ProgrammableMarketing/upliftai.org (opens in a new tab)
PDF

Views

  • Plan
  • Roadmap
  • Research

On this page

  • The brief
  • Start here
  • Idea portfolio
  • Effort vs impact
  • This week · 8
  • This quarter · 11
  • Big bets · 2
  • Wildcards · 1
  • How we can help
Generated Nov 10, 2026
  1. Programmable
  2. Playbooks
  3. Uplift AI
Uplift AI logo

growth playbook · Nov 10, 2026

upliftai.org (opens in a new tab)

Uplift AI growth playbook

Diagnosis

Uplift AI is a YC S25 company with a $3.5 million seed round and real proof: Khan Academy has published thousands of Urdu videos with its voices, and more than 1,000 developers build on its APIs. Its newest move is AI calling for Pakistani businesses, led by COD order confirmation. Almost none of that proof is on the site, though.

There are no logos, case studies or result numbers. Industry pages all play the same three sample calls, prices live in a homepage anchor, and the e-commerce offer asks merchants to integrate an API and wait 30 days, while keypad robocall apps install in minutes on Shopify and have hundreds of reviews. Real visitors also fail Core Web Vitals despite a 99 lab score.

This playbook first puts proof, prices and sector-specific samples on the site and fixes speed and the sitemap. Then it builds a self-serve Shopify path for COD confirmation, template agents for small businesses, self-serve concurrency and new language and sector pages. The big bets are a community language programme built on the tuning platform now being hired for, and a voice hotline for people who cannot read.

Growth Score

6/10

Biggest Opportunity

COD confirmation in Pakistan is a proven, paid-for job that merchants already buy as keypad robocalls on Shopify. A real Urdu conversation that fixes addresses before dispatch is a clear upgrade, if merchants can install it in minutes.

Biggest Risk

Global voice platforms already sell Urdu text-to-speech, and cheap local IVR apps own COD confirmation, so Uplift has to prove its quality and outcomes in public before buyers treat it as interchangeable.

01

Start here

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

  1. 1

    Give every industry page a sample call from its own sector

    Write the four missing call scripts from each page's own "Where to start" section.

    High impact First result in 2 weeks
  2. 2

    Put the Khan Academy story on the site: the strongest proof Uplift has is only in the press

    Ask the Khan Academy contact for approval to publish the numbers and a quote.

    High impact First result in 3 weeks
  3. 3

    Publish a real pricing page that turns plan minutes into calls and voiceovers

    Write the per-plan translation: calls, voiceover minutes and hours.

    Medium impact First result in 3 weeks

02

Your idea portfolio

22 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.

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

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 · 2Big bets · 6Fill-ins · 10Reconsider · 4
Effort →

Showing 22 of 22 ideas

This week 8

Quick wins with a clear first action.

  • 02

    The healthcare, telecom and e-commerce pages all play the same three samples: a courier follow-up, an e-commerce order confirmation and a real estate site visit. A clinic manager on the healthcare page never hears an appointment reminder, and a telecom buyer never hears a billing call. The sample call is the product, and right now it is the wrong one on most pages.

    High impactLow effort First result in 2 weeks CEO

    First action

    Write the four missing call scripts from each page's own "Where to start" section.

    How to measure it

    Industry page call bookings

    Now
    –
    Target
    +30%
    By
    60 days

    Steps

    1. 1

      Write scripts

      One per sector, from the use cases each page already lists.

      CEO 1 day

    2. 2

      Record in Studio

      Use the voices a customer in that sector would pick.

      CEO 1 day

    3. 3

      Add transcripts

      Show the transcript and outcome card under each player.

      Founding Design Engineer 1 day

    4. 4

      Reorder samples

      Sector sample first, the others below.

      Founding Design Engineer 2 hours

    5. 5

      Track plays

      Plays and booking clicks per page.

      CTO 2 hours

    6. 6

      Refresh

      Swap in a real, consented customer call once one exists.

      CEO quarterly

    First result
    2 weeks
    Cost
    $0, using Studio
    Owner
    CEO

    Smallest test

    Add a healthcare appointment sample this week and compare bookings with the month before.

    Why it works

    Record one realistic sample per sector in Studio: an appointment confirmation and reschedule for healthcare, a payment reminder with a promise-to-pay for financial services, a package or balance question for telecom, a missed-delivery reschedule for couriers. Put each one first on its page, with the transcript and outcome card underneath, and keep the other samples below.

    Buyers judge voice AI by listening. A sample that matches their own call cuts the doubt before "Book a call" and makes the page feel built for them.

    Risks and how to handle them

    • Samples sound staged

      Include a realistic interruption or a code-switch to English, as real callers do.

    Evidence

    • Your websiteIndustry pages (healthcare, telecom and e-commerce, checked November 10, 2026): each plays the same three sample calls (courier delivery follow-up, e-commerce order confirmation, real estate site visit), so the healthcare and telecom pages have no sample from their own sector; no named customers or metrics; each promises production within 30 days.Source
    • Your websiteAI calling page: "AI calling in Pakistan. Built around your customers."; Urdu-first, including "the English and Punjabi words customers mix in"; inbound and outbound calls with a recording, transcript, summary and outcome for each; "Starts at $0.04 per minute", final pricing through a "Talk to Team" call; no customers named and no mention of WhatsApp.Source
  • 03

    Seed coverage says Khan Academy has published 2,500 Urdu educational videos with Uplift's voices. Yet the homepage, the Urdu page and every industry page show no logos, testimonials or case studies at all. A buyer comparing Uplift with a global voice platform never sees the one fact that settles quality.

    High impactLow effort First result in 3 weeks CEO

    First action

    Ask the Khan Academy contact for approval to publish the numbers and a quote.

    How to measure it

    Signup conversion

    Now
    –
    Target
    +15%
    By
    60 days

    Steps

    1. 1

      Get approval

      Numbers, logo use and a quote.

      CEO 1 week

    2. 2

      Write the story

      Problem, workflow, numbers, time saved.

      CEO 2 days

    3. 3

      Add clips

      Two short before-and-after segments.

      Founding Design Engineer 1 day

    4. 4

      Build the page

      A /customers page plus a logo strip on the homepage and Urdu page.

      Founding Design Engineer 2 days

    5. 5

      Add to the sitemap

      So search and AI assistants find it.

      CTO 1 hour

    6. 6

      Line up the next story

      The agriculture deployment or the first calling customer.

      CEO 1 month

    First result
    3 weeks
    Cost
    $0
    Owner
    CEO

    Smallest test

    Add one line with the 2,500-video figure and a press link under the homepage headline.

    Why it works

    With Khan Academy's approval, publish a short case study: the problem (Urdu video at scale), the workflow, the numbers (from 104 videos at launch to 2,500), the time saved, and two before-and-after clips. Add a logo strip and one quote to the homepage and the Urdu TTS page, and a line on developer adoption (more than 1,000 developers).

    Third-party proof from a name buyers trust is worth more than any feature list, and it answers "is the Urdu good enough?" before the free trial.

    Risks and how to handle them

    • Approval takes time

      Publish only what is already public in the press first, then expand.

    Evidence

    • PressSeed round coverage: $3.5 million led by Y Combinator and Indus Valley Capital; Khan Academy has used the technology to publish 2,500 educational videos in Urdu; a global crop-science company is rolling out voice tools for farmers; more than 1,000 developers build on the APIs; next: more languages, especially in MENA, and education, public-sector and enterprise integrations.Source
    • PressKhan Academy launch coverage: 104 Urdu-dubbed videos at the time with thousands more planned; the company says dubbing is 10 times faster than traditional voiceover.Source
    • Your websiteHomepage: H1 "Finally, AI Voices that sound like Pakistan"; "We make world-leading voice models for regional languages"; three products (Text to Speech Studio, Developer API, AI Calling in "Early Preview"); "100+ AI Voices"; plans Free $0 (10 min TTS, 10 min AI calling), Starter $5/month (100 min TTS, 125 min calling), Pro $50 ("Popular"; 25 hours, 1,250 min), Growth $300 (200 hours, 7,500 min) and custom Enterprise with on-prem; a "Call me now" phone demo and two sample calls (a property-expo invitation, and a sales call about a crop disease offering a free expert visit); no customer logos, testimonials or case studies on the page.Source
  • 04

    Prices live only in a homepage anchor (/#pricing). The sitemap has no pricing page, the Urdu TTS page shows no prices, and the AI calling page says "Starts at $0.04 per minute" but sends buyers to a call for final pricing. The plans are actually simple: Starter's 125 calling minutes and Pro's 1,250 both work out to $0.04 a minute. Nobody has done that math for the buyer.

    Medium impactLow effort First result in 3 weeks CEO

    First action

    Write the per-plan translation: calls, voiceover minutes and hours.

    How to measure it

    New paid subscriptions

    Now
    –
    Target
    +20%
    By
    60 days

    Steps

    1. 1

      Do the math

      Calls per plan at a typical 1-2 minute confirmation call.

      CEO 2 hours

    2. 2

      Write the page

      Plans, translation rows, licence, overage, FAQ.

      CEO 1 day

    3. 3

      Design it

      Reuse the homepage plan cards.

      Founding Design Engineer 1 day

    4. 4

      Link it

      From the Urdu, AI calling and industry pages and the docs.

      CTO 2 hours

    5. 5

      Add markup

      Sitemap entry and product price markup.

      CTO 2 hours

    6. 6

      Review monthly

      Which plan people pick, and adjust the translation rows.

      CEO monthly

    First result
    3 weeks
    Cost
    $0
    Owner
    CEO

    Smallest test

    Add the calls-per-plan row to the homepage pricing block today.

    Why it works

    Create /pricing with the five plans, plus a translation row for each: about how many confirmation calls, voiceover minutes or audiobook hours each plan covers. Add what the commercial licence includes, what overage costs, and when Enterprise makes sense. Link it from the Urdu, AI calling and industry pages and add it to the sitemap.

    Small businesses and creators buy when they can see the cost of their own job. A crawlable pricing page also answers "Urdu TTS price" searches and AI-assistant questions.

    Risks and how to handle them

    • Showing prices invites comparison with $2 keypad apps

      Show the cost per confirmed order and per avoided return, not just per minute.

    Evidence

    • Your websiteHomepage: H1 "Finally, AI Voices that sound like Pakistan"; "We make world-leading voice models for regional languages"; three products (Text to Speech Studio, Developer API, AI Calling in "Early Preview"); "100+ AI Voices"; plans Free $0 (10 min TTS, 10 min AI calling), Starter $5/month (100 min TTS, 125 min calling), Pro $50 ("Popular"; 25 hours, 1,250 min), Growth $300 (200 hours, 7,500 min) and custom Enterprise with on-prem; a "Call me now" phone demo and two sample calls (a property-expo invitation, and a sales call about a crop disease offering a free expert visit); no customer logos, testimonials or case studies on the page.Source
    • Your websiteAI calling page: "AI calling in Pakistan. Built around your customers."; Urdu-first, including "the English and Punjabi words customers mix in"; inbound and outbound calls with a recording, transcript, summary and outcome for each; "Starts at $0.04 per minute", final pricing through a "Talk to Team" call; no customers named and no mention of WhatsApp.Source
    • Your websiteSitemap (checked November 10, 2026): 24 URLs, all with the same lastmod; 12 are signed-in app screens under /app/ (studio, calling, billing, team, API keys and others); one language page (Urdu), six industry pages and the AI calling page; no blog, pricing page, case studies, or agriculture or education pages.Source
  • 05

    Half of the 24 URLs in upliftai.org's sitemap are signed-in app screens: studio, calling, billing, team, API keys, webhooks and others. All 24 share the same lastmod, so the date tells search engines nothing. Crawlers spend their visits on login walls, while the pages that sell are a minority of the map.

    Medium impactLow effort First result in 2 weeks CTO

    First action

    Filter /app/ routes out of the sitemap generator.

    How to measure it

    Marketing pages indexed

    Now
    –
    Target
    All public pages
    By
    30 days

    Steps

    1. 1

      Filter routes

      Exclude /app/* from the sitemap.

      CTO 1 hour

    2. 2

      Add noindex

      On signed-in screens.

      CTO 1 hour

    3. 3

      Fix lastmod

      Use each page's real edit date.

      CTO 2 hours

    4. 4

      Resubmit

      In Search Console.

      CTO 15 minutes

    5. 5

      Check coverage

      Indexed pages after two weeks.

      CTO 2 weeks

    6. 6

      Keep it current

      Add each new public page at launch.

      CTO ongoing

    First result
    2 weeks
    Cost
    $0
    Owner
    CTO

    Smallest test

    Ship the sitemap filter today.

    Why it works

    Remove /app/* from the sitemap, mark those routes noindex, and set real lastmod dates on the marketing pages. Add the new pricing, customer and language pages as they ship, so the sitemap becomes a list of what Uplift wants found.

    A clean sitemap with honest dates helps search engines and AI crawlers focus on the pages that convert, which matters for a site with only a dozen public pages.

    Risks and how to handle them

    • Deep links to the app stop showing in search

      Keep "Login" and "Start Free" links on public pages; nobody needs billing screens in search results.

    Evidence

    • Your websiteSitemap (checked November 10, 2026): 24 URLs, all with the same lastmod; 12 are signed-in app screens under /app/ (studio, calling, billing, team, API keys and others); one language page (Urdu), six industry pages and the AI calling page; no blog, pricing page, case studies, or agriculture or education pages.Source
  • 06

    PageSpeed gives the homepage a lab score of 99, but real visitors fail Core Web Vitals: field LCP is 3,552 ms, INP is 443 ms, and time to first byte is 1,737 ms. Real-visitor LCP got 20% slower than the previous period. The LCP element is just the headline text, yet the page ships 1,021 KB of JavaScript (266 KB unused) with 750 ms of render-blocking requests. Visitors on mobile networks wait for the headline, and voice-sample taps feel slow.

    Medium impactMedium effort First result in 1 month CTO

    First action

    Check why the first byte takes 1.7 seconds for real visitors and turn on edge caching for static marketing pages.

    How to measure it

    Field Core Web Vitals

    Now
    –
    Target
    LCP under 2.5 s and INP under 200 ms
    By
    60 days

    Steps

    1. 1

      Edge cache

      Serve public pages from the CDN cache.

      CTO 1 day

    2. 2

      Static headline

      Render the H1 server-side without waiting for scripts.

      CTO 1 day

    3. 3

      Defer the gallery

      Load the 100+ voice cards and players on scroll or tap.

      Founding Design Engineer 3 days

    4. 4

      Split JavaScript

      Drop the 266 KB that the homepage doesn't use.

      CTO 3 days

    5. 5

      Remove blocking

      Inline critical CSS, preload the font.

      CTO 1 day

    6. 6

      Watch field data

      Track the 28-day field numbers.

      CTO monthly

    First result
    1 month
    Cost
    Engineering time
    Owner
    CTO

    Smallest test

    Turn on edge caching and compare time to first byte after a week.

    Why it works

    Cache the marketing pages at the edge to cut the 1.7-second first byte, render the headline as static HTML, defer the voice gallery and players until interaction, split unused JavaScript and remove render-blocking requests. Re-test with field data after 28 days.

    First byte and headline speed decide whether a mobile visitor stays long enough to hear a voice. Passing Core Web Vitals also helps search ranking.

    Risks and how to handle them

    • Deferring players delays the first sample

      Preload only the first featured voice.

    Evidence

    • PageSpeedHomepage tested November 10, 2026: lab score 99, but real visitors fail Core Web Vitals, with field LCP 3,552 ms and INP 443 ms and a time to first byte of 1,737 ms; the LCP element is the H1 text; 1,021 KB of JavaScript (266 KB unused) and 750 ms of render-blocking requests; real-visitor LCP got 20% slower than the previous period.Source
  • 07

    Step one in the docs is "Visit upliftai.org to generate your API key", with no link to the key page (which exists at /app/developer/api-keys) and no word on the free allowance. More than 1,000 developers already build on the APIs. Every one of them had to find the key on their own before hearing a single word of Urdu from their own code.

    Medium impactLow effort First result in 2 weeks CTO

    First action

    Change the docs line to a direct link to /app/developer/api-keys.

    How to measure it

    Day-1 API activation

    Now
    –
    Target
    +25%
    By
    45 days

    Steps

    1. 1

      Direct link

      Docs button to the key page.

      CTO 1 hour

    2. 2

      State the allowance

      Free minutes next to the button.

      CTO 1 hour

    3. 3

      Pre-fill the example

      Key and voice ID in the first snippet for signed-in users.

      CTO 2 days

    4. 4

      Close the loop

      Success message and the next two tutorials.

      Founding Design Engineer 1 day

    5. 5

      Track first calls

      Time from signup to first successful request.

      CTO 1 day

    6. 6

      Email stuck users

      A key but no call after 24 hours gets a one-line help email.

      CEO ongoing

    First result
    2 weeks
    Cost
    $0
    Owner
    CTO

    Smallest test

    Swap the line for a direct link today and compare key creation the following week.

    Why it works

    Replace the instruction with a button to the API key page, show the free allowance next to it, and pre-fill the first cURL example with the signed-in user's key and a default voice. End the quickstart with "you just generated Urdu speech" and links to the WhatsApp and LiveKit tutorials.

    Developer activation is mostly about time to first success. Removing a search step and a blank key field gets more signups to a working call.

    Risks and how to handle them

    • Showing keys in docs pages

      Show the key only to the signed-in owner, masked by default.

    Evidence

    • Your websiteDeveloper docs: REST and WebSocket text-to-speech with "~150-200ms Latency"; tutorials for a WhatsApp bot and a LiveKit voice agent; step one is "Visit upliftai.org to generate your API key", with no direct link to the key page; no pricing or free allowance on the page.Source
    • Your websiteSitemap (checked November 10, 2026): 24 URLs, all with the same lastmod; 12 are signed-in app screens under /app/ (studio, calling, billing, team, API keys and others); one language page (Urdu), six industry pages and the AI calling page; no blog, pricing page, case studies, or agriculture or education pages.Source
    • PressSeed round coverage: $3.5 million led by Y Combinator and Indus Valley Capital; Khan Academy has used the technology to publish 2,500 educational videos in Urdu; a global crop-science company is rolling out voice tools for farmers; more than 1,000 developers build on the APIs; next: more languages, especially in MENA, and education, public-sector and enterprise integrations.Source
  • 08

    The Urdu page says "Free-tier output does not include a commercial license" and that commercial use depends on the subscription, but it doesn't name the plan that includes it. The page targets YouTube voiceovers, social video, lessons and business announcements, which are mostly commercial uses. The licence is the clearest reason for a creator to pay $5, and it is buried in an FAQ.

    Medium impactLow effort First result in 3 weeks Founding Design Engineer

    First action

    Confirm which plans include the commercial licence and put it on the plan cards.

    How to measure it

    Free-to-paid conversion in Studio

    Now
    –
    Target
    +30%
    By
    60 days

    Steps

    1. 1

      Confirm the terms

      Which plan includes commercial use.

      CEO 1 day

    2. 2

      Update plan cards

      A "Commercial licence" line on paid plans.

      Founding Design Engineer 2 hours

    3. 3

      Design the prompt

      Two-option choice at download for free users.

      Founding Design Engineer 2 days

    4. 4

      Ship it

      Behind a flag for half of free users.

      CTO 2 days

    5. 5

      Add receipts

      Licence line on paid downloads.

      CTO 1 day

    6. 6

      Compare

      Upgrade rate with and without the prompt.

      CEO 4 weeks

    First result
    3 weeks
    Cost
    $0
    Owner
    Founding Design Engineer

    Smallest test

    Add the licence line to the plan cards and the Urdu page FAQ this week.

    Why it works

    State on the pricing cards that Starter and above include a commercial licence. In Studio, when a free user clicks Download MP3, show a short choice: "Personal use" or "Publishing on YouTube, social or for a business? Starter, $5/month, includes the commercial licence." Add a licence line to paid users' download receipts.

    The upgrade prompt appears at the moment of need and is honest. Creators who plan to monetise want the licence, and $5 is an easy yes.

    Risks and how to handle them

    • Prompt annoys hobby users

      Show it once per session, with "Personal use" as a one-click path.

    Builds on

    Evidence

    • Your websiteUrdu text-to-speech page: Urdu script or Roman Urdu in, MP3 out; use cases for voice agents, YouTube and social video, lessons, audiobooks and announcements; "Free-tier output does not include a commercial license", with commercial use depending on the subscription but no plan named; no prices on the page; voice personas labelled Punjabi, Sindhi, Balochi and Pashtun, but no page for any language besides Urdu.Source
    • Your websiteHomepage: H1 "Finally, AI Voices that sound like Pakistan"; "We make world-leading voice models for regional languages"; three products (Text to Speech Studio, Developer API, AI Calling in "Early Preview"); "100+ AI Voices"; plans Free $0 (10 min TTS, 10 min AI calling), Starter $5/month (100 min TTS, 125 min calling), Pro $50 ("Popular"; 25 hours, 1,250 min), Growth $300 (200 hours, 7,500 min) and custom Enterprise with on-prem; a "Call me now" phone demo and two sample calls (a property-expo invitation, and a sales call about a crop disease offering a free expert visit); no customer logos, testimonials or case studies on the page.Source
  • 21

    A November 2025 scan of upliftai.org found Google Analytics and PostHog. The current homepage scan, after the move to Next.js, detects no analytics tool at all. Meanwhile the site is changing fast: a new headline, new AI calling and industry pages. If tracking really did drop out in the rebuild, none of the ideas in this plan can be measured.

    Medium impactLow effort First result in 2 weeks CTO

    First action

    Check whether any analytics is still firing on the new site.

    How to measure it

    Attributed signups

    Now
    –
    Target
    90%
    By
    30 days

    Steps

    1. 1

      Audit tracking

      What fires on the marketing site and the app.

      CTO 2 hours

    2. 2

      Define events

      Seven funnel steps, from sample played to first payment.

      CTO 1 day

    3. 3

      Instrument

      Site and app, with a consent banner if needed.

      CTO 3 days

    4. 4

      Tag sources

      UTMs on LinkedIn, docs and partner links.

      CEO 1 day

    5. 5

      Build the dashboard

      Weekly funnel by page and source.

      CTO 1 day

    6. 6

      Review weekly

      Pick the next fix from the biggest drop.

      CEO weekly

    First result
    2 weeks
    Cost
    $0 to low monthly tool cost
    Owner
    CTO

    Smallest test

    Add sample-played and Start Free events this week.

    Why it works

    Confirm what is tracked today. Then instrument one funnel: page view, voice sample played, Start Free, signup, first generation or first API call, first calling campaign, first payment. Tag the source (search, LinkedIn, HN, partner) and the landing page, and review it weekly.

    With a three-person team, knowing which page and channel produce paying users is the only way to choose what to build next.

    Risks and how to handle them

    • Tracking slows the page

      Load analytics after the page is interactive; keep it to one tool.

    Evidence

    • ResearchSite scan: LinkedIn is the only social profile linked; no blog, newsletter or video channel; a November 2025 scan found Google Analytics and PostHog, and the current homepage scan finds no analytics tool; the headline moved from "Realistic Text to Speech Models for Pakistani Languages" to "AI Voices that sound like Pakistan".Source

This quarter 11

The core plays for the next three months.

  • 01

    Uplift's e-commerce page promises "Nothing ships without a yes", but getting there means "Your store or order system posts orders to Uplift AI's API" and a deployment that targets production "within 30 days" after a sales call. No store platform is named. Pakistani merchants are already buying this job inside Shopify: Robo Calls Pakistan has 133 reviews at 5.0 for press-1-to-confirm calls, and a newer app starts at $2 a month plus $0.020 per answered call. Those apps take keypad answers. Uplift's agent can hold a real Urdu and English conversation and record address corrections before dispatch, but only merchants willing to integrate an API can get it.

    High impactMedium effort First result in 6 weeks CTO

    First action

    Write the order-to-call-to-tag flow on one page and list the outcome tags a merchant would act on.

    How to measure it

    Paying Shopify installs

    Now
    –
    Target
    50 stores
    By
    90 days after listing

    Steps

    1. 1

      Define the flow

      New COD order, call within minutes, outcomes written back as tags and a note, retries for no answer.

      CTO 2 days

    2. 2

      Build the app

      Order webhook, OAuth install, tags, Flow trigger and an outcome log inside Shopify admin.

      CTO 3 weeks

    3. 3

      Design onboarding

      Pick a voice, preview the script in Urdu, set retry rules, place a test call to your own phone.

      Founding Design Engineer 1 week

    4. 4

      Pilot with stores

      Five stores run it side by side with their current confirmation method for two weeks.

      CEO 2 weeks

    5. 5

      Write the listing

      Lead with address correction and real call recordings; show per-minute pricing plainly.

      CEO 2 days

    6. 6

      Submit and launch

      App review, then announce on LinkedIn and to e-commerce leads from the industry page.

      CTO 2 weeks

    7. 7

      Collect reviews

      Ask pilot stores for honest reviews once they have a week of results.

      CEO ongoing

    First result
    6 weeks
    Cost
    Engineering time; no paid media needed
    Owner
    CTO

    Smallest test

    Run the flow by hand for one Shopify store for a week, using the existing API and order export, and count address corrections.

    Why it works

    Ship a Shopify app that installs in minutes. It listens for new COD orders, calls the buyer with Uplift's Urdu agent, and writes the outcome back as order tags and notes (confirmed, address changed, cancelled, unreachable), with a Shopify Flow trigger so merchants can hold unconfirmed orders. Bill through the existing plans (Starter covers 125 calling minutes, Pro 1,250) and lead the listing with what keypads cannot do: "fixes the address on the call". Keep the 30-day managed deployment for large stores and couriers.

    Merchants searching the App Store for COD confirmation already have budget and intent, and the reviews on rival apps show the category converts. A conversational agent is a visible upgrade over press-1, and app-store installs give Uplift a self-serve channel with no sales call.

    Ready to use

    ChecklistShopify COD app launch list

    [ ] Webhook on new COD orders
    [ ] Call within 5 minutes, with retry rules
    [ ] Outcome tags: confirmed / address-changed / cancelled / unreachable
    [ ] Corrected address written to the order note
    [ ] Shopify Flow trigger to hold unconfirmed orders
    [ ] Test call to the merchant's own phone in onboarding
    [ ] Three real Urdu call recordings on the listing
    [ ] Pricing in plain terms: minutes per plan, cost per call
    [ ] Five pilot stores with before/after return counts

    Risks and how to handle them

    • Keypad apps are cheaper per call

      Sell on address corrections and avoided returns, and show the cost of one returned parcel next to the price of a call.

    • Concurrency of 1 call by default slows big stores

      Raise concurrency automatically for app installs on paid plans (see idea 9).

    Evidence

    • Your websiteE-commerce page: "Confirm every COD order before you ship it." and "Nothing ships without a yes."; the agent records address corrections and declines before dispatch; set-up is "Your store or order system posts orders to Uplift AI's API", with no store platform named; "We scope, build, test, and target production within 30 days"; main button "Book a call"; no customer proof or result numbers.Source
    • Competitor researchShopify App Store: Robo Calls Pakistan (Devsol, Lahore; launched December 2022) is rated 5.0 from 133 reviews for recorded confirmation calls where buyers press 1 to confirm, 2 to cancel or 3 for a callback, with order tags and Shopify Flow. Robo Call Pakistan (Punjani Inc., Karachi; launched April 2025; 35 reviews) starts at $2 a month plus $0.020 per answered call, with an address validator and auto-cancel for unverified orders.Source
    • Your websiteHomepage: H1 "Finally, AI Voices that sound like Pakistan"; "We make world-leading voice models for regional languages"; three products (Text to Speech Studio, Developer API, AI Calling in "Early Preview"); "100+ AI Voices"; plans Free $0 (10 min TTS, 10 min AI calling), Starter $5/month (100 min TTS, 125 min calling), Pro $50 ("Popular"; 25 hours, 1,250 min), Growth $300 (200 hours, 7,500 min) and custom Enterprise with on-prem; a "Call me now" phone demo and two sample calls (a property-expo invitation, and a sales call about a crop disease offering a free expert visit); no customer logos, testimonials or case studies on the page.Source
    • Your websiteVoice agent limits: default concurrency is 1 call, "shared by inbound and outbound", "Raised on request, into the thousands"; outbound calls use a pool of Pakistani numbers, with a dedicated number by email "within a day"; "Every limit can be raised on request" by email.Source
  • 09

    Voice agents run one call at a time by default, "shared by inbound and outbound", and more is "Raised on request, into the thousands" by emailing the founders. Dedicated numbers are also by email, "within a day". A store with 300 COD orders a morning hits the limit on day one and has to write in. Concurrency is the thing growing customers most want to pay for, and today it costs nothing and needs a founder's inbox.

    High impactMedium effort First result in 1 month CTO

    First action

    Pull the last three months of concurrency and number requests and how quickly each was answered.

    How to measure it

    Add-on revenue

    Now
    –
    Target
    10% of calling revenue
    By
    1 quarter

    Steps

    1. 1

      Review requests

      Who asked, for how many lines and how fast it was granted.

      CEO 1 day

    2. 2

      Set tiers

      Concurrency per plan and add-on prices.

      CEO 2 days

    3. 3

      Enforce in code

      Per-account concurrency from the plan.

      CTO 1 week

    4. 4

      Billing screen

      Buy lines and a dedicated number.

      Founding Design Engineer 1 week

    5. 5

      Queue indicator

      Waiting calls and an "add lines" button in the dashboard.

      Founding Design Engineer 3 days

    6. 6

      Update docs

      Limits page shows plan defaults.

      CTO 1 day

    First result
    1 month
    Cost
    Engineering time; telecom number costs passed through
    Owner
    CTO

    Smallest test

    Offer a paid "extra lines" add-on by email to the next five customers who ask.

    Why it works

    Tie concurrency to plans: Starter 2, Pro 5, Growth 20, Enterprise custom, with paid add-ons for more lines and a dedicated number bought in the billing screen. Show live queue length in the calling dashboard with an "add lines" button when calls wait.

    Usage-based limits that customers can raise themselves turn a support ticket into revenue, and they make Pro and Growth worth more than their minutes.

    Risks and how to handle them

    • Existing customers lose free capacity

      Grandfather current grants for six months.

    Evidence

    • Your websiteVoice agent limits: default concurrency is 1 call, "shared by inbound and outbound", "Raised on request, into the thousands"; outbound calls use a pool of Pakistani numbers, with a dedicated number by email "within a day"; "Every limit can be raised on request" by email.Source
    • Your websiteHomepage: H1 "Finally, AI Voices that sound like Pakistan"; "We make world-leading voice models for regional languages"; three products (Text to Speech Studio, Developer API, AI Calling in "Early Preview"); "100+ AI Voices"; plans Free $0 (10 min TTS, 10 min AI calling), Starter $5/month (100 min TTS, 125 min calling), Pro $50 ("Popular"; 25 hours, 1,250 min), Growth $300 (200 hours, 7,500 min) and custom Enterprise with on-prem; a "Call me now" phone demo and two sample calls (a property-expo invitation, and a sales call about a crop disease offering a free expert visit); no customer logos, testimonials or case studies on the page.Source
  • 10

    One of the two AI calling demos on the homepage is a sales call that warns a farmer about a crop disease and offers a free expert visit. Seed coverage says a global crop-science company is rolling out Uplift's voice tools for farmers. Yet the six industry pages cover e-commerce, couriers, healthcare, real estate, finance and telecom, and there is no agriculture page. Uplift's most distinctive use case, reaching rural customers who may not read, has nowhere to land.

    Medium impactLow effort First result in 1 month CEO

    First action

    Draft the page from the existing crop-disease demo script.

    How to measure it

    Agriculture page bookings

    Now
    –
    Target
    10 qualified calls
    By
    1 quarter

    Steps

    1. 1

      Write the page

      Use cases, sample, FAQ, 30-day launch plan.

      CEO 2 days

    2. 2

      Record samples

      Outbreak alert and dealer follow-up in two languages.

      CEO 1 day

    3. 3

      Design and publish

      Reuse the industry template.

      Founding Design Engineer 2 days

    4. 4

      Add to the nav

      Industries menu and sitemap.

      CTO 1 hour

    5. 5

      Outreach

      Twenty agribusinesses, input companies and agri-lenders on LinkedIn.

      CEO 3 weeks

    6. 6

      Seek a case study

      Ask the existing agri customer what can be published.

      CEO 1 month

    First result
    1 month
    Cost
    $0
    Owner
    CEO

    Smallest test

    Publish the page and send it to ten agri contacts.

    Why it works

    Publish /industries/agriculture: advisory and outbreak alerts, dealer and distributor follow-ups, field-visit booking and input reminders in Urdu, Punjabi and Sindhi. Use the crop-disease demo as the sample, add a farmer-facing FAQ (no internet needed, works on any phone), and describe the agri deployment generically until it can be named.

    Agribusinesses, input companies and agri-fintechs need to reach farmers by phone, and few voice vendors speak to them. The page converts the demo people already hear into a sector pipeline.

    Risks and how to handle them

    • Agronomy advice without experts

      Keep the agent to approved scripts and route crop questions to the customer's experts, as the healthcare page does for clinical questions.

    Evidence

    • Your websiteSitemap (checked November 10, 2026): 24 URLs, all with the same lastmod; 12 are signed-in app screens under /app/ (studio, calling, billing, team, API keys and others); one language page (Urdu), six industry pages and the AI calling page; no blog, pricing page, case studies, or agriculture or education pages.Source
    • PressSeed round coverage: $3.5 million led by Y Combinator and Indus Valley Capital; Khan Academy has used the technology to publish 2,500 educational videos in Urdu; a global crop-science company is rolling out voice tools for farmers; more than 1,000 developers build on the APIs; next: more languages, especially in MENA, and education, public-sector and enterprise integrations.Source
    • PressLaunch HN ("Uplift (YC S25) – Voice models for under-served languages"): public TTS APIs for Urdu, Sindhi and Balochi; Khan Academy uses it to dub videos into Urdu; a prototype WhatsApp bot reached "800 people" in two days; commenters asked what happens when bigger platforms add these languages, how it benchmarks (the founders had not benchmarked systematically) and how misuse is prevented; offline models and telco hotlines are on the roadmap; language requests go through a form.Source
  • 11

    Education is where Uplift has its best proof: Khan Academy went from 104 dubbed videos to 2,500, and the company says dubbing is 10 times faster. The Urdu page lists "Lessons & learning", and the seed coverage names education platforms as a next integration. But there is no education page, and the industry menu doesn't include the sector Uplift has done the most in.

    Medium impactLow effort First result in 1 month CEO

    First action

    Write the page around the Khan Academy numbers.

    How to measure it

    Paying education accounts

    Now
    –
    Target
    10
    By
    1 quarter

    Steps

    1. 1

      Write the page

      Proof, workflow, hourly price, FAQ.

      CEO 2 days

    2. 2

      Record samples

      Teacher and narrator voices on a real lesson.

      CEO 1 day

    3. 3

      Publish

      Industry template, nav and sitemap.

      Founding Design Engineer 2 days

    4. 4

      Free first lesson

      An offer form for ed-tech companies and creators.

      CTO 1 day

    5. 5

      Outreach

      Thirty ed-tech companies and large Urdu course creators.

      CEO 3 weeks

    6. 6

      Convert

      Follow up each free lesson with a plan quote.

      CEO ongoing

    First result
    1 month
    Cost
    Free lesson minutes
    Owner
    CEO

    Smallest test

    Dub one free lesson for three ed-tech companies.

    Why it works

    Publish /industries/education with the Khan Academy story, a dubbing workflow (script in, reviewed Urdu audio out), pricing per hour of finished audio on Pro and Growth, and samples in a teacher's and a narrator's voice. Offer ed-tech companies and course creators a free first lesson dubbed.

    Ed-tech and course creators in Pakistan have large back catalogues in English, and a proven, priced workflow turns them into steady Growth-plan customers.

    Risks and how to handle them

    • Pronunciation errors on technical terms

      Use phrase replacements and offer a review pass.

    Builds on

    Evidence

    • PressKhan Academy launch coverage: 104 Urdu-dubbed videos at the time with thousands more planned; the company says dubbing is 10 times faster than traditional voiceover.Source
    • PressSeed round coverage: $3.5 million led by Y Combinator and Indus Valley Capital; Khan Academy has used the technology to publish 2,500 educational videos in Urdu; a global crop-science company is rolling out voice tools for farmers; more than 1,000 developers build on the APIs; next: more languages, especially in MENA, and education, public-sector and enterprise integrations.Source
    • Your websiteSitemap (checked November 10, 2026): 24 URLs, all with the same lastmod; 12 are signed-in app screens under /app/ (studio, calling, billing, team, API keys and others); one language page (Urdu), six industry pages and the AI calling page; no blog, pricing page, case studies, or agriculture or education pages.Source
  • 12

    The homepage says Uplift makes "voice models for regional languages", and the Launch HN post says public APIs exist for Urdu, Sindhi and Balochi. But the site has exactly one language page, Urdu, and the Urdu page's Punjabi, Sindhi and Balochi labels are personas (accents), not languages. Searchers for Sindhi or Balochi text-to-speech, where few vendors compete, find nothing from Uplift.

    Medium impactMedium effort First result in 2 months CTO

    First action

    List which languages are production-ready in Studio and the API today.

    How to measure it

    Language page signups

    Now
    –
    Target
    300 a month
    By
    2 quarters

    Steps

    1. 1

      Confirm readiness

      Which languages are live, beta or planned.

      CTO 1 day

    2. 2

      Write native copy

      Have native speakers write each page in its own script, plus English.

      CEO 2 weeks

    3. 3

      Record samples

      Three per language.

      CEO 1 week

    4. 4

      Publish

      Pages, hreflang where translated, sitemap.

      Founding Design Engineer 1 week

    5. 5

      Request form

      Collect demand for the next languages.

      CTO 1 day

    6. 6

      Measure

      Signups and API usage per language.

      CTO monthly

    First result
    2 months
    Cost
    Native-speaker review time
    Owner
    CTO

    Smallest test

    Publish a Sindhi page and track signups for a month.

    Why it works

    For each language that is production-ready, publish /text-to-speech/<language> with native-script samples, input options, use cases (local news, lessons, government and NGO announcements) and an honest status ("live" or "beta"). Add a language request form, as the founders already run for HN readers, to gather demand for Pashto and others.

    Underserved languages have little search competition, and every page is a doorway to the same Studio and API. Honest status labels keep trust.

    Risks and how to handle them

    • Quality below Urdu

      Label beta languages clearly and gather feedback through the tuning platform.

    Evidence

    • Your websiteUrdu text-to-speech page: Urdu script or Roman Urdu in, MP3 out; use cases for voice agents, YouTube and social video, lessons, audiobooks and announcements; "Free-tier output does not include a commercial license", with commercial use depending on the subscription but no plan named; no prices on the page; voice personas labelled Punjabi, Sindhi, Balochi and Pashtun, but no page for any language besides Urdu.Source
    • PressLaunch HN ("Uplift (YC S25) – Voice models for under-served languages"): public TTS APIs for Urdu, Sindhi and Balochi; Khan Academy uses it to dub videos into Urdu; a prototype WhatsApp bot reached "800 people" in two days; commenters asked what happens when bigger platforms add these languages, how it benchmarks (the founders had not benchmarked systematically) and how misuse is prevented; offline models and telco hotlines are on the roadmap; language requests go through a form.Source
    • ResearchYC profile: "builds foundational voice models for regional languages such as Urdu, Bengali, and Greek" for voice e-commerce, voice banking and AI tutors.Source
  • 13

    Uplift confirms COD orders by phone call only. WhatsApp-first rivals like eGrow confirm COD orders by message instead. Uplift already has a WhatsApp bot tutorial in its docs, and its prototype WhatsApp bot reached 800 people in two days, but the AI calling page never mentions WhatsApp. An "unreachable" order is the one most likely to come back.

    High impactMedium effort First result in 2 months CTO

    First action

    Measure what share of calls end "unreachable" today.

    How to measure it

    Confirmation coverage

    Now
    –
    Target
    +15 points
    By
    1 quarter

    Steps

    1. 1

      Size the gap

      Unreachable share by customer.

      CTO 1 day

    2. 2

      Set up messaging

      WhatsApp Business API account and templates.

      CTO 2 weeks

    3. 3

      Build the fallback

      Voice note plus buttons after the last retry.

      CTO 2 weeks

    4. 4

      Write back

      Replies update the order outcome.

      CTO 1 week

    5. 5

      Show in the dashboard

      One timeline per order across call and message.

      Founding Design Engineer 1 week

    6. 6

      Price and launch

      Add-on on Pro, Growth and the Shopify app.

      CEO 1 week

    First result
    2 months
    Cost
    Messaging fees per conversation
    Owner
    CTO

    Smallest test

    Send manual WhatsApp follow-ups for one customer's unreachable orders for a week.

    Why it works

    When retries run out, send a WhatsApp voice note in the same Urdu voice with the order summary and two reply buttons (confirm or change), and write the reply back to the order. Offer it as an add-on on Pro and Growth and in the Shopify app.

    Merchants don't care about the channel, they care about confirmed orders. Voice first, then a voice note on WhatsApp, covers buyers who miss calls and buyers who don't read.

    Risks and how to handle them

    • Message template approval

      Use order-update templates and get them approved early.

    Builds on

    Evidence

    • Competitor researchWhatsApp-first rival eGrow sends an automated Urdu and English WhatsApp message right after a COD order to confirm details or fix the address; its blog claims, without a source, that COD is "60-80%" of online purchases in Pakistan and return-to-origin rates are "often ranging from 20-40%".Source
    • Your websiteDeveloper docs: REST and WebSocket text-to-speech with "~150-200ms Latency"; tutorials for a WhatsApp bot and a LiveKit voice agent; step one is "Visit upliftai.org to generate your API key", with no direct link to the key page; no pricing or free allowance on the page.Source
    • PressLaunch HN ("Uplift (YC S25) – Voice models for under-served languages"): public TTS APIs for Urdu, Sindhi and Balochi; Khan Academy uses it to dub videos into Urdu; a prototype WhatsApp bot reached "800 people" in two days; commenters asked what happens when bigger platforms add these languages, how it benchmarks (the founders had not benchmarked systematically) and how misuse is prevented; offline models and telco hotlines are on the roadmap; language requests go through a form.Source
    • Your websiteAI calling page: "AI calling in Pakistan. Built around your customers."; Urdu-first, including "the English and Punjabi words customers mix in"; inbound and outbound calls with a recording, transcript, summary and outcome for each; "Starts at $0.04 per minute", final pricing through a "Talk to Team" call; no customers named and no mention of WhatsApp.Source
  • 14

    The new Founding Design Engineer's 90-day goal is "a way for small-business owners to create and improve voice agents without prompt engineering or ML knowledge". Today every industry page leads with "Book a call" and a 30-day deployment, which suits enterprises but not a clinic or a small store on the $5 or $50 plan. The self-serve plans exist, but the self-serve agent doesn't yet.

    High impactMedium effort First result in 2 months Founding Design Engineer

    First action

    Pick the two templates with the most demand from recent sales calls.

    How to measure it

    Week-1 calling activation

    Now
    –
    Target
    25% of new signups
    By
    1 quarter

    Steps

    1. 1

      Choose templates

      Five, starting with COD and appointments.

      CEO 2 days

    2. 2

      Write the questions

      Plain-language set-up for each template.

      Founding Design Engineer 1 week

    3. 3

      Build the flow

      Questions, test call, CSV upload, launch.

      Founding Design Engineer 4 weeks

    4. 4

      Guardrails

      Approved actions and escalation for each template.

      CTO 1 week

    5. 5

      Link from pages

      Each industry page opens its template.

      CTO 1 day

    6. 6

      Watch sessions

      Where owners drop off, and fix it weekly.

      Founding Design Engineer weekly

    First result
    2 months
    Cost
    Design and engineering time
    Owner
    Founding Design Engineer

    Smallest test

    Launch the COD template for ten free-plan users and watch them set it up.

    Why it works

    Ship five templates in the calling app (COD confirmation, appointment reminder, delivery reschedule, payment reminder, lead follow-up). Each asks a few plain questions (business name, hours, what the agent may change), lets the owner hear a test call on their own phone, then upload a CSV of contacts. Link each industry page's "Start Free" straight to its template.

    Small businesses buy what they can try in one sitting. Templates turn industry page visitors into active callers without founder time.

    Risks and how to handle them

    • Poorly set-up agents annoy customers

      Require a test call and a small first batch before full upload.

    Evidence

    • HiringFounding Design Engineer (Lahore): owns brand identity and the hardest interfaces, including audio annotation tools; within 90 days, ship a way for small-business owners to create and improve voice agents without prompt engineering or ML knowledge; within a year, one design system across brand, products and internal tools. YC lists a team of 3.Source
    • Your websiteIndustry pages (healthcare, telecom and e-commerce, checked November 10, 2026): each plays the same three sample calls (courier delivery follow-up, e-commerce order confirmation, real estate site visit), so the healthcare and telecom pages have no sample from their own sector; no named customers or metrics; each promises production within 30 days.Source
    • Your websiteE-commerce page: "Confirm every COD order before you ship it." and "Nothing ships without a yes."; the agent records address corrections and declines before dispatch; set-up is "Your store or order system posts orders to Uplift AI's API", with no store platform named; "We scope, build, test, and target production within 30 days"; main button "Book a call"; no customer proof or result numbers.Source
  • 15

    On Launch HN, commenters asked what happens when bigger platforms add these languages and how Uplift benchmarks, and the founders said they had not benchmarked systematically. Since then a global voice platform has published an Urdu page claiming the "best Urdu text to speech". Uplift's whole pitch is "sounds like Pakistan", but there is no public evidence a buyer can check.

    Medium impactMedium effort First result in 6 weeks CTO

    First action

    Write the 50-sentence test set around real call content: addresses, amounts, dates.

    How to measure it

    Benchmark-driven signups

    Now
    –
    Target
    500
    By
    1 quarter after publishing

    Steps

    1. 1

      Build the test set

      50 sentences from real call and video use.

      CTO 3 days

    2. 2

      Render audio

      Each system's default best Urdu voice.

      CTO 2 days

    3. 3

      Recruit listeners

      A few hundred native speakers across regions.

      CEO 2 weeks

    4. 4

      Run blind ratings

      Randomised pairs, naturalness and errors.

      Founding Engineer, Frontier Tuning Platform 1 week

    5. 5

      Publish

      Method, audio and results, including losses.

      Founding Design Engineer 1 week

    6. 6

      Repeat

      Quarterly, adding languages.

      CTO quarterly

    First result
    6 weeks
    Cost
    Small listener incentives
    Owner
    CTO

    Smallest test

    Run 20 sentences with 50 listeners and share the results on LinkedIn.

    Why it works

    Run a blind test: 50 everyday Urdu and Roman-Urdu sentences (names, numbers, addresses, code-switching), rendered by Uplift and the main alternatives, rated by a few hundred native listeners for naturalness and correctness. Publish the method, the audio and the results page, and repeat it each quarter, including for Sindhi and Balochi.

    A transparent test answers the buyer's first question with evidence, earns press and links, and holds up even when results are mixed, because the method is open.

    Risks and how to handle them

    • Uplift doesn't win every category

      Publish anyway; an honest method is the point, and it shows where to improve.

    Evidence

    • PressLaunch HN ("Uplift (YC S25) – Voice models for under-served languages"): public TTS APIs for Urdu, Sindhi and Balochi; Khan Academy uses it to dub videos into Urdu; a prototype WhatsApp bot reached "800 people" in two days; commenters asked what happens when bigger platforms add these languages, how it benchmarks (the founders had not benchmarked systematically) and how misuse is prevented; offline models and telco hotlines are on the roadmap; language requests go through a form.Source
    • Competitor researchElevenLabs runs an Urdu page ("Create Realistic Urdu Text to Speech") among 90+ languages, claims it "offers the best Urdu text to speech (TTS) online" and gives 10,000 characters free a month.Source
  • 16

    Uplift sells to financial services, healthcare and telecom, where buyers need security and data answers before a pilot. Each industry FAQ answers this in one or two lines ("minimum data scope, access, and retention are agreed before launch"). Launch HN commenters asked how misuse is prevented. There is no single page a compliance reviewer can read.

    Medium impactLow effort First result in 1 month CEO

    First action

    Collect the security and data questions from the last five enterprise conversations.

    How to measure it

    Enterprise sales cycle

    Now
    –
    Target
    -25%
    By
    2 quarters

    Steps

    1. 1

      Gather questions

      From recent enterprise calls.

      CEO 2 days

    2. 2

      Document data flows

      Recordings, transcripts, retention, access.

      CTO 1 week

    3. 3

      Write the use policy

      Consent, disclosure and prohibited uses.

      CEO 3 days

    4. 4

      Legal review

      Against the terms of service and privacy policy.

      CEO 1 week

    5. 5

      Publish and link

      From industry pages and Enterprise.

      Founding Design Engineer 1 day

    6. 6

      Keep it current

      Update with each new certification or region.

      CTO quarterly

    First result
    1 month
    Cost
    Legal review
    Owner
    CEO

    Smallest test

    Send a draft trust document to the next enterprise prospect and note which questions remain.

    Why it works

    Publish /trust covering data flows and retention (recordings, transcripts, sessions), where data is processed, on-prem options, who can review calls, consent and disclosure scripts for outbound calls, voice-use rules (no impersonating real people, no political or misleading content) and how to report misuse. Link it from every industry page and the Enterprise plan.

    Enterprise deals stall in security review. A clear page shortens it and shows the founders took misuse seriously from day one.

    Risks and how to handle them

    • Over-promising on compliance

      State only what is true today and mark plans as plans.

    Evidence

    • Your websiteIndustry pages (healthcare, telecom and e-commerce, checked November 10, 2026): each plays the same three sample calls (courier delivery follow-up, e-commerce order confirmation, real estate site visit), so the healthcare and telecom pages have no sample from their own sector; no named customers or metrics; each promises production within 30 days.Source
    • PressLaunch HN ("Uplift (YC S25) – Voice models for under-served languages"): public TTS APIs for Urdu, Sindhi and Balochi; Khan Academy uses it to dub videos into Urdu; a prototype WhatsApp bot reached "800 people" in two days; commenters asked what happens when bigger platforms add these languages, how it benchmarks (the founders had not benchmarked systematically) and how misuse is prevented; offline models and telco hotlines are on the roadmap; language requests go through a form.Source
    • Your websiteVoice agent limits: default concurrency is 1 call, "shared by inbound and outbound", "Raised on request, into the thousands"; outbound calls use a pool of Pakistani numbers, with a dedicated number by email "within a day"; "Every limit can be raised on request" by email.Source
  • 17

    The e-commerce page promises fewer avoidable returns but shows no numbers, while keypad apps compete at $2 a month plus $0.020 per answered call. Merchants compare call prices. What they should compare is the cost of a returned parcel (shipping both ways, packing, stock tied up) against a $0.04-a-minute call that can also fix the address.

    Medium impactLow effort First result in 1 month CEO

    First action

    Get return costs from two or three pilot merchants to set honest defaults.

    How to measure it

    E-commerce page conversion

    Now
    –
    Target
    +20%
    By
    60 days

    Steps

    1. 1

      Collect inputs

      Real return costs from pilot merchants.

      CEO 1 week

    2. 2

      Write the formula

      Conservative reduction range, clearly labelled as an estimate.

      CEO 1 day

    3. 3

      Build it

      Four inputs, one result card.

      Founding Design Engineer 3 days

    4. 4

      Connect

      Pass results to the booking form and app install.

      CTO 1 day

    5. 5

      Test

      Compare conversion with and without it.

      CTO 4 weeks

    6. 6

      Replace estimates

      Use real customer results once available.

      CEO 1 quarter

    First result
    1 month
    Cost
    $0
    Owner
    CEO

    Smallest test

    Add a static example ("1,000 orders a month at 20% returns...") to the page.

    Why it works

    Add a calculator: monthly COD orders, average order value, current return rate and cost per return. It shows returns per month, their cost, and the cost of calling every order, with a conservative range for the reduction. Pre-fill a sensible default and let merchants change it. Pass the result into "Book a call" or the Shopify install.

    It moves the conversation from price per call to money saved per month, where a conversational agent wins.

    Risks and how to handle them

    • Over-promising savings

      Default to a low reduction and show the assumption.

    Evidence

    • Your websiteE-commerce page: "Confirm every COD order before you ship it." and "Nothing ships without a yes."; the agent records address corrections and declines before dispatch; set-up is "Your store or order system posts orders to Uplift AI's API", with no store platform named; "We scope, build, test, and target production within 30 days"; main button "Book a call"; no customer proof or result numbers.Source
    • Competitor researchShopify App Store: Robo Calls Pakistan (Devsol, Lahore; launched December 2022) is rated 5.0 from 133 reviews for recorded confirmation calls where buyers press 1 to confirm, 2 to cancel or 3 for a callback, with order tags and Shopify Flow. Robo Call Pakistan (Punjani Inc., Karachi; launched April 2025; 35 reviews) starts at $2 a month plus $0.020 per answered call, with an address validator and auto-cancel for unverified orders.Source
    • Competitor researchWhatsApp-first rival eGrow sends an automated Urdu and English WhatsApp message right after a COD order to confirm details or fix the address; its blog claims, without a source, that COD is "60-80%" of online purchases in Pakistan and return-to-origin rates are "often ranging from 20-40%".Source
  • 18

    Uplift publishes @upliftai/assistant-widget and @upliftai/assistants-react on npm, but they get 11 and 14 downloads a week, and the marketing site never mentions an embeddable website agent. Every product page talks about phone calls and Studio. Uplift already has a drop-in way to put an Urdu voice agent on any website, and almost nobody knows.

    Medium impactLow effort First result in 1 month CTO

    First action

    Put the widget live on upliftai.org, answering questions about plans and voices.

    How to measure it

    Active widget installs

    Now
    –
    Target
    100 domains
    By
    1 quarter

    Steps

    1. 1

      Dogfood

      Run the widget on upliftai.org.

      CTO 3 days

    2. 2

      Write the guide

      One-tag install, React option, config.

      CTO 2 days

    3. 3

      Snippet in the app

      Copy-paste tag tied to each assistant.

      Founding Design Engineer 1 week

    4. 4

      Add templates

      Store FAQ and clinic booking.

      CTO 1 week

    5. 5

      Announce

      LinkedIn post and developer email.

      CEO 1 day

    6. 6

      Track installs

      Active domains and minutes used.

      CTO monthly

    First result
    1 month
    Cost
    $0
    Owner
    CTO

    Smallest test

    Add the widget to the Urdu page and count conversations in a week.

    Why it works

    Write a docs page and a short landing section: "Add an Urdu voice assistant to your site with one script tag", with a live demo on upliftai.org itself, a copy-paste snippet tied to the user's assistant, and templates for store FAQs and clinic booking. Track installs by domain.

    A visible, working widget is self-demonstrating marketing, and each install puts Uplift's voice in front of the customer's visitors.

    Risks and how to handle them

    • Widget adds weight to customer sites

      Load on click only, with a tiny launcher.

    Evidence

    • Researchnpm: @upliftai/assistants-react (14 weekly downloads), @upliftai/assistant-widget (11) and @upliftai/sdk-js (6).Source
    • Your websiteHomepage: H1 "Finally, AI Voices that sound like Pakistan"; "We make world-leading voice models for regional languages"; three products (Text to Speech Studio, Developer API, AI Calling in "Early Preview"); "100+ AI Voices"; plans Free $0 (10 min TTS, 10 min AI calling), Starter $5/month (100 min TTS, 125 min calling), Pro $50 ("Popular"; 25 hours, 1,250 min), Growth $300 (200 hours, 7,500 min) and custom Enterprise with on-prem; a "Call me now" phone demo and two sample calls (a property-expo invitation, and a sales call about a crop disease offering a free expert visit); no customer logos, testimonials or case studies on the page.Source

Big bets 2

Larger plays that can change the trajectory.

  • 19

    Uplift is hiring a founding engineer to build "the platform that lets people around the world improve audio models for their own languages and use cases", and the 90-day goal is an outside community building a production model with minimal help. Language requests already arrive through the form the founders opened on Launch HN. The YC profile already names Bengali and Greek. The platform could become a growth engine, not only a tool.

    High impactHigh effort First result in 1 quarter Founding Engineer

    First action

    Choose two pilot communities from the language request form with the most demand.

    How to measure it

    Community languages live

    Now
    –
    Target
    3
    By
    2 quarters

    Steps

    1. 1

      Pick pilots

      Two languages with strong requests and a willing partner.

      CEO 2 weeks

    2. 2

      Ship self-service

      Contribute audio, train, evaluate.

      Founding Engineer, Frontier Tuning Platform 1 month

    3. 3

      Agree terms

      Data rights, consent, credits or revenue share.

      CEO 1 month

    4. 4

      Design the contributor flow

      Recording and labelling that stays fast and pleasant.

      Founding Design Engineer 1 month

    5. 5

      Public leaderboard

      Progress per language.

      Founding Design Engineer 2 weeks

    6. 6

      Launch pages

      A language page for each one that reaches production.

      CTO ongoing

    7. 7

      Open applications

      After two pilots succeed.

      CEO 2 quarters

    First result
    1 quarter
    Cost
    Training compute and partner credits
    Owner
    Founding Engineer, Frontier Tuning Platform

    Smallest test

    Run one language pilot with one university partner.

    Why it works

    Launch a community programme: universities, language societies and NGOs apply with a language, contribute recorded audio through the platform, and get a free hosted model plus revenue share or credits when businesses use it. Publish a public leaderboard of community languages, and give each language a page on upliftai.org when it reaches production quality.

    Each community brings data, native-speaker evaluation and its own audience, so Uplift can add languages faster than it could hire for them, and every new language opens a market.

    Risks and how to handle them

    • Consent and data ownership disputes

      Clear contributor agreements and consent recorded per speaker.

    • Quality varies

      A published quality bar before any language is marked production.

    Builds on

    Evidence

    • HiringFounding Engineer, Frontier Tuning Platform (Lahore): "You will build the platform that lets people around the world improve audio models for their own languages and use cases", for people without ML expertise or big budgets; a self-service workflow to contribute audio, train and evaluate within 30 days, one or two real customers within 60, and an outside community or customer building a production model with minimal help within 90.Source
    • PressLaunch HN ("Uplift (YC S25) – Voice models for under-served languages"): public TTS APIs for Urdu, Sindhi and Balochi; Khan Academy uses it to dub videos into Urdu; a prototype WhatsApp bot reached "800 people" in two days; commenters asked what happens when bigger platforms add these languages, how it benchmarks (the founders had not benchmarked systematically) and how misuse is prevented; offline models and telco hotlines are on the roadmap; language requests go through a form.Source
    • ResearchYC profile: "builds foundational voice models for regional languages such as Urdu, Bengali, and Greek" for voice e-commerce, voice banking and AI tutors.Source
  • 20

    Uplift's mission starts from people who cannot read, and on Launch HN the founders put offline models and telco hotlines on the roadmap because phone coverage is wider than internet access. Uplift already has inbound calling and a telecom industry page. A hotline that answers questions by voice on any phone would reach the people text and apps miss.

    High impactHigh effort First result in 2 quarters CEO

    First action

    Write a one-page pilot proposal for a telecom operator's customer-service hotline.

    How to measure it

    Hotline callers

    Now
    –
    Target
    50,000 a month
    By
    2 quarters after launch

    Steps

    1. 1

      Write the proposal

      Scope, savings model, pilot terms.

      CEO 1 week

    2. 2

      Find the partner

      One operator willing to run a pilot.

      CEO 1 quarter

    3. 3

      Build the flows

      Operator FAQs first, then partner content.

      CTO 1 month

    4. 4

      Design for voice-only

      Short menus, confirmations, no reading required.

      Founding Design Engineer 2 weeks

    5. 5

      Scale capacity

      Concurrency and monitoring for peak hours.

      CTO 1 month

    6. 6

      Measure and publish

      Calls handled, escalations, satisfaction.

      CEO monthly

    First result
    2 quarters
    Cost
    Engineering time and telecom integration costs
    Owner
    CEO

    Smallest test

    Run a WhatsApp or phone pilot on one topic for a few thousand callers, as the earlier WhatsApp prototype did.

    Why it works

    Pilot a short-code hotline with one telecom operator: callers ask in Urdu or a regional language and hear answers about the operator's services first, then partner content (farm advisories, health appointment booking, government service steps). The operator gets lower call-centre load and a new service; Uplift gets volume, data on real speech and a flagship case study.

    It puts the product where the mission is, uses inbound calling that already exists, and gives a telecom buyer a reason to sign a large contract.

    Risks and how to handle them

    • Long telecom sales cycles

      Start with a contained pilot on the operator's own support questions.

    • Wrong answers on public-service topics

      Approved content only, with transfers to humans.

    Builds on

    Evidence

    • PressLaunch HN ("Uplift (YC S25) – Voice models for under-served languages"): public TTS APIs for Urdu, Sindhi and Balochi; Khan Academy uses it to dub videos into Urdu; a prototype WhatsApp bot reached "800 people" in two days; commenters asked what happens when bigger platforms add these languages, how it benchmarks (the founders had not benchmarked systematically) and how misuse is prevented; offline models and telco hotlines are on the roadmap; language requests go through a form.Source
    • Your websiteIndustry pages (healthcare, telecom and e-commerce, checked November 10, 2026): each plays the same three sample calls (courier delivery follow-up, e-commerce order confirmation, real estate site visit), so the healthcare and telecom pages have no sample from their own sector; no named customers or metrics; each promises production within 30 days.Source
    • Your websiteAI calling page: "AI calling in Pakistan. Built around your customers."; Urdu-first, including "the English and Punjabi words customers mix in"; inbound and outbound calls with a recording, transcript, summary and outcome for each; "Starts at $0.04 per minute", final pricing through a "Talk to Team" call; no customers named and no mention of WhatsApp.Source

Wildcards 1

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

  • 22

    Uplift's library has more than 100 characters, from news anchors and storytellers to regional personas, and its Urdu page sells voiceovers for YouTube, reels and audiobooks. But the only social profile linked from the site is LinkedIn, with no blog, newsletter or video channel. The product's best advertisement, how it sounds, is only heard by people who already visit the site.

    Medium impactMedium effort First result in 1 month Founding Design Engineer

    First action

    Produce ten clips in Studio across five voices and post them over two weeks.

    How to measure it

    Social-sourced signups

    Now
    –
    Target
    1,000
    By
    1 quarter

    Steps

    1. 1

      Pick formats

      Five repeatable clip formats.

      Founding Design Engineer 2 days

    2. 2

      Produce a batch

      Ten clips voiced in Studio.

      Founding Design Engineer 1 week

    3. 3

      Open accounts

      YouTube, TikTok, Instagram, linked from the site footer.

      CEO 1 day

    4. 4

      Post on a schedule

      Daily for a month.

      Founding Design Engineer 1 month

    5. 5

      Creator submissions

      Feature creators' scripts with credit.

      CEO ongoing

    6. 6

      Measure

      Signups by tracked link per format.

      CTO monthly

    First result
    1 month
    Cost
    Team time; Studio minutes
    Owner
    Founding Design Engineer

    Smallest test

    Post ten clips and compare signups against LinkedIn posts.

    Why it works

    Start a YouTube Shorts, TikTok and Instagram channel where every clip is voiced in Studio: a daily Urdu proverb by the storyteller voice, a news-style explainer, a regional accent challenge, a before-and-after of a creator's script. Each clip says "voiced with Uplift Studio" and links to the free plan. Invite creators to submit scripts and feature the best ones.

    Short video is where Urdu creators, Uplift's Studio buyers, spend time, and each clip shows the quality while recruiting the people who would pay for it.

    Risks and how to handle them

    • Clips mistaken for real people or news

      Label every clip as an AI voice and avoid real news or public figures.

    Evidence

    • ResearchSite scan: LinkedIn is the only social profile linked; no blog, newsletter or video channel; a November 2025 scan found Google Analytics and PostHog, and the current homepage scan finds no analytics tool; the headline moved from "Realistic Text to Speech Models for Pakistani Languages" to "AI Voices that sound like Pakistan".Source
    • Your websiteUrdu text-to-speech page: Urdu script or Roman Urdu in, MP3 out; use cases for voice agents, YouTube and social video, lessons, audiobooks and announcements; "Free-tier output does not include a commercial license", with commercial use depending on the subscription but no plan named; no prices on the page; voice personas labelled Punjabi, Sindhi, Balochi and Pashtun, but no page for any language besides Urdu.Source
    • Your websiteHomepage: H1 "Finally, AI Voices that sound like Pakistan"; "We make world-leading voice models for regional languages"; three products (Text to Speech Studio, Developer API, AI Calling in "Early Preview"); "100+ AI Voices"; plans Free $0 (10 min TTS, 10 min AI calling), Starter $5/month (100 min TTS, 125 min calling), Pro $50 ("Popular"; 25 hours, 1,250 min), Growth $300 (200 hours, 7,500 min) and custom Enterprise with on-prem; a "Call me now" phone demo and two sample calls (a property-expo invitation, and a sales call about a crop disease offering a free expert visit); no customer logos, testimonials or case studies on the page.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.

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)
More example playbooks and reviews

More growth playbooks

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

  • Outtake growth strategyAn agentic AI platform for next-generation digital risk protection: it finds and takes down impersonation and fraud across domains, social…21 growth ideas
  • Kamiwaza growth strategyAn enterprise AI orchestration platform deployed on-premises or in the customer's own cloud: distributed inference, data connections…20 growth ideas
  • Vint & York growth strategyAn independent, New York-designed eyewear brand selling vintage-inspired prescription glasses and sunglasses online, with single-vision…22 growth ideas
  • Lucy growth strategyTobacco-free nicotine products for adults: Breakers pouches with a burstable flavor capsule, nicotine pouches and gum, FDA-approved Quit…22 growth ideas
  • Chowbus growth strategyA restaurant technology platform (POS, kiosks, online ordering, loyalty, AI marketing tools and financing) for independent, culturally…21 growth ideas
  • Growth Hacking Studio growth strategyAn Istanbul-based studio, run by founder Muhammed Tüfekyapan, that builds Growth Suite, a Built for Shopify app that shows one real…23 growth ideas
  • Sitefire growth strategyA marketing suite for the agentic web: it tracks how AI answers mention a brand, finds the pages and sources that get cited, writes briefs…23 growth ideas
  • Normal Computing growth strategyAn applied AI company building thermodynamic, physics-based AI inference hardware (Normal ASICs and reference systems) and Normal EDA, an…22 growth ideas

See every playbook and website review

Independent research. Programmable wrote this playbook from public sources, as they stood on November 10, 2026, and cites them throughout. Uplift AI did not commission, review or endorse it, and Programmable is not affiliated with Uplift AI. 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 Uplift AI runs its business.

Something wrong, or do you represent Uplift AI? 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

    Instrument the funnel, clean the sitemap, fix real-visitor speed, add a direct API-key path, a pricing page and a clear commercial licence, sector-specific sample calls and the Khan Academy story.

  2. Days 31 to 609 ideas

    Ship the Shopify COD app, self-serve concurrency, agriculture and education pages, call-then-WhatsApp fallback, template agents, a trust page, a return-cost calculator and the widget embed.

  3. Days 61 to 905 ideas

    Language pages beyond Urdu, a public listening test, a Studio-voiced social channel, and commit to the big bets: the community language programme and the voice hotline.

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 urdu voices that sound Pakistani rather than robotic. The most cited problem: COD orders that are refused or returned.

  • Roman Urdu input and code-switching between Urdu and English
  • Busy front desks and call queues
  • In their words: “COD confirmation”

Market

Competitive landscape

4 competitors compared: Robo Calls Pakistan (Devsol), Robo Call Pakistan (Punjani Inc.), eGrow and ElevenLabs.

  • Gap to exploit: Conversational (not keypad) COD confirmation that installs self-serve
  • Robo Calls Pakistan (Devsol): Keypad answers only, so it cannot capture address changes in conversation
  • Robo Call Pakistan (Punjani Inc.): IVR options, English listed as the only language

Hiring signals

Uplift AI is a team of three hiring two founding roles in Lahore, and both point at self-serve growth.

  • Founding Design Engineer: Self-serve voice agents for small-business owners within 90 days; brand and design system. Ideas 14, 18, 22.
  • Founding Engineer, Frontier Tuning Platform: Communities build production models for their own languages. Ideas 12, 15, 19.

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 firstNew accounts get 100,000 free enrichment credits with code BETA100K (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)
  • ProgrammableMediaAn AI note taker, interpreter and webinar host for your calls (opens programmable.media 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