Tesora growth playbook

Add one action, worded to match each page, that does two things: books a 30-minute call with the Head of Sales or Chief Actuary, and offers a secure upload for the rater, filing or loss triangles the page asks for (into the same single-tenant setup the security page describes, with an NDA option). Put a plain email address and the booking link in the footer and on /security, make "Give us a call" a real number, and add the same link to the .md twins so AI assistants that read them can point buyers to it. Ask three questions on the form (line of business, segment, what they would send first) so the GTM Insurance Executive starts with context.

The pages already persuade; they just end in a dead end. In a five-person company with a new GTM hire, the cheapest pipeline is the visitors who already decided to talk. Matching the button to each page's promise ("Send us your rater") turns a generic demo request into the start of a parallel run, which is how Tesora wins.

[ ] Booking link + email on homepage, /rating, /security
[ ] Page-specific button on /comparative, /mgas, /admitted, /non-admitted, /reserving, /consultants, /workbench, /benchmark, /cas
[ ] "Give us a call" shows a real number
[ ] Form: line of business, segment, what you would send first
[ ] Optional secure upload + NDA checkbox, with where files go and when they are deleted
[ ] Same link in every .md twin and the footer
[ ] One-business-day response owner
[ ] Requests tracked by page

• Buyers hesitate to upload confidential raters before an NDA

  Make the upload optional, offer the NDA first and say on the form where files go and when they are deleted.

• Low-quality requests take founder time

  Use the three form questions to route students and vendors elsewhere.

• Your websiteProduct and segment pages (fetched twice, November 9, 2026): every page ends with an invitation, "Give us a call to see how you can speed it up" (/rating, no phone number), "Start with a rater you already know" (/comparative), "Send us the rater you run today, and we will deploy it in parallel" (/mgas), "Bring your next rate change." (/admitted), "Bring last quarter's close." (/reserving), "Bring your next engagement." (/consultants) and "Book time with us." (/security), but no page has a form, upload, calendar link, phone number or contact email; the only addresses on the site are privacy@ in the privacy policy and support@ in the terms.Source
• Your websiteHomepage (checked November 9, 2026): H1 "Tesora: Frontier AI for Actuaries", "Complete complex actuarial work in half the time." and "Same headcount and standards, twice the completed work."; two proof stats read "a month, saved" ("Per actuary, measured on our own workflows") and "the output per actuary" with no figure in the page text or in its /index.md twin; closing headings "Bring us your hardest problem" and "Start with one team. Keep building." with no link, button or form; backed by Foundation Capital and Y Combinator S25.Source
• Competitor researchRivals' own homepages: Akur8, "The Global Actuarial AI Platform", "350+ insurers worldwide", agentic features, more than 150 consulting partners and "Book a free demo"; hyperexponential, "The #1 pricing and underwriting platform for commercial P&C insurance", over $75bn of premium and "Book a Demo"; Earnix, 100+ insurers, "Rate creation in under 4 days" and "Schedule a demo".See the research Source
• HiringGTM Insurance Executive (New York, $125K–$250K plus 0.50%–2.50% equity): owns the full sales cycle into P&C carriers and MGAs, selling to chief actuaries, heads of underwriting and pricing leaders, from first touch to signed contract; runs multi-stakeholder deals across actuarial, underwriting, IT and procurement; works with the founders on demos and helps build the sales playbook. The company blurb on the post still says Tesora automates insurance "from submission to bind".Source

Render the numbers as plain text in the HTML (animate on top if wanted), write them into index.md, and add a one-line method note ("measured on our own workflows" already exists; add what was measured and when). Check every other page and .md twin for the same pattern.

The headline promise ("half the time", "twice the completed work") is only as credible as the figure under it, and actuaries read for the number. Tesora built .md twins so AI assistants quote it accurately; a blank stat defeats that.

• A number invites scrutiny

  Publish the method beside it, as the benchmark page already does.

• Your websiteHomepage (checked November 9, 2026): H1 "Tesora: Frontier AI for Actuaries", "Complete complex actuarial work in half the time." and "Same headcount and standards, twice the completed work."; two proof stats read "a month, saved" ("Per actuary, measured on our own workflows") and "the output per actuary" with no figure in the page text or in its /index.md twin; closing headings "Bring us your hardest problem" and "Start with one team. Keep building." with no link, button or form; backed by Foundation Capital and Y Combinator S25.Source
• Your websiteSitemap (46 URLs, checked November 9, 2026): 23 pages, each with a .md twin: homepage, about, careers, blog, CAS, security, four segment pages (consultants, admitted, non-admitted, MGAs), five product pages (workbench, comparative, rating, reserving, best-insurance-pricing-software), the benchmark, an "AI for actuaries" guide and four how-to guides; no pricing, contact, customers, case-study, reinsurer or trust-center page.Source

Publish /trust (or link the hosted trust center) from /security, the footer and every segment page: the SOC 2 Type II report behind a short NDA request, the subprocessor list in the open, the pen-test summary on request, and a pre-filled security questionnaire. Add /trust to the sitemap and a .md twin.

Security review is the longest step in selling AI to insurers. Letting a buyer's IT team get the report the same day removes weeks from the cycle and signals maturity beyond a five-person team.

• Sharing the report too widely

  Keep it behind an NDA click-through and log each download.

• Your websiteSecurity page: "SOC 2 Type II audited" with controls monitored by Rippling, a single-tenant virtual machine per customer, an on-prem or own-account option, "SOX-aligned" and built for the ASOPs; "The current report is on the trust center", with the subprocessor list and a penetration-test summary under NDA, but the page links to no trust center and /trust returns only the site title; "Send the questionnaire over if you have one."Source
• HiringGTM Insurance Executive (New York, $125K–$250K plus 0.50%–2.50% equity): owns the full sales cycle into P&C carriers and MGAs, selling to chief actuaries, heads of underwriting and pricing leaders, from first touch to signed contract; runs multi-stakeholder deals across actuarial, underwriting, IT and procurement; works with the founders on demos and helps build the sales playbook. The company blurb on the post still says Tesora automates insurance "from submission to bind".Source

Rewrite the blurb on the YC job posts, LinkedIn and any directory profiles to match the site: the actuarial workbench, the segments served, CAS Executive Partner status and the benchmark. Keep one approved 50-word and 150-word description and reuse it everywhere.

A consistent description helps the right actuaries and engineers apply, stops AI assistants and directories mixing old and new positioning, and makes the company look settled in its market after a pivot.

• Old listings persist

  Prioritise the profiles candidates and buyers actually see.

• HiringGTM Insurance Executive (New York, $125K–$250K plus 0.50%–2.50% equity): owns the full sales cycle into P&C carriers and MGAs, selling to chief actuaries, heads of underwriting and pricing leaders, from first touch to signed contract; runs multi-stakeholder deals across actuarial, underwriting, IT and procurement; works with the founders on demos and helps build the sales playbook. The company blurb on the post still says Tesora automates insurance "from submission to bind".Source
• Y Combinator profileCompany page: "Frontier AI for Actuaries"; AI agents for actuaries "at P&C carriers, reinsurers, and MGAs" that pull data, run methods and deploy raters while actuaries review; deploys inside the client's stack; team of 5, San Francisco, S25 batch.Source
• Your websiteSite history: in November 2025 the homepage headline was "Tesora - Automated Supplier Discovery & Due Diligence"; it is now "Tesora: Frontier AI for Actuaries"; domain registered February 2025; no newsletter, live chat, testimonials or case-study section found.Source

Log every booth conversation the same day (segment, line, the close pain they described). Within a week send a short note that repeats their pain in their words, links the reserving case study, and offers to run their year-end close in parallel with a fixed scope. Invite the ones who are not ready to review benchmark tasks instead.

Event leads decay fast; a follow-up that quotes the actuary's own problem and offers a concrete, timed run beats a generic deck, and year-end gives a natural deadline.

• Too many runs for a small team at year-end

  Cap at three and schedule the rest for Q1 close.

• #1 Give every "send us your rater" invitation somewhere to go: a secure upload and a booking link on every page that asks for one

• Your websiteCAS page: Tesora sponsors the CAS Annual Meeting, November 8 to 11, 2026, Hilton Hawaiian Village, Honolulu, at booth 9; "No demo unless you want one" and "We'd rather hear what your last close was like."Source
• Your websiteReserving page: triangles on ingest, chain ladder and Bornhuetter-Ferguson side by side, anomalies escalated before factor selection; case study: a regional carrier's year-end close where the workbench flagged that gross came in below net in the supplied data before any selection, and the actuary reached year-over-year review "days earlier than in prior studies"; CTA "Bring last quarter's close."Source

Find the largest element on the homepage (likely the hero or an animated stat), serve it in the first HTML response, preload the main font, and defer any animation or chart scripts until after first paint. Recheck the field data after four weeks.

Buyers from carriers often browse on locked-down corporate laptops and networks; a fast text page also feeds search and the crawler-readable figures from idea 2.

• Low traffic makes field data noisy

  Track lab results weekly alongside it.

• Your websiteChrome UX Report field data for tesora.ai (real visitors, 75th percentile): Largest Contentful Paint 3,021 ms, above Google's 2,500 ms "good" line and 63% slower than the previous period; Interaction to Next Paint 109 ms and Cumulative Layout Shift 0.07, both good.Source
• Your websiteHomepage (checked November 9, 2026): H1 "Tesora: Frontier AI for Actuaries", "Complete complex actuarial work in half the time." and "Same headcount and standards, twice the completed work."; two proof stats read "a month, saved" ("Per actuary, measured on our own workflows") and "the output per actuary" with no figure in the page text or in its /index.md twin; closing headings "Bring us your hardest problem" and "Start with one team. Keep building." with no link, button or form; backed by Foundation Capital and Y Combinator S25.Source

Write a 1,200-word post on that catch: what the data looked like, how the anomaly check works, what a general assistant would have done, and what the actuary did next. Anonymise fully, end with "Bring last quarter's close" and the new contact path, and share it on LinkedIn and with CAS contacts.

Actuaries trust specifics. A concrete near-miss story shows the audit agents' value better than any feature list and gives the sales team something to send.

• Customer recognisable

  Change line, size and dates; get written approval.

• Your websiteGuides and blog: four how-to guides (SERFF objection responses, Excel rater to API, GLM pricing in Python, Bornhuetter-Ferguson), each roughly 500 to 600 words with no author, date or call to action; the "AI for actuaries" guide has no call to action either; the blog "Actuarially Intelligent" promises "The problems we ran into putting frontier models on actuarial work" but lists no posts yet.Source
• Your websiteReserving page: triangles on ingest, chain ladder and Bornhuetter-Ferguson side by side, anomalies escalated before factor selection; case study: a regional carrier's year-end close where the workbench flagged that gross came in below net in the supplied data before any selection, and the actuary reached year-over-year review "days earlier than in prior studies"; CTA "Bring last quarter's close."Source

Build /customers with each story as a one-page case: segment, line, starting point, what Tesora did, measured result (hours, days, factors found off-market, premium or profit impact from the comparative study) and a quote from the actuary, named or anonymous. Link each from its segment and product page.

Insurance buyers buy what peers already use. Two specific, numbered stories from regional carriers are more persuasive to a mid-size carrier or MGA than a large logo wall.

• Carriers decline to be named

  Publish anonymised with segment, size band and line.

• Your websiteComparative raters page: rebuilds a rater, "yours or a competitor's", from Excel or SERFF filing PDFs "in hours", drawing on "over a decade of public SERFF rate and rule filings"; case study: "A regional carrier asked us to rebuild its own rater for one line in one state, along with its competitors'" and "left with full visibility into which of its factors were competitive"; no financial result given; CTA "Start with a rater you already know."Source
• Your websiteSegment pages: MGAs, captives and RRGs ("Pick one program and we will start there"), consultants ("Each client's data stays separate from every other client's", "Bring your next engagement", no partner or reseller offer), admitted carriers ("every rate change tested against the market before it is filed") and E&S carriers; none shows a customer, number or quote.Source
• Competitor researchRivals' own homepages: Akur8, "The Global Actuarial AI Platform", "350+ insurers worldwide", agentic features, more than 150 consulting partners and "Book a free demo"; hyperexponential, "The #1 pricing and underwriting platform for commercial P&C insurance", over $75bn of premium and "Book a Demo"; Earnix, 100+ insurers, "Rate creation in under 4 days" and "Schedule a demo".See the research Source

Add two short forms: "Email me the results" (name, company type, role) and "Review tasks" for credentialed actuaries. Add a one-question poll for carriers: "Which task would you hand to a model first?" Publish the poll results with the scores. Send results to the list first, with a link to try the workbench on their own data.

The scores release is a one-time news moment; a list built beforehand turns it into pipeline. The poll gives the sales team each company's first use case.

• List is mostly researchers

  Ask company type and route only carriers and MGAs to sales.

• Your websiteTesora Actuarial Benchmark (November 2026): "a set of 92 actuarial tasks, with answer keys and scorers" on synthetic data; first release has 23 tasks; first runs cover 18 models from nine labs; "Scores will follow once the first runs are reviewed"; a human baseline of credentialed actuaries answering 180 instances is planned; tasks, generators and scorers to follow under an open license; it invites actuaries to review tasks, carriers to say which work they would hand to a model first and researchers to run the suite, with no sign-up or email capture.Source

Recruit 20 to 30 credentialed actuaries from carriers, MGAs and consultancies (not students) for the baseline, with a fee or a charity donation, credit in the published results, and an early look at the workbench. After the baseline, invite each to try their own data. Track how many bring their team.

People trust what they helped measure. A participant who scored the tasks can explain the workbench internally better than any salesperson, and the published list of contributors signals credibility.

• Employers restrict participation

  Use personal time, allow anonymous credit and get permission up front.

• Selection looks biased toward prospects

  Publish how participants were chosen.

• #9 Add a "get the results first" list and a task-reviewer signup to the benchmark page

• Your websiteTesora Actuarial Benchmark (November 2026): "a set of 92 actuarial tasks, with answer keys and scorers" on synthetic data; first release has 23 tasks; first runs cover 18 models from nine labs; "Scores will follow once the first runs are reviewed"; a human baseline of credentialed actuaries answering 180 instances is planned; tasks, generators and scorers to follow under an open license; it invites actuaries to review tasks, carriers to say which work they would hand to a model first and researchers to run the suite, with no sign-up or email capture.Source
• Your websiteAbout page: a five-person team: a cofounder CEO (former PE investor and McKinsey consultant), a cofounder CTO (Stanford AI researcher), a Chief Actuary & CPO who was chief actuary at W. R. Berkley E&S and a capital modeler at Gen Re, a Head of Sales with 25+ years in actuarial software (most recently Akur8 and Milliman) and a founding engineer with three actuarial exams; Tesora became an Executive Partner of the Casualty Actuarial Society for 2026-2027 from October 1, 2026.Source

Build a public sandbox on synthetic data from the benchmark generators: a reserving triangle and a small rater, each with a hidden error. Visitors pick one, run the audit agents, and see the flag, the citation and the fix. Let them add their own planted error to the synthetic file. End with "Now try it on your book" and the contact path.

Actuaries are sceptical of AI and believe what they can test. A synthetic sandbox needs no NDA or security review, so it lets a buyer feel the product before procurement starts.

• A public miss

  Show misses honestly with what the actuary would do; that builds trust.

• Cost of anonymous runs

  Require a work email after the first run.

• #1 Give every "send us your rater" invitation somewhere to go: a secure upload and a booking link on every page that asks for one

• Your websiteWorkbench page: a coordinator, specialist agents and "read-only review agents"; an MCP server so actuaries can call Tesora from the AI assistants they already use, with their own permissions; works with Excel, Python or R; "paid for results, not tokens", with trials but no prices; a challenge, "Try to break it": submit an analysis with a planted mistake and see which audit agent catches it; no customer names or results.Source
• Your websiteTesora Actuarial Benchmark (November 2026): "a set of 92 actuarial tasks, with answer keys and scorers" on synthetic data; first release has 23 tasks; first runs cover 18 models from nine labs; "Scores will follow once the first runs are reviewed"; a human baseline of credentialed actuaries answering 180 instances is planned; tasks, generators and scorers to follow under an open license; it invites actuaries to review tasks, carriers to say which work they would hand to a model first and researchers to run the suite, with no sign-up or email capture.Source

For each target carrier or MGA, rebuild their filed rater and two competitors' for one line in one state from public filings, and send a two-page snapshot: three factors where they are most out of line with the market, with citations to the filing pages. Offer the full comparative study as the paid next step.

It turns a cold email into useful work for a pricing actuary, shows the citation-to-source promise in action and needs no data-sharing agreement. It also gives the new GTM executive a repeatable opener for the playbook they are hired to build.

• Seen as presumptuous about a carrier's pricing

  Frame it as a market view from public filings with citations, not advice.

• Filing access terms

  Check each state's public-access terms before reusing filings in outreach.

• Your websiteComparative raters page: rebuilds a rater, "yours or a competitor's", from Excel or SERFF filing PDFs "in hours", drawing on "over a decade of public SERFF rate and rule filings"; case study: "A regional carrier asked us to rebuild its own rater for one line in one state, along with its competitors'" and "left with full visibility into which of its factors were competitive"; no financial result given; CTA "Start with a rater you already know."Source
• HiringGTM Insurance Executive (New York, $125K–$250K plus 0.50%–2.50% equity): owns the full sales cycle into P&C carriers and MGAs, selling to chief actuaries, heads of underwriting and pricing leaders, from first touch to signed contract; runs multi-stakeholder deals across actuarial, underwriting, IT and procurement; works with the founders on demos and helps build the sales playbook. The company blurb on the post still says Tesora automates insurance "from submission to bind".Source

Define three pilots, one per entry point: Close (one quarter, up to N lines), Rater (one filed rater to API, in parallel), Competitive (one line, one state, two rivals). Each has a fixed fee credited against the first year, a four- to eight-week timeline, data needed and two success measures agreed in writing (for example days to year-over-year review, premium match to filed). Publish a one-page summary on each product page.

A defined pilot shortens procurement, helps the forward-deployed engineer plan, and makes "paid for results" concrete. A small fee filters tyre-kickers and makes conversion to annual a decision about results, not a new negotiation.

• Fixed scope too narrow for complex carriers

  Allow a priced extension, not a free one.

• #1 Give every "send us your rater" invitation somewhere to go: a secure upload and a booking link on every page that asks for one

• Your websiteSegment pages: MGAs, captives and RRGs ("Pick one program and we will start there"), consultants ("Each client's data stays separate from every other client's", "Bring your next engagement", no partner or reseller offer), admitted carriers ("every rate change tested against the market before it is filed") and E&S carriers; none shows a customer, number or quote.Source
• Your websiteWorkbench page: a coordinator, specialist agents and "read-only review agents"; an MCP server so actuaries can call Tesora from the AI assistants they already use, with their own permissions; works with Excel, Python or R; "paid for results, not tokens", with trials but no prices; a challenge, "Try to break it": submit an analysis with a planted mistake and see which audit agent catches it; no customer names or results.Source
• HiringForward Deployed 'Actuarial' Engineer (San Francisco or New York, $100K–$200K plus 0.50%–2.00% equity): maps actuaries' and underwriters' rating, reserving and pricing workflows, builds models, pipelines and features, and owns "deployments end to end, from first conversation to production use" for specialty P&C carriers and MGAs, to "rebuild the underwriting and actuarial workflows that the industry still runs on spreadsheets".Source

Add /pricing explaining the model: what counts as a result (a deployed rater, a closed quarter, a competitive study), how pilots are priced and credited, what is included (dedicated VM, support, audit trail) and typical ranges by team size. Add a cost question to the homepage FAQ.

Pricing pages are among the most visited pages on B2B sites; even a range lets a pricing lead get budget approval before the first call and filters out mismatched buyers.

• Ranges anchor negotiations low

  Use "from" pricing and tie it to scope.

• #13 Package the parallel run as a fixed-scope, fixed-price pilot with success criteria agreed up front

• Your websiteWorkbench page: a coordinator, specialist agents and "read-only review agents"; an MCP server so actuaries can call Tesora from the AI assistants they already use, with their own permissions; works with Excel, Python or R; "paid for results, not tokens", with trials but no prices; a challenge, "Try to break it": submit an analysis with a planted mistake and see which audit agent catches it; no customer names or results.Source
• Your websiteSitemap (46 URLs, checked November 9, 2026): 23 pages, each with a .md twin: homepage, about, careers, blog, CAS, security, four segment pages (consultants, admitted, non-admitted, MGAs), five product pages (workbench, comparative, rating, reserving, best-insurance-pricing-software), the benchmark, an "AI for actuaries" guide and four how-to guides; no pricing, contact, customers, case-study, reinsurer or trust-center page.Source

Add author (with credentials), reviewed date and a closing next step to every guide. Deepen each to a working reference: a worked example, a downloadable template (an objection-response letter skeleton, a sample triangle with BF worked through, GLM notebook) and a short FAQ. Add a guide a month on topics next to the product: loss trend selection, on-level premium, Cape Cod, rate-filing exhibits.

Actuarial searches are low-volume but high-intent, and the reader is often the buyer. Signed, dated, deeper guides rank better, get cited by AI assistants and give each reader a reason to come back.

• Founder time

  Draft from engagement notes and have the Chief Actuary review rather than write.

• Your websiteGuides and blog: four how-to guides (SERFF objection responses, Excel rater to API, GLM pricing in Python, Bornhuetter-Ferguson), each roughly 500 to 600 words with no author, date or call to action; the "AI for actuaries" guide has no call to action either; the blog "Actuarially Intelligent" promises "The problems we ran into putting frontier models on actuarial work" but lists no posts yet.Source
• Your websiteAbout page: a five-person team: a cofounder CEO (former PE investor and McKinsey consultant), a cofounder CTO (Stanford AI researcher), a Chief Actuary & CPO who was chief actuary at W. R. Berkley E&S and a capital modeler at Gen Re, a Head of Sales with 25+ years in actuarial software (most recently Akur8 and Milliman) and a founding engineer with three actuarial exams; Tesora became an Executive Partner of the Casualty Actuarial Society for 2026-2027 from October 1, 2026.Source

Write one page per common incumbent (start with Akur8, Earnix and WTW Radar): what the incumbent does well, the gaps around it (competitor raters from filings, rater-to-API, close), how data moves between them, and a joint workflow example. Keep it factual, dated and sourced, and run it past legal for trademark use. The Head of Sales knows the Akur8 side from experience.

"Alongside" pages capture high-intent searches from teams who already pay for pricing software, avoid a rip-and-replace fight and give sales a ready answer to "we already have Akur8".

• Former-employer sensitivities

  Use only public information and keep the tone generous.

• Your website"Actuarial software in P&C" page: maps each problem to specialist tools (Akur8, WTW Radar, Earnix, SAS, hyperexponential, Guidewire PricingCenter) and to Tesora, says specialist tools "usually solve only one" problem, offers a vendor demo checklist ("including us"), and says "Tesora works alongside it"; no author or date.Source
• Competitor researchRivals' own homepages: Akur8, "The Global Actuarial AI Platform", "350+ insurers worldwide", agentic features, more than 150 consulting partners and "Book a free demo"; hyperexponential, "The #1 pricing and underwriting platform for commercial P&C insurance", over $75bn of premium and "Book a Demo"; Earnix, 100+ insurers, "Rate creation in under 4 days" and "Schedule a demo".See the research Source

Build a five-page packet: one page each for the chief actuary (results, ASOPs, sign-off stays with the actuary), IT (architecture, SSO, data residency, on-prem), risk and audit (SOX role separation, audit trail, model version per change), procurement (pilot terms, pricing model, references) and a mutual action plan template. Share it as a single link after every first call.

Enterprise deals stall when the champion cannot answer colleagues' questions. A ready packet lets the actuary sell for you and gives the new sales hire a repeatable motion for the playbook they are building.

• Packet goes stale

  Date it and review monthly.

• #3 Link the trust center from the security page and make /trust a real page with a report request

• HiringGTM Insurance Executive (New York, $125K–$250K plus 0.50%–2.50% equity): owns the full sales cycle into P&C carriers and MGAs, selling to chief actuaries, heads of underwriting and pricing leaders, from first touch to signed contract; runs multi-stakeholder deals across actuarial, underwriting, IT and procurement; works with the founders on demos and helps build the sales playbook. The company blurb on the post still says Tesora automates insurance "from submission to bind".Source
• Your websiteSecurity page: "SOC 2 Type II audited" with controls monitored by Rippling, a single-tenant virtual machine per customer, an on-prem or own-account option, "SOX-aligned" and built for the ASOPs; "The current report is on the trust center", with the subprocessor list and a penetration-test summary under NDA, but the page links to no trust center and /trust returns only the site title; "Send the questionnaire over if you have one."Source

Offer three to five boutique actuarial consultancies a partner tier: discounted seats for client engagements, co-branded deliverables, training, a listing on a /partners page and a referral fee or revenue share when a client licenses Tesora directly. Start with firms serving MGAs, captives and RRGs.

Consultants are trusted advisers to exactly the carriers that lack in-house tools; when their deliverables cite Tesora, every engagement becomes a demo.

• Channel conflict with direct sales

  Register deals and pay referral fees on direct conversions.

• Your websiteSegment pages: MGAs, captives and RRGs ("Pick one program and we will start there"), consultants ("Each client's data stays separate from every other client's", "Bring your next engagement", no partner or reseller offer), admitted carriers ("every rate change tested against the market before it is filed") and E&S carriers; none shows a customer, number or quote.Source
• Competitor researchRivals' own homepages: Akur8, "The Global Actuarial AI Platform", "350+ insurers worldwide", agentic features, more than 150 consulting partners and "Book a free demo"; hyperexponential, "The #1 pricing and underwriting platform for commercial P&C insurance", over $75bn of premium and "Book a Demo"; Earnix, 100+ insurers, "Rate creation in under 4 days" and "Schedule a demo".See the research Source
• Your websiteAbout page: a five-person team: a cofounder CEO (former PE investor and McKinsey consultant), a cofounder CTO (Stanford AI researcher), a Chief Actuary & CPO who was chief actuary at W. R. Berkley E&S and a capital modeler at Gen Re, a Head of Sales with 25+ years in actuarial software (most recently Akur8 and Milliman) and a founding engineer with three actuarial exams; Tesora became an Executive Partner of the Casualty Actuarial Society for 2026-2027 from October 1, 2026.Source

After each deployment, save what was learned as a starter kit by line and segment: data mappings for common loss-run and triangle formats, exhibit templates, standard checks, and a deployment checklist. Start every new pilot from the closest kit and measure days to first result. Feed recurring gaps to the product roadmap.

Forward-deployed models scale only when each project leaves something behind. Faster time to first result makes pilots cheaper to run and easier to win.

• Customer data leaks into kits

  Kits hold structure and templates only, never customer data.

• #13 Package the parallel run as a fixed-scope, fixed-price pilot with success criteria agreed up front

• HiringForward Deployed 'Actuarial' Engineer (San Francisco or New York, $100K–$200K plus 0.50%–2.00% equity): maps actuaries' and underwriters' rating, reserving and pricing workflows, builds models, pipelines and features, and owns "deployments end to end, from first conversation to production use" for specialty P&C carriers and MGAs, to "rebuild the underwriting and actuarial workflows that the industry still runs on spreadsheets".Source
• Your websiteWorkbench page: a coordinator, specialist agents and "read-only review agents"; an MCP server so actuaries can call Tesora from the AI assistants they already use, with their own permissions; works with Excel, Python or R; "paid for results, not tokens", with trials but no prices; a challenge, "Try to break it": submit an analysis with a planted mistake and see which audit agent catches it; no customer names or results.Source

Package the MCP connection as a self-serve starting point: list it in the major assistant connector directories, publish a setup guide, and offer a free tier on synthetic or public filing data (for example, "rebuild this public rater and cite every factor"). When a user wants their own book, route them to a pilot. Track which firms connect and hand warm accounts to sales.

Meeting actuaries inside tools they already open every day lowers the first step from a procurement process to a connection, while Tesora keeps the audited, cited core that general assistants lack.

• Users bring confidential data to the free tier

  Block uploads in the free scope; own data only in a dedicated environment.

• Undercuts the "harness, not chat" message

  Show the citations and audit trail in every response.

• #11 Make "Try to break it" a real self-serve demo: a synthetic book with planted mistakes anyone can run

• Your websiteWorkbench page: a coordinator, specialist agents and "read-only review agents"; an MCP server so actuaries can call Tesora from the AI assistants they already use, with their own permissions; works with Excel, Python or R; "paid for results, not tokens", with trials but no prices; a challenge, "Try to break it": submit an analysis with a planted mistake and see which audit agent catches it; no customer names or results.Source
• Your websiteHomepage (checked November 9, 2026): H1 "Tesora: Frontier AI for Actuaries", "Complete complex actuarial work in half the time." and "Same headcount and standards, twice the completed work."; two proof stats read "a month, saved" ("Per actuary, measured on our own workflows") and "the output per actuary" with no figure in the page text or in its /index.md twin; closing headings "Bring us your hardest problem" and "Start with one team. Keep building." with no link, button or form; backed by Foundation Capital and Y Combinator S25.Source

Release the full 92 tasks openly, host a leaderboard updated with each major model release, accept outside submissions with reproducible runs, and publish a short quarterly report for chief actuaries: what models can now be trusted with and what still needs an actuary. Present it at CAS events as an Executive Partner and invite carriers to contribute anonymised task ideas.

A neutral, well-run benchmark earns press, academic citations and invitations to speak, and positions Tesora as the people who know where AI is safe in actuarial work, which is the buyer's main question.

• Seen as marketing, not science

  Publish method first, accept outside runs and include Tesora's own harness on equal terms.

• Models saturate the tasks

  Keep a held-out set and add the harder tiers over time.

• #9 Add a "get the results first" list and a task-reviewer signup to the benchmark page
• #10 Recruit the benchmark's human baseline from working carrier actuaries, and they become your first champions

• Your websiteTesora Actuarial Benchmark (November 2026): "a set of 92 actuarial tasks, with answer keys and scorers" on synthetic data; first release has 23 tasks; first runs cover 18 models from nine labs; "Scores will follow once the first runs are reviewed"; a human baseline of credentialed actuaries answering 180 instances is planned; tasks, generators and scorers to follow under an open license; it invites actuaries to review tasks, carriers to say which work they would hand to a model first and researchers to run the suite, with no sign-up or email capture.Source
• Your websiteAbout page: a five-person team: a cofounder CEO (former PE investor and McKinsey consultant), a cofounder CTO (Stanford AI researcher), a Chief Actuary & CPO who was chief actuary at W. R. Berkley E&S and a capital modeler at Gen Re, a Head of Sales with 25+ years in actuarial software (most recently Akur8 and Milliman) and a founding engineer with three actuarial exams; Tesora became an Executive Partner of the Casualty Actuarial Society for 2026-2027 from October 1, 2026.Source

Start with one or two lines (for example personal auto and homeowners) in a handful of states: publish filed rate changes, effective dates and the most changed factors, each cited to the filing page. Offer a free monthly email digest by line and state, and a paid upgrade: a full rebuilt rater and your book priced through it.

A monthly digest of competitors' filings is useful to every pricing team, builds a list of exactly the right buyers and shows the citation-to-source promise at scale.

• Filing reuse terms vary by state

  Confirm terms first and start only where reuse is clearly allowed.

• Errors in a public tracker

  Cite every figure and correct openly.

• #12 Open outbound with a free competitor snapshot: where a carrier's filed rates sit against two rivals in one state

• Your websiteComparative raters page: rebuilds a rater, "yours or a competitor's", from Excel or SERFF filing PDFs "in hours", drawing on "over a decade of public SERFF rate and rule filings"; case study: "A regional carrier asked us to rebuild its own rater for one line in one state, along with its competitors'" and "left with full visibility into which of its factors were competitive"; no financial result given; CTA "Start with a rater you already know."Source
• Your websiteSite history: in November 2025 the homepage headline was "Tesora - Automated Supplier Discovery & Due Diligence"; it is now "Tesora: Frontier AI for Actuaries"; domain registered February 2025; no newsletter, live chat, testimonials or case-study section found.Source

Test demand with one /reinsurers page: treaty pricing from cedent data, loss-run and triangle ingestion from many cedents with anomaly flags, experience and exposure rating with every value cited, and a pilot offer ("Send us one cedent submission"). Run ten conversations with reinsurer actuaries before building anything new.

Reinsurers receive messy data from many cedents and price it under time pressure, which is exactly where cited, audited ingestion pays off. One page costs little and tells Tesora whether the segment is real.

• Spreading a small team too thin

  Cap at one pilot until a second segment is proven.

• Y Combinator profileCompany page: "Frontier AI for Actuaries"; AI agents for actuaries "at P&C carriers, reinsurers, and MGAs" that pull data, run methods and deploy raters while actuaries review; deploys inside the client's stack; team of 5, San Francisco, S25 batch.Source
• Your websiteSitemap (46 URLs, checked November 9, 2026): 23 pages, each with a .md twin: homepage, about, careers, blog, CAS, security, four segment pages (consultants, admitted, non-admitted, MGAs), five product pages (workbench, comparative, rating, reserving, best-insurance-pricing-software), the benchmark, an "AI for actuaries" guide and four how-to guides; no pricing, contact, customers, case-study, reinsurer or trust-center page.Source
• Your websiteAbout page: a five-person team: a cofounder CEO (former PE investor and McKinsey consultant), a cofounder CTO (Stanford AI researcher), a Chief Actuary & CPO who was chief actuary at W. R. Berkley E&S and a capital modeler at Gen Re, a Head of Sales with 25+ years in actuarial software (most recently Akur8 and Milliman) and a founding engineer with three actuarial exams; Tesora became an Executive Partner of the Casualty Actuarial Society for 2026-2027 from October 1, 2026.Source

Offer a free student workbench on synthetic data only: worked exam-style problems with every figure cited, the planted-mistake challenge, and a certificate for completing a set. Partner with a few university actuarial programs and student CAS chapters. Use it as a hiring pipeline and ask participants to share it with their employer's pricing or reserving team.

Students spread tools fast, employers notice what new hires already know, and Tesora gets a recruiting channel without job ads. Using synthetic data keeps it free of confidentiality issues.

• Seen as an exam shortcut

  Position it as practice with cited workings, not answers to live exams.

• #11 Make "Try to break it" a real self-serve demo: a synthetic book with planted mistakes anyone can run

• Your websiteCareers page: "Remote, New York hub"; roles listed on Y Combinator Work at a Startup; "We hire ahead of the role more often than not"; students can "book a quick chat and tell us your school, program and exam progress".Source
• Your websiteTesora Actuarial Benchmark (November 2026): "a set of 92 actuarial tasks, with answer keys and scorers" on synthetic data; first release has 23 tasks; first runs cover 18 models from nine labs; "Scores will follow once the first runs are reviewed"; a human baseline of credentialed actuaries answering 180 instances is planned; tasks, generators and scorers to follow under an open license; it invites actuaries to review tasks, carriers to say which work they would hand to a model first and researchers to run the suite, with no sign-up or email capture.Source