 months Software Engineer

    First action

    Write the first five integration tasks and run them by hand in one agent.

    How to measure it

    First-try success rate

    Now
    –
    Target
    90% on the five tasks
    By
    90 days

    Steps

    1. 1

      Write the tasks

      Five realistic tasks across pre-recorded, streaming, Voice Agent API and LLM Gateway.

      Software Engineer, Agent Experience 2 days

    2. 2

      Automate the run

      A harness that runs each agent in a clean repo and checks a working call.

      Software Engineer, Agent Experience 2 weeks

    3. 3

      Record results

      Success, time to first call, and the failure reason.

      Software Engineer, Agent Experience 1 day

    4. 4

      Fix what fails

      Docs, skills, MCP tools and samples, one failure at a time.

      Software Engineer, Agent Experience 3 weeks

    5. 5

      Publish the page

      Monthly score, method, prompts and outputs.

      Marketing 2 days

    6. 6

      Add competitors

      Same tasks against two other APIs, with method shown.

      Software Engineer, Agent Experience 1 week

    7. 7

      Use in sales

      A slide and a line in the AE outbound kit.

      Digital Native AI Account Executive 1 hour

    First result
    2 months
    Cost
    Agent API costs, under $500 a month
    Owner
    Software Engineer, Agent Experience

    Smallest test

    Run five tasks in one agent by hand and publish the result as a blog post.

    Why it works

    Run a fixed set of tasks ("add live captions to a Next.js app", "transcribe a call with speakers", "build a voice agent with turn detection") through three coding agents every month, publish success rate and time to first working call, show the prompt and the code, and fix what fails. Add competitors' APIs once the method is stable.

    Developers increasingly let the agent choose the API. A public score gives the agent experience work a target, gives AEs a proof point for AI-native buyers, and becomes a page agents themselves cite.

    Risks and how to handle them

    • A competitor comparison looks biased

      Publish prompts, code and failures, and invite vendors to submit fixes.

    • Score drops when agents change

      Show the score over time and date every run.

    Evidence

    • HiringSoftware Engineer, Agent Experience: "More and more developers don't read our docs or click through our dashboard. Their coding agent does." Owns the skills, MCP server, llms.txt, code samples and docs an agent reads; the north star is that "an agent should go from zero to a working, production-grade AssemblyAI integration on the first try". Builds evals that run real coding agents on real integration tasks and tracks "success rate, time to first working call, and activation from agent-driven signups".Source
    • Your websitellms.txt already exists and tells agents to "prefer the Markdown sources linked in this file over crawling the HTML marketing site"; it points coding assistants to "Build with AI coding agents" and a docs MCP server. The use-case pages include "Dictation for Coding Agents".Source
    • Your websiteSitemap (1,655 URLs): 26 /compare pages (Deepgram, Speechmatics, Gladia, Soniox, ElevenLabs and Scribe v2, OpenAI Realtime, Whisper, AWS, Google and Azure, plus LLM gateways such as OpenRouter, LiteLLM, Portkey and the Vercel, Cloudflare and Kong AI gateways), 99 /languages pages, 31 /solutions, 27 /partners, 24 /customers, about 800 blog posts and 700+ docs URLs.Source
  • 08

    The Agent Experience role tracks "activation from agent-driven signups", but today those signups look like ordinary organic sign-ups. If a growing share of developers arrive because Claude Code or Cursor picked AssemblyAI, the sales team needs to see that channel to staff it.

    High impactMedium effort First result in 1 month Senior GTM AI Engineer

    First action

    Check which signup events can be tied to MCP or Markdown docs traffic today.

    How to measure it

    Agent-driven signups measured weekly

    Now
    –
    Target
    a weekly report in place
    By
    30 days

    Steps

    1. 1

      Find the signals

      MCP server sessions, .md doc fetches, llms.txt hits and agent user agents.

      Software Engineer, Agent Experience 2 days

    2. 2

      Join to signups

      Tie sessions to accounts within a short window.

      Senior GTM AI Engineer 1 week

    3. 3

      Define the channel

      Rules for "agent-driven" agreed with marketing and sales.

      Marketing 2 hours

    4. 4

      Weekly report

      Agent-driven signups, activation and first-month spend, in Looker.

      Senior GTM AI Engineer 2 days

    5. 5

      Route to AEs

      Funded companies with an agent-driven signup go to the digital native queue.

      Senior GTM AI Engineer 2 days

    6. 6

      Review monthly

      With the Head of Sales: volume, conversion and staffing.

      Head of Sales, Digital Native 1 hour a month

    First result
    1 month
    Cost
    $0
    Owner
    Senior GTM AI Engineer

    Smallest test

    Count last month's signups that fetched a .md doc page first.

    Why it works

    Tag signups that arrive through the MCP server, docs Markdown and llms.txt fetches, or an API key created from an agent session; report them weekly as a channel; and send agent-driven signups from funded companies into the digital native AE queue with what was built.

    A channel that is not measured cannot get budget. Showing that agents bring in companies would justify the agent experience team and give AEs warm accounts with a working integration already in place.

    Risks and how to handle them

    • Attribution is fuzzy

      Report it as a lower bound and show the rules.

    • Privacy

      Use request metadata only, never code or audio contents.

    Evidence

    • HiringSoftware Engineer, Agent Experience: "More and more developers don't read our docs or click through our dashboard. Their coding agent does." Owns the skills, MCP server, llms.txt, code samples and docs an agent reads; the north star is that "an agent should go from zero to a working, production-grade AssemblyAI integration on the first try". Builds evals that run real coding agents on real integration tasks and tracks "success rate, time to first working call, and activation from agent-driven signups".Source
    • HiringSenior GTM AI Engineer (the first): a software engineer inside the revenue org who builds "the pipelines, internal tools, and agents that decide which companies we go after": "Finding every company building voice AI before they contact us", account research, first-email drafts, "Watching product usage and telling an AE when a customer is about to expand or churn", and turning product usage data in BigQuery and Looker into usage alerts, expansion triggers and account health scores. Judged on whether pipeline and revenue move.Source
    • Your websitellms.txt already exists and tells agents to "prefer the Markdown sources linked in this file over crawling the HTML marketing site"; it points coding assistants to "Build with AI coding agents" and a docs MCP server. The use-case pages include "Dictation for Coding Agents".Source
  • 09

    The Head of GTM, Europe must decide which countries come first. AssemblyAI already holds the best evidence: which European accounts use the EU endpoint, which languages they transcribe most and where self-serve spend is growing.

    High impactMedium effort First result in 6 weeks Head of GTM

    First action

    Pull EU-endpoint accounts and spend by billing country for the last six months.

    How to measure it

    Pipeline from the two chosen countries

    Now
    –
    Target
    2x in two quarters
    By
    6 months

    Steps

    1. 1

      Pull usage

      EU-endpoint accounts, spend and growth by billing country.

      Senior GTM AI Engineer 2 days

    2. 2

      Add language volume

      Aggregate hours by detected European language; no content.

      Senior GTM AI Engineer 2 days

    3. 3

      Add signups and deals

      Signups and open opportunities by country.

      Senior GTM AI Engineer 1 day

    4. 4

      Score countries

      Weight demand, growth and language fit, and rank.

      Head of GTM, Europe 1 day

    5. 5

      Pick two

      With the VP of Customer Solutions; write down why.

      Head of GTM, Europe 1 day

    6. 6

      Build the account list

      Top 50 accounts in each chosen country.

      Head of GTM, Europe 1 week

    7. 7

      Review quarterly

      Re-run the scorecard every quarter.

      Head of GTM, Europe 1 day a quarter

    First result
    6 weeks
    Cost
    $0
    Owner
    Head of GTM, Europe

    Smallest test

    Pull the EU-endpoint numbers by country and see whether two countries stand out.

    Why it works

    Build a one-page country scorecard from aggregate data: EU-endpoint accounts and spend by country, language-detection volume by European language, signups from each country, and open deals. Pick two countries to staff first, and set the first quarter's target account list from the top accounts there.

    The job post says Europe is growing from a footprint "covered largely from the US"; the accounts that found AssemblyAI on their own show where demand already is, which makes the first European hires productive faster.

    Risks and how to handle them

    • Usage reflects the past, not the future

      Add a short list of known large prospects to the scorecard.

    • Privacy

      Use aggregate counts by country and language only.

    Evidence

    • HiringHead of GTM, Europe (London): "Europe is already one of our fastest growing regions." Calling platforms, contact center software and voice agent companies are moving workloads to AssemblyAI "for multilingual accuracy, often replacing incumbents or self-hosted models after head-to-head evals". Today the region is "covered largely from the US"; this is the first senior leader on the ground. Duties: choose which countries come first (UK & Ireland, DACH, France, Benelux, Nordics, Southern Europe), build Sales, Applied AI and Forward Deployed Engineering, handle "GDPR, the EU AI Act, data residency, DPAs" so compliance "becomes an advantage", represent AssemblyAI at conferences and meetups, and report "which languages, accents, and deployment requirements are winning or losing deals".Source
    • DocsDocs, "Cloud Endpoints and Data Residency": the EU endpoint api.eu.assemblyai.com "guarantees your data never leaves the European Union"; streaming is processed in AWS regions in the US and the EU; Sync STT, Dictation and LLM Gateway each have their own data-residency page; there is a self-serve BAA, and the legal pages include a biometric data addendum.Source
    • HiringSenior GTM AI Engineer (the first): a software engineer inside the revenue org who builds "the pipelines, internal tools, and agents that decide which companies we go after": "Finding every company building voice AI before they contact us", account research, first-email drafts, "Watching product usage and telling an AE when a customer is about to expand or churn", and turning product usage data in BigQuery and Looker into usage alerts, expansion triggers and account health scores. Judged on whether pipeline and revenue move.Source
  • 10

    Every European enterprise deal will ask the same questions in security and legal review. The answers exist across docs, legal pages and a self-serve BAA, but nothing hands a European procurement team one complete pack.

    High impactMedium effort First result in 2 months Head of GTM

    First action

    Collect the last ten European security questionnaires and list the repeated questions.

    How to measure it

    Median security review time

    Now
    –
    Target
    -30%
    By
    2 quarters

    Steps

    1. 1

      Collect questions

      The last ten European questionnaires and RFPs.

      Head of GTM, Europe 2 days

    2. 2

      Draft answers

      Per product, from the docs and legal pages.

      Legal 1 week

    3. 3

      EU AI Act section

      Counsel writes how AssemblyAI answers buyers' questions under it, without overstating.

      Legal 2 weeks

    4. 4

      Design the pack

      A short PDF and a web page with the same content.

      Marketing 1 week

    5. 5

      Add to sequences

      A link in every European AE and SDR sequence.

      Head of GTM, Europe 1 hour

    6. 6

      Keep it current

      A named owner and a date on each section.

      Legal quarterly

    First result
    2 months
    Cost
    Legal time
    Owner
    Head of GTM, Europe

    Smallest test

    Answer the ten most common questions in a one-page document and send it in two live deals.

    Why it works

    With counsel, build a downloadable pack and a matching page: where data is processed per product, retention and model-training opt-out, the DPA and sub-processors, the biometric data addendum for voice, and written answers to the EU AI Act questions buyers send. Offer it on the EU residency page and in every European AE sequence.

    The job post asks the Head of GTM, Europe to make "GDPR, the EU AI Act, data residency, DPAs" an advantage. A pack sent on the first call removes weeks from review and signals that the London team knows the buyer's world.

    Risks and how to handle them

    • Legal interpretation changes

      Date every answer and review each quarter.

    • Overpromising

      Counsel signs off every sentence.

    Evidence

    • HiringHead of GTM, Europe (London): "Europe is already one of our fastest growing regions." Calling platforms, contact center software and voice agent companies are moving workloads to AssemblyAI "for multilingual accuracy, often replacing incumbents or self-hosted models after head-to-head evals". Today the region is "covered largely from the US"; this is the first senior leader on the ground. Duties: choose which countries come first (UK & Ireland, DACH, France, Benelux, Nordics, Southern Europe), build Sales, Applied AI and Forward Deployed Engineering, handle "GDPR, the EU AI Act, data residency, DPAs" so compliance "becomes an advantage", represent AssemblyAI at conferences and meetups, and report "which languages, accents, and deployment requirements are winning or losing deals".Source
    • DocsDocs, "Cloud Endpoints and Data Residency": the EU endpoint api.eu.assemblyai.com "guarantees your data never leaves the European Union"; streaming is processed in AWS regions in the US and the EU; Sync STT, Dictation and LLM Gateway each have their own data-residency page; there is a self-serve BAA, and the legal pages include a biometric data addendum.Source
    • HiringEnterprise Account Executive, described as "AssemblyAI's founding Enterprise Account Executive": own a book of strategic enterprise accounts, lead model evaluations and win enterprise engineering and product leaders, through procurement, legal and security review.Source
  • 11

    The European job post says customers move to AssemblyAI "often replacing incumbents or self-hosted models after head-to-head evals", and there is already a Whisper compare page. Teams running open models themselves rarely count the GPU, on-call and accuracy cost.

    High impactMedium effort First result in 2 months Forward Deployed Engineering

    First action

    Write the one-page eval scope and success criteria.

    How to measure it

    Evals converted to production

    Now
    –
    Target
    40%
    By
    2 quarters

    Steps

    1. 1

      Scope the eval

      Audio sample size, metrics, a one-week timeline and what the customer provides.

      Forward Deployed Engineering 2 days

    2. 2

      Cost model

      A simple calculator for GPU hours, engineer time and incidents.

      Finance Engineer 3 days

    3. 3

      Report template

      Side-by-side accuracy, latency and cost with a migration plan.

      Forward Deployed Engineering 2 days

    4. 4

      Promote it

      On the Whisper compare page, the docs and in AE outbound.

      Marketing 2 days

    5. 5

      Find teams

      Companies with job posts for speech ML or Whisper fine-tuning.

      Senior GTM AI Engineer 1 week

    6. 6

      Run the first five

      With named accounts from the digital native book.

      Forward Deployed Engineering 6 weeks

    7. 7

      Publish the pattern

      An anonymized write-up of what self-hosted really cost.

      Marketing 1 week

    First result
    2 months
    Cost
    Engineer time
    Owner
    Forward Deployed Engineering

    Smallest test

    Offer the eval to three accounts known to run Whisper.

    Why it works

    Package the eval: AssemblyAI's engineers run the customer's own audio through their self-hosted model and AssemblyAI side by side, report word error rate, diarization, latency and total cost (GPUs, engineering time, incidents), and leave a migration plan. Promote it on the Whisper compare page and to companies hiring for speech ML roles.

    Self-hosted teams are technical buyers who trust their own data over a vendor's benchmark. A free, scoped eval on their audio is the evaluation-driven motion the sales posts describe, aimed at a defined group.

    Risks and how to handle them

    • Customer's model wins on their audio

      Report it honestly and keep the relationship for a later release.

    • Too many free evals

      Qualify on volume before committing engineers.

    Evidence

    • HiringHead of GTM, Europe (London): "Europe is already one of our fastest growing regions." Calling platforms, contact center software and voice agent companies are moving workloads to AssemblyAI "for multilingual accuracy, often replacing incumbents or self-hosted models after head-to-head evals". Today the region is "covered largely from the US"; this is the first senior leader on the ground. Duties: choose which countries come first (UK & Ireland, DACH, France, Benelux, Nordics, Southern Europe), build Sales, Applied AI and Forward Deployed Engineering, handle "GDPR, the EU AI Act, data residency, DPAs" so compliance "becomes an advantage", represent AssemblyAI at conferences and meetups, and report "which languages, accents, and deployment requirements are winning or losing deals".Source
    • Your websiteSitemap (1,655 URLs): 26 /compare pages (Deepgram, Speechmatics, Gladia, Soniox, ElevenLabs and Scribe v2, OpenAI Realtime, Whisper, AWS, Google and Azure, plus LLM gateways such as OpenRouter, LiteLLM, Portkey and the Vercel, Cloudflare and Kong AI gateways), 99 /languages pages, 31 /solutions, 27 /partners, 24 /customers, about 800 blog posts and 700+ docs URLs.Source
    • HiringHead of Sales, Digital Native (San Francisco): scale "a small team of AEs and BDRs today to a larger sales org within the next year", own the land number for digital native and AI-native companies, hire the front-line managers, design territories and target account tiers, and drive "an outbound-first culture where pipeline is created, not waited for", with "custom benchmarks, competitive insight, and use-case-specific demos". The motion is "technical, product-led, and evaluation-driven": get builders to run an eval, then convert wins into production usage.Source
  • 12

    AssemblyAI has partner pages for LiveKit, Vapi, Daily, Twilio, Telnyx, Stream and Recall: the frameworks and platforms where AI-native teams build voice agents. A developer who starts from a partner's template uses whatever speech-to-text that template ships with.

    High impactMedium effort First result in 2 months Head of Sales

    First action

    Check which partner quickstarts already include AssemblyAI and which don't.

    How to measure it

    Partner-sourced accounts

    Now
    –
    Target
    +50%
    By
    2 quarters

    Steps

    1. 1

      Audit templates

      Which partner quickstarts and examples include AssemblyAI today.

      Applied AI Engineering 2 days

    2. 2

      Fill the gaps

      Open pull requests or send working examples to each partner.

      Applied AI Engineering 3 weeks

    3. 3

      Co-publish

      One joint build guide a month with a partner.

      Marketing monthly

    4. 4

      Office hours

      A joint live session per partner each quarter.

      Digital Native AI Account Executive quarterly

    5. 5

      Tag signups

      UTM and referral tags on every partner link.

      Senior GTM AI Engineer 2 days

    6. 6

      Report

      Partner-sourced accounts and spend in the weekly pipeline review.

      Head of Sales, Digital Native weekly

    First result
    2 months
    Cost
    $0 plus engineer time
    Owner
    Head of Sales, Digital Native

    Smallest test

    Open pull requests to two partner template repos.

    Why it works

    For each platform, make sure the official quickstarts and templates offer AssemblyAI with working code, co-publish one build guide a month, run a joint office hour, and tag signups that come through partner docs so partner-sourced accounts show up in the digital native pipeline.

    The Head of Sales owns a land number in a segment where builders start from frameworks. Being in the template means landing before the AE ever calls.

    Risks and how to handle them

    • Partners favor their own or rival speech-to-text

      Offer AssemblyAI as an option with clear benefits, not as a replacement.

    Evidence

    • Your websitePartner pages: LiveKit, Vapi, Daily, Twilio, Telnyx, Stream, Recall, Genesys, Amazon Connect, AWS, Google Cloud, Anthropic, OpenRouter, Zoom, Zapier, Make, n8n, Pipedream and others. Logos and stories on the site include Zoom, CallRail, Calabrio, Siro, Metaview, Ashby, LiveKit, EliseAI and Commure.Source
    • HiringHead of Sales, Digital Native (San Francisco): scale "a small team of AEs and BDRs today to a larger sales org within the next year", own the land number for digital native and AI-native companies, hire the front-line managers, design territories and target account tiers, and drive "an outbound-first culture where pipeline is created, not waited for", with "custom benchmarks, competitive insight, and use-case-specific demos". The motion is "technical, product-led, and evaluation-driven": get builders to run an eval, then convert wins into production usage.Source
    • HiringDigital Native AI Account Executive, posted three times (San Francisco, New York, London): own a book of AI-native startup accounts, generate pipeline through "personalized outreach, conferences, exclusive events, and finding warm introductions", send "custom benchmarks, competitive insights, use-case-specific demos that earn replies", and win CTOs, CPOs and VPs of AI. Expect "to be on the road 50%+ of each month".Source
  • 13

    The SF Tech Week truck named each flavor after a customer partner and collected a work email for every ticket: a customer showcase, a lead form and something people photograph, in one. The AE posts ask for "exclusive events" and 50% travel in exactly the cities where the AEs sit.

    Medium impactMedium effort First result in 2 months Digital Native AI Account Exe…

    First action

    Ask Finance for the cost and ticket count from the SF Tech Week truck.

    How to measure it

    Meetings from tickets claimed

    Now
    –
    Target
    10% of tickets
    By
    each event

    Steps

    1. 1

      Review SF

      Cost, tickets claimed, accounts reached and meetings.

      Marketing 1 day

    2. 2

      Pick the weeks

      A tech week in New York and one in London.

      Marketing 1 day

    3. 3

      Name the flavors

      Ask local customers for permission to use their names.

      Digital Native AI Account Executive 1 week

    4. 4

      Plan the stops

      Near offices of target accounts.

      Digital Native AI Account Executive 2 days

    5. 5

      Ticket form

      Name, work email and the account owner assigned automatically.

      Senior GTM AI Engineer 1 day

    6. 6

      Follow up

      An AE note within three days to every ticket at a target account.

      Digital Native AI Account Executive 1 week

    7. 7

      Share

      Customer posts and a short recap video.

      Marketing 1 day

    First result
    2 months
    Cost
    $10K-$20K per city
    Owner
    Digital Native AI Account Executive

    Smallest test

    Run one afternoon in New York near five target accounts.

    Why it works

    Run the same format during a big tech week in New York and in London, naming flavors after customers there, with each stop near offices on the target account list. Follow every ticket with a same-week note from the AE who owns that account.

    It is cheaper than a booth, makes customers the heroes (which gets them to share it), and gives AEs a reason to visit target accounts in person.

    Risks and how to handle them

    • Fun but no pipeline

      Measure meetings at target accounts, not tickets.

    • Permits and weather

      Book through an operator who handles permits; have a rain plan.

    Evidence

    • Your websiteSF Tech Week event page: an ice cream truck at three San Francisco stops on October 6, with flavors named after customer partners ("Granilla" for Granola, "Cookies & Commure", "Retell in a Nutshell", "Triple-Berry Triplex", "Trick or Tolan", "Area Cocoa"); visitors claim a ticket with their name and work email.Source
    • HiringDigital Native AI Account Executive, posted three times (San Francisco, New York, London): own a book of AI-native startup accounts, generate pipeline through "personalized outreach, conferences, exclusive events, and finding warm introductions", send "custom benchmarks, competitive insights, use-case-specific demos that earn replies", and win CTOs, CPOs and VPs of AI. Expect "to be on the road 50%+ of each month".Source
    • HiringHead of GTM, Europe (London): "Europe is already one of our fastest growing regions." Calling platforms, contact center software and voice agent companies are moving workloads to AssemblyAI "for multilingual accuracy, often replacing incumbents or self-hosted models after head-to-head evals". Today the region is "covered largely from the US"; this is the first senior leader on the ground. Duties: choose which countries come first (UK & Ireland, DACH, France, Benelux, Nordics, Southern Europe), build Sales, Applied AI and Forward Deployed Engineering, handle "GDPR, the EU AI Act, data residency, DPAs" so compliance "becomes an advantage", represent AssemblyAI at conferences and meetups, and report "which languages, accents, and deployment requirements are winning or losing deals".Source
  • 14

    The Startup Program gives 12 months of recurring credits and phases standard pricing in "with no migration and no cliff". The moment a startup's credits run down is when it decides whether to stay, and whether to sign a commitment.

    High impactLow effort First result in 1 month Head of Sales

    First action

    List program companies whose credits end in the next 90 days.

    How to measure it

    Graduates signing a commitment

    Now
    –
    Target
    30%
    By
    2 quarters

    Steps

    1. 1

      List graduates

      Credit end date, usage trend and funding for each company.

      Senior GTM AI Engineer 2 days

    2. 2

      Rank

      By usage growth and funding stage.

      Senior GTM AI Engineer 1 day

    3. 3

      Assign

      Each to an AE 60 days before credits end.

      Head of Sales, Digital Native 1 hour

    4. 4

      Commitment offer

      A simple startup-to-scale volume plan.

      Head of Sales, Digital Native 1 week

    5. 5

      Reach out

      AE call with a usage review, not a pitch.

      Digital Native AI Account Executive monthly

    6. 6

      Track

      Graduates retained, committed or churned.

      Senior GTM AI Engineer monthly

    First result
    1 month
    Cost
    $0
    Owner
    Head of Sales, Digital Native

    Smallest test

    Call the ten largest companies whose credits end next quarter.

    Why it works

    Each month, list program companies within 60 days of the end of their credits, ranked by usage growth and funding; give each to a digital native AE with its usage history, and offer a startup-to-scale commitment with volume pricing.

    These are the AI-native companies the Head of Sales wants, already in production. A timely, helpful call from an AE beats finding them later when a competitor is offering bigger credits.

    Risks and how to handle them

    • Startups feel pressured

      Lead with a usage review and a cost forecast, not a contract.

    Evidence

    • Your websiteStartup Program: pre-seed to Series A (bootstrapped welcome), building with voice and launching within six months; "12 months of recurring free credits"; "Startup discounts continue after your credits run out. Standard pricing phases in with no migration and no cliff".Source
    • HiringHead of Sales, Digital Native (San Francisco): scale "a small team of AEs and BDRs today to a larger sales org within the next year", own the land number for digital native and AI-native companies, hire the front-line managers, design territories and target account tiers, and drive "an outbound-first culture where pipeline is created, not waited for", with "custom benchmarks, competitive insight, and use-case-specific demos". The motion is "technical, product-led, and evaluation-driven": get builders to run an eval, then convert wins into production usage.Source
    • HiringSenior GTM AI Engineer (the first): a software engineer inside the revenue org who builds "the pipelines, internal tools, and agents that decide which companies we go after": "Finding every company building voice AI before they contact us", account research, first-email drafts, "Watching product usage and telling an AE when a customer is about to expand or churn", and turning product usage data in BigQuery and Looker into usage alerts, expansion triggers and account health scores. Judged on whether pipeline and revenue move.Source
  • 15

    Most speech-to-text customers send transcripts to an LLM for summaries or cleanup, often at frontier-model prices. The Voice Standard Bench shows many of those tasks run as well on open models AssemblyAI hosts, on the same API key.

    High impactMedium effort First result in 2 months Applied AI Engineering

    First action

    Pick ten customers whose products obviously summarize calls or meetings.

    How to measure it

    Accounts adopting LLM Gateway after a review

    Now
    –
    Target
    30% of reviews
    By
    2 quarters

    Steps

    1. 1

      Pick accounts

      Customers with a clear summary or extraction step.

      Applied AI Engineering 1 day

    2. 2

      Offer the review

      Through the account owner, with a data-handling note.

      Digital Native AI Account Executive 1 week

    3. 3

      Run the test

      Their prompts and sample transcripts across candidate models.

      Applied AI Engineering 2 days per account

    4. 4

      Report

      Quality, latency and cost, and the recommended model.

      Applied AI Engineering 1 day per account

    5. 5

      Switch

      Help set up LLM Gateway and monitor the first two weeks.

      Forward Deployed Engineering 1 week

    6. 6

      Publish

      An anonymized case on what customers saved.

      Marketing 1 week

    First result
    2 months
    Cost
    Engineer time
    Owner
    Applied AI Engineering

    Smallest test

    Run the review for two friendly customers.

    Why it works

    Offer every customer above a spend line a free review: with permission, test a sample of their prompts and transcripts on the bench's candidate models, show quality, speed and cost, and set up the switch on LLM Gateway. Count it as expansion revenue for the account team.

    It turns a cost saving for the customer into new revenue for AssemblyAI, and makes a second product sticky in accounts where speech-to-text is already won.

    Risks and how to handle them

    • Quality drops in production

      Monitor and keep an easy switch back.

    • Customer data sensitivity

      Use only samples the customer provides, under the existing agreement.

    Evidence

    • Your websiteVoice Standard Bench (blog, October 13, 2026): LLM Gateway "now has 50+ models", on the same API key as speech-to-text, including low-cost open models hosted on AssemblyAI's own infrastructure; the bench scores LLMs on summarization, information extraction and text correction at four transcript lengths, with cost and speed side by side. "The main finding: most voice tasks don't need frontier models, or frontier-model prices."Source
    • HiringHead of Sales, Digital Native (San Francisco): scale "a small team of AEs and BDRs today to a larger sales org within the next year", own the land number for digital native and AI-native companies, hire the front-line managers, design territories and target account tiers, and drive "an outbound-first culture where pipeline is created, not waited for", with "custom benchmarks, competitive insight, and use-case-specific demos". The motion is "technical, product-led, and evaluation-driven": get builders to run an eval, then convert wins into production usage.Source
    • HiringSenior GTM AI Engineer (the first): a software engineer inside the revenue org who builds "the pipelines, internal tools, and agents that decide which companies we go after": "Finding every company building voice AI before they contact us", account research, first-email drafts, "Watching product usage and telling an AE when a customer is about to expand or churn", and turning product usage data in BigQuery and Looker into usage alerts, expansion triggers and account health scores. Judged on whether pipeline and revenue move.Source

Big bets 3

Larger plays that can change the trajectory.

  • 16

    The 99 language pages are written in English about the language. A German contact center buyer, or a French voice agent startup, reads the pitch in English and does not see local customers.

    High impactHigh effort First result in 2 quarters Head of GTM

    First action

    Confirm the two countries with the scorecard (idea 9).

    How to measure it

    Pipeline from localized pages

    Now
    –
    Target
    25% of European pipeline
    By
    3 quarters

    Steps

    1. 1

      Confirm the countries

      From the scorecard.

      Head of GTM, Europe 1 week

    2. 2

      Choose the pages

      Home, pricing, enterprise, EU residency, top three use cases, docs quickstart.

      Marketing 1 week

    3. 3

      Translate and review

      Native speakers, with technical review.

      Marketing 6 weeks

    4. 4

      Local stories

      Two customers per country.

      Head of GTM, Europe 2 months

    5. 5

      Set hreflang

      Tell search engines which page is for which country.

      Software Engineer, Agent Experience 2 days

    6. 6

      EU endpoint default

      In every code sample on the localized pages.

      Software Engineer, Agent Experience 1 day

    7. 7

      Measure

      Signups, EU-endpoint use and pipeline by country.

      Senior GTM AI Engineer monthly

    First result
    2 quarters
    Cost
    $40K-$80K
    Owner
    Head of GTM, Europe

    Smallest test

    Translate the home and EU residency pages for one country and run local ads to them.

    Why it works

    For the two countries the scorecard picks, translate the core pages (home, pricing, enterprise, EU residency, the main use cases and the docs quickstart), add two local customer stories each, set the EU endpoint as the default in samples, and hire or contract a native-speaking content lead.

    Speechmatics and Gladia sell to the same European buyers on accents and multilingual speech. A site in the buyer's language, with local proof and residency built in, says AssemblyAI is staying.

    Risks and how to handle them

    • Translation goes stale

      Translate only pages that change slowly and set an update owner.

    • Low traffic

      Launch one country first and decide on the second by results.

    Evidence

    • Your websiteLanguage pages for German, French, Spanish and Dutch (in English): the German page promises "From call centers in Berlin to media companies in Vienna" and "cloud or self-hosted, fully compliant", but none of the four mentions EU data residency, the EU endpoint or GDPR.Source
    • HiringHead of GTM, Europe (London): "Europe is already one of our fastest growing regions." Calling platforms, contact center software and voice agent companies are moving workloads to AssemblyAI "for multilingual accuracy, often replacing incumbents or self-hosted models after head-to-head evals". Today the region is "covered largely from the US"; this is the first senior leader on the ground. Duties: choose which countries come first (UK & Ireland, DACH, France, Benelux, Nordics, Southern Europe), build Sales, Applied AI and Forward Deployed Engineering, handle "GDPR, the EU AI Act, data residency, DPAs" so compliance "becomes an advantage", represent AssemblyAI at conferences and meetups, and report "which languages, accents, and deployment requirements are winning or losing deals".Source
    • Competitor researchDeepgram (homepage title "Enterprise Voice AI: STT, TTS & Agent APIs"; headline "Voice AI built for real conversations."), Speechmatics ("speech-to-text APIs for real conversations, not clean demos ... accents, multiple speakers, and multilingual speech"), Gladia ("AI audio infrastructure ... through a single API"), Soniox ("realtime speech-to-text, text-to-speech, and translation in 60+ languages") and ElevenLabs Scribe ("Transcribe in 99 languages").Source
  • 17

    The GTM AI engineer's brief is literally "Finding every company building voice AI before they contact us." The public signals are there: job posts for voice and speech roles, products that launch with dictation, notetaking or voice agents, public repos that add speech SDKs, and partner marketplaces.

    High impactHigh effort First result in 2 quarters Senior GTM AI Engineer

    First action

    List the five public signals and test each on 20 known customers to see which would have found them early.

    How to measure it

    Opportunities sourced from map accounts

    Now
    –
    Target
    30% of new digital native pipeline
    By
    3 quarters

    Steps

    1. 1

      Pick signals

      Job posts, product launches, public repos, partner marketplaces, funding news.

      Senior GTM AI Engineer 1 week

    2. 2

      Backtest

      Would each signal have found 20 current customers before they signed up?

      Senior GTM AI Engineer 1 week

    3. 3

      Build collectors

      Weekly jobs that pull and normalize each signal.

      Senior GTM AI Engineer 4 weeks

    4. 4

      Match and score

      Company matching and a fit score with reasons.

      Senior GTM AI Engineer 3 weeks

    5. 5

      Deliver to AEs

      Top accounts in the CRM queue with the reason and an opener.

      Senior GTM AI Engineer 2 weeks

    6. 6

      Measure

      Opportunities and wins from map accounts each month.

      Head of Sales, Digital Native monthly

    7. 7

      Retire weak signals

      Drop signals that don't lead to meetings.

      Senior GTM AI Engineer quarterly

    First result
    2 quarters
    Cost
    Data sources and engineer time
    Owner
    Senior GTM AI Engineer

    Smallest test

    Backtest one signal (job posts) against 20 recent customers.

    Why it works

    Build a weekly pipeline that collects these signals, matches them to companies, scores them for fit (funding, use case, likely volume) and puts the top accounts in each AE's queue with the reason and a suggested opener tied to their use case.

    With a sales team set to grow quickly, the scarce resource is knowing where to point new reps. A ranked list with reasons makes each new AE productive in weeks, which is what the Head of Sales is asked to do.

    Risks and how to handle them

    • Noisy matches

      Show the reason with every account so AEs can judge it.

    • Terms of use for data sources

      Use public data and official APIs within their terms.

    Evidence

    • HiringSenior GTM AI Engineer (the first): a software engineer inside the revenue org who builds "the pipelines, internal tools, and agents that decide which companies we go after": "Finding every company building voice AI before they contact us", account research, first-email drafts, "Watching product usage and telling an AE when a customer is about to expand or churn", and turning product usage data in BigQuery and Looker into usage alerts, expansion triggers and account health scores. Judged on whether pipeline and revenue move.Source
    • HiringHead of Sales, Digital Native (San Francisco): scale "a small team of AEs and BDRs today to a larger sales org within the next year", own the land number for digital native and AI-native companies, hire the front-line managers, design territories and target account tiers, and drive "an outbound-first culture where pipeline is created, not waited for", with "custom benchmarks, competitive insight, and use-case-specific demos". The motion is "technical, product-led, and evaluation-driven": get builders to run an eval, then convert wins into production usage.Source
    • HiringGTM Recruiter: "our go-to-market team is growing fast"; hires Account Executives, BDRs, Customer Success, Forward Deployed Engineers and Marketers, 5-6 open roles at a time; "we're hiring a lot of people". Also open: a Finance Engineer for deal desk and sales-compensation systems, and inference engineers in the UK and North America. 15 open roles in all, 6 of them in sales.Source
  • 18

    Contact center software is named as a source of European growth, and AssemblyAI has partner pages for Genesys and Amazon Connect. Large contact centers buy speech features inside their platform, often through its marketplace.

    High impactHigh effort First result in 3 quarters Enterprise Account Executive

    First action

    Check what the Genesys and Amazon Connect partner pages offer today and what a marketplace listing requires.

    How to measure it

    Partner-sourced enterprise deals

    Now
    –
    Target
    5 in three quarters
    By
    3 quarters

    Steps

    1. 1

      Review the integrations

      What exists today for Genesys and Amazon Connect.

      Forward Deployed Engineering 1 week

    2. 2

      Listing requirements

      What each marketplace needs to list.

      Enterprise Account Executive 1 week

    3. 3

      Build the gap

      A ready integration and setup guide.

      Forward Deployed Engineering 6 weeks

    4. 4

      Reference customer

      One contact center live on each platform.

      Enterprise Account Executive 2 months

    5. 5

      Co-sell plan

      With each platform's partner team.

      Enterprise Account Executive 1 month

    6. 6

      Target list

      Large contact centers on each platform, US and Europe.

      Senior GTM AI Engineer 2 weeks

    7. 7

      Run the motion

      Joint calls and a quarterly review with each partner.

      Enterprise Account Executive ongoing

    First result
    3 quarters
    Cost
    Integration work and marketplace fees
    Owner
    Enterprise Account Executive

    Smallest test

    Get one existing contact center customer to agree to be the reference for one platform.

    Build the prospect list. The list compiled from public sources, verified, deduplicated and delivered to your CRM. Usually live within days.

    Book a working session

    Why it works

    Make AssemblyAI available as a ready integration for Genesys and Amazon Connect customers (marketplace listing, setup guide, a reference customer), co-sell with each platform's partner team, and give the founding Enterprise AE a target list of large contact centers on those platforms.

    The founding Enterprise AE needs a way into large companies that does not depend on each one building from scratch; the contact center platform is that route, in the US and in Europe.

    Risks and how to handle them

    • Long platform cycles

      Start with the integration that has an existing customer pulling for it.

    • Channel conflict

      Agree deal registration rules with each partner.

    Evidence

    • Your websitePartner pages: LiveKit, Vapi, Daily, Twilio, Telnyx, Stream, Recall, Genesys, Amazon Connect, AWS, Google Cloud, Anthropic, OpenRouter, Zoom, Zapier, Make, n8n, Pipedream and others. Logos and stories on the site include Zoom, CallRail, Calabrio, Siro, Metaview, Ashby, LiveKit, EliseAI and Commure.Source
    • HiringEnterprise Account Executive, described as "AssemblyAI's founding Enterprise Account Executive": own a book of strategic enterprise accounts, lead model evaluations and win enterprise engineering and product leaders, through procurement, legal and security review.Source
    • HiringHead of GTM, Europe (London): "Europe is already one of our fastest growing regions." Calling platforms, contact center software and voice agent companies are moving workloads to AssemblyAI "for multilingual accuracy, often replacing incumbents or self-hosted models after head-to-head evals". Today the region is "covered largely from the US"; this is the first senior leader on the ground. Duties: choose which countries come first (UK & Ireland, DACH, France, Benelux, Nordics, Southern Europe), build Sales, Applied AI and Forward Deployed Engineering, handle "GDPR, the EU AI Act, data residency, DPAs" so compliance "becomes an advantage", represent AssemblyAI at conferences and meetups, and report "which languages, accents, and deployment requirements are winning or losing deals".Source

Wildcards 3

Unexpected experiments, each with a cheap way to test it.

  • 19

    The Head of GTM, Europe must report "which languages, accents, and deployment requirements are winning or losing deals". Most vendors keep that list private. Published, it is the most honest multilingual roadmap in the market.

    Medium impactLow effort First result in 2 months Head of GTM

    First action

    Draft the first note from the last quarter's lost European evaluations.

    How to measure it

    Audio samples submitted after each note

    Now
    –
    Target
    50
    By
    each quarter

    Steps

    1. 1

      Collect win/loss

      From European and multilingual evaluations, anonymized.

      Head of GTM, Europe 1 week

    2. 2

      Research input

      What is being worked on and what can be said publicly.

      Research 2 days

    3. 3

      Write it

      Wins, losses, examples and next steps.

      Marketing 2 days

    4. 4

      Submission form

      A way for customers to send hard audio under clear terms.

      Software Engineer, Agent Experience 2 days

    5. 5

      Publish and share

      Blog, newsletter and AE outreach in Europe.

      Marketing 1 day

    6. 6

      Follow up

      Tell submitters when their language improves.

      Research quarterly

    First result
    2 months
    Cost
    $0
    Owner
    Head of GTM, Europe

    Smallest test

    Publish a short note on two languages that improved and one still in progress.

    Why it works

    Each quarter, publish which languages and accents improved, where AssemblyAI still loses evaluations (with anonymized examples), and what the research team is working on; invite customers to send hard audio for the next release.

    Multilingual buyers distrust "99 languages" claims. Saying where the model is weaker, and what is being fixed, earns trust and brings in exactly the audio research needs.

    Risks and how to handle them

    • Competitors use the losses in sales

      Publish losses only alongside what is being fixed and the date.

    Evidence

    • HiringHead of GTM, Europe (London): "Europe is already one of our fastest growing regions." Calling platforms, contact center software and voice agent companies are moving workloads to AssemblyAI "for multilingual accuracy, often replacing incumbents or self-hosted models after head-to-head evals". Today the region is "covered largely from the US"; this is the first senior leader on the ground. Duties: choose which countries come first (UK & Ireland, DACH, France, Benelux, Nordics, Southern Europe), build Sales, Applied AI and Forward Deployed Engineering, handle "GDPR, the EU AI Act, data residency, DPAs" so compliance "becomes an advantage", represent AssemblyAI at conferences and meetups, and report "which languages, accents, and deployment requirements are winning or losing deals".Source
    • Your websitePricing (read October 18, 2026): Universal-3.5 Pro $0.21/hr, "works across 18 languages, with native code switching"; Universal-2 $0.15/hr, "Supports 99 languages"; add-ons such as keyterms prompting $0.05/hr, speaker diarization $0.02/hr and Medical Mode $0.15/hr; "Start free, pay-as-you-go after that – no commitments required".Source
    • Competitor researchDeepgram (homepage title "Enterprise Voice AI: STT, TTS & Agent APIs"; headline "Voice AI built for real conversations."), Speechmatics ("speech-to-text APIs for real conversations, not clean demos ... accents, multiple speakers, and multilingual speech"), Gladia ("AI audio infrastructure ... through a single API"), Soniox ("realtime speech-to-text, text-to-speech, and translation in 60+ languages") and ElevenLabs Scribe ("Transcribe in 99 languages").Source
  • 20

    A vendor's own benchmark is always suspected of favoring the vendor. The Voice Standard Bench scores LLMs, not speech models, which makes it easier for AssemblyAI to run it fairly for the whole industry.

    Medium impactMedium effort First result in 3 months Marketing

    First action

    Write the submission rules and invite two model providers.

    How to measure it

    Sites linking to the bench

    Now
    –
    Target
    50
    By
    6 months

    Steps

    1. 1

      Publish the method

      Tasks, data, scoring and how to reproduce.

      Research 2 weeks

    2. 2

      Submission rules

      Who can submit and how results are verified.

      Marketing 1 week

    3. 3

      Invite providers

      Five model providers on LLM Gateway.

      Partnerships 2 weeks

    4. 4

      Monthly run

      Re-score every model and date the results.

      Research monthly

    5. 5

      Customer option

      Run the bench on a customer's own transcripts.

      Applied AI Engineering 3 weeks

    6. 6

      Track citations

      Links, mentions and AI assistant answers.

      Marketing monthly

    First result
    3 months
    Cost
    Compute costs
    Owner
    Marketing

    Smallest test

    Publish the method and invite one outside provider.

    Why it works

    Publish the method and test sets, accept submissions from model providers, re-run monthly with dated results, and add a "bring your own transcripts" option for customers. Keep AssemblyAI's hosted models scored by the same rules.

    An independent-looking reference becomes a page others link to, that AI assistants cite, and that brings model providers into partnership conversations, while every row still points to LLM Gateway.

    Risks and how to handle them

    • Seen as biased

      Publish everything needed to reproduce the results.

    • Providers dispute scores

      A documented appeal and re-run process.

    Evidence

    • Your websiteVoice Standard Bench (blog, October 13, 2026): LLM Gateway "now has 50+ models", on the same API key as speech-to-text, including low-cost open models hosted on AssemblyAI's own infrastructure; the bench scores LLMs on summarization, information extraction and text correction at four transcript lengths, with cost and speed side by side. "The main finding: most voice tasks don't need frontier models, or frontier-model prices."Source
    • programmable.infoSite scan: Chrome UX Report mobile LCP p75 5,175 ms, 39% slower than a year earlier (INP 129 ms, CLS 0.03); the homepage runs a live "Try your own voice" recorder and code sample above the fold. The assemblyai npm package has about 289,000 weekly downloads (Deepgram's packages about 922,000); PyPI package last released September 24, 2026; 71 public GitHub repos.Source
  • 21

    AssemblyAI has a "Dictation for Coding Agents" use case and a new Dictation API built to return text "ready to send". The people choosing APIs spend their day prompting Claude Code and Cursor; many would rather speak the prompt.

    Medium impactLow effort First result in 2 months Software Engineer

    First action

    Build a prototype that dictates into the terminal and try it with five engineers internally.

    How to measure it

    Weekly active users

    Now
    –
    Target
    2,000
    By
    3 months

    Steps

    1. 1

      Prototype

      Push-to-talk dictation into the terminal through the Dictation API.

      Software Engineer, Agent Experience 1 week

    2. 2

      Internal test

      Five engineers for two weeks.

      Software Engineer, Agent Experience 2 weeks

    3. 3

      Polish

      Install in one command, a free limit and an API-key prompt.

      Software Engineer, Agent Experience 1 week

    4. 4

      Open-source it

      Public repo with a clear licence and the integration explained.

      Software Engineer, Agent Experience 1 day

    5. 5

      Launch

      Developer communities, the blog and the agent docs.

      Marketing 1 week

    6. 6

      Measure

      Installs, weekly users and signups that create a key.

      Senior GTM AI Engineer monthly

    First result
    2 months
    Cost
    API usage within the free limit
    Owner
    Software Engineer, Agent Experience

    Smallest test

    Share the internal prototype with one developer community and count installs.

    Why it works

    Ship a small open-source tool (a CLI or editor extension) that dictates prompts into coding agents using the Dictation API, free up to a generous limit, with the source code as a reference integration. Promote it where the Agent Experience work already reaches developers.

    It puts AssemblyAI in the developer's hands every day, as a user rather than a vendor, and the source is itself a sample that agents learn from.

    Risks and how to handle them

    • Becomes a support burden

      Keep it small and community-supported, with clear limits.

    Evidence

    • Your websitellms.txt already exists and tells agents to "prefer the Markdown sources linked in this file over crawling the HTML marketing site"; it points coding assistants to "Build with AI coding agents" and a docs MCP server. The use-case pages include "Dictation for Coding Agents".Source
    • HiringSoftware Engineer, Agent Experience: "More and more developers don't read our docs or click through our dashboard. Their coding agent does." Owns the skills, MCP server, llms.txt, code samples and docs an agent reads; the north star is that "an agent should go from zero to a working, production-grade AssemblyAI integration on the first try". Builds evals that run real coding agents on real integration tasks and tracks "success rate, time to first working call, and activation from agent-driven signups".Source
    • programmable.infoSite scan: Chrome UX Report mobile LCP p75 5,175 ms, 39% slower than a year earlier (INP 129 ms, CLS 0.03); the homepage runs a live "Try your own voice" recorder and code sample above the fold. The assemblyai npm package has about 289,000 weekly downloads (Deepgram's packages about 922,000); PyPI package last released September 24, 2026; 71 public GitHub repos.Source

03

How Programmable can help

Where an idea in this playbook is something Programmable builds or runs, it's listed here with what we'd do. Everything else is yours to run.

Have it built

ProgrammableAgency

1 of the 21 ideas are things Programmable builds and runs for its customers: pages, outreach, prospect lists, booking flows and monitors. Pick one, or send the whole list.

Prospect lists1The list compiled from public sources, verified, deduplicated and delivered to your CRM.

A 30-minute session with the engineer who would build it, not a sales call. You leave with a plan whether or not you hire us.

Book a working session

Usually live within days.

Tools for the rest

ProgrammableTalk

Fill the events in this plan

These ideas are events. Talk gets registrants for them and tells you who's coming and what their company does before the day.

Plan an event (opens programmable.talk in a new tab)
ProgrammableDigital

Show it, don't just describe it

These ideas need something to show: a demo, a video, an animation, graphics or a deck. Digital turns screenshots or the site into all of them, in the company's own brand.

Make a demo (opens programmable.digital in a new tab)
Website review of assemblyai.com More example playbooks and reviews

More growth playbooks

Other companies we have researched, each read from what they are hiring for.

  • SmartSites growth strategyA New Jersey-founded digital marketing agency: SEO and AI search, PPC (Google, Meta, ChatGPT Ads and more), email and SMS, social media…21 growth ideas
  • Posh growth strategyA social events platform where organizers create, sell and grow IRL events and communities, and people discover parties, festivals, run…21 growth ideas
  • Morpho growth strategyAn open, noncustodial credit network: Morpho Blue (variable rate), Morpho Midnight (fixed rate, fixed term) and Morpho Vaults, used by…20 growth ideas
  • GTCL Wellness growth strategyA Singapore wellness company (GTCL Pte Ltd, founded 2013) behind Anaya (stingless bee honey, propolis and anti-mould leather care), Byra…21 growth ideas
  • RELEX Solutions growth strategyAn AI-native, unified supply chain and retail planning platform: demand forecasting, replenishment, inventory, merchandising, pricing…22 growth ideas
  • OLAPLEX growth strategyPrestige hair care built on patented bond-building technology: treatments, shampoos, conditioners, masks, stylers and new lash, brow and…21 growth ideas
  • Verve growth strategyVerve Group is an omnichannel advertising platform: intent data from search and AI chat, premium mobile, CTV and web supply, a…21 growth ideas
  • Weee! growth strategyAmerica's largest online Asian supermarket, with free next-day grocery delivery, a Latino store (MásGusto), a local marketplace for…21 growth ideas

See every playbook and website review

Independent research. Programmable wrote this playbook from public sources, as they stood on October 18, 2026, and cites them throughout. AssemblyAI did not commission, review or endorse it, and Programmable is not affiliated with AssemblyAI. Company and product names are trademarks of their owners, used only to identify them. The ideas are suggestions and opinion, not professional, legal or financial advice, and not statements about how AssemblyAI runs its business.

Something wrong, or do you represent AssemblyAI? Email [email protected] and we will correct it or take it down.

30/60/90 roadmap

What to focus on in each month, and the ideas that carry it.

  1. Days 1 to 308 ideas

    EU residency page and language-page block, one set of scale numbers, benchmark outreach, old pages, the first usage alert and mobile speed; start the agent-signup channel and the startup graduation handoff.

  2. Days 31 to 6010 ideas

    The coding-agent integration score, Europe's country scorecard, the European buyer's pack, the self-hosted Whisper eval, partner templates, events in New York and London, LLM Gateway reviews, and the wildcards.

  3. Days 61 to 903 ideas

    Commit to the big bets with the first 60 days of data: localized sites for two countries, the voice-builder map and the contact center platforms.

The research behind the plan

3 research areas, each summarised in a few lines. Open any of them for the full detail.

Customers

What customers say

Customers value accuracy on real, messy audio and speaker diarization. The most cited problem: users of voice agents "repeating themselves".

  • Fast time to first API call and good SDKs and docs
  • Cost of LLM post-processing at frontier-model prices
  • In their words: “voice AI infrastructure”

Market

Competitive landscape

6 competitors compared: Deepgram, Speechmatics, Gladia, Soniox and 2 more.

  • Gap to exploit: Proof of how well coding agents integrate each API
  • Deepgram: Positions on speed and the full stack more than multilingual accuracy
  • Speechmatics: Smaller developer community

Hiring signals

Six of AssemblyAI's 15 open roles are in sales, and a GTM Recruiter is hiring "a lot of people".

  • Head of Sales, Digital Native: Scale a small AE and BDR team into a multi-layer org within a year; outbound-first with custom benchmarks…
  • Digital Native AI Account Executive (San Francisco, New York, London): Three openings; personalized outreach, conferences and exclusive events…
  • Enterprise Account Executive: The founding enterprise seller; procurement, legal and security review. Ideas 2, 10, 18.

From the team that wrote this playbook

All products
  • ProgrammableAgencyWorking AI systems in days, not quarters (opens programmable.agency in a new tab)
  • ProgrammableInfoAn agent that finds, researches and asks first (opens programmable.info in a new tab)
  • ProgrammableDigitalScreenshots in, an animated product demo out (opens programmable.digital in a new tab)
  • ProgrammableWatchKnow the day a supplier falls out of compliance (opens programmable.watch in a new tab)
  • ProgrammableTalkGet registrants for your events, and know who’s coming (opens programmable.talk in a new tab)
  • ProgrammablePageText what you need, your customer gets a page that works (opens programmable.page in a new tab)
  • ProgrammableCommunityForums with labelled AI members, never empty (opens programmable.community in a new tab)
ProgrammableMarketing
ExamplesLeaderboardsProductsTermsPrivacyCookiesDPALegalContact

© 2026 AI-Powered Growth Intelligence