Domain layer over a frontier model

Here is the part the demo skips. Most vertical tools run on the same OpenAI or Anthropic models as your general assistant. The brain is rented. What you pay for is the layer on top: proprietary data, workflow, integrations, evaluation and accountability.

• The intelligence is often rented
• The domain layer is the product
• Judge the layer, not the demo

Price, lock-in and overlap

Specialists cost more. They embed your data deeply. And they often overlap with the general assistant you already pay for. So the call is economic and architectural. Capability is only one input.

• Your Copilot may already do 80%
• Migration is rarely easy
• Buy only for the part that clears the bar

Harvey Legal

Agents for law firms and enterprise legal teams, grounded in case law, contracts and firm knowledge. Harvey is the most active vendor here, with roughly 28 posts between September 3 and September 17 alone. Its homepage, read October 1, 2026, claims more than 200,000 professionals, 3,000 or more firms and in-house teams, more than 70 countries and 80 or more AmLaw 100 firms (vendor claims). Two of those moved in two weeks: firms from 2,400 to 3,000, AmLaw 100 from 75 to 80. So read the homepage again on the day you quote it. What shipped: Agentic Search across iManage, SharePoint and Google Drive went generally available on September 30. What hasn’t: Harvey II, announced August 18, carries no availability language at all. And the Everlaw MCP integration is still “are building.” On data, Harvey says it doesn’t train on your work. It doesn’t say how long it keeps it. And there’s no price anywhere.
· Harvey’s training wording: “We don’t use inputs, outputs, or uploaded documents to train underlying models.”
· harvey.ai/pricing returns a 404, read October 1, 2026.

• 3,000 or more firms and teams, 80 or more AmLaw 100 (vendor claims)
• No training on inputs, outputs or uploaded documents
• No published customer-data retention period
• Harvey II has no availability language and no price

Glean Enterprise knowledge

Work AI that unifies permission-aware search, an assistant and agents across your company apps. Start with a small tell. Glean’s own pages disagree about its connector count: “more than 250” on the platform overview, “275+” on the pricing page. Both are undated. Quote whichever you read, with the date. That pricing page carries no plan names, no prices, no seat model and no minimum. Just “Get a demo.” Glean’s May 28, 2026 press release puts the business at $300M ARR (vendor claim). August 26 brought six launch posts in one day. Read Transform’s status carefully, because the launch doesn’t. Glean’s navigation lists it “Coming soon,” and glean.com/products/transform is still a 404. September brought something shipped instead: interactive artifacts on September 9, available the same day. On September 29 Glean joined OpenAI’s B2B marketplace. And its release notes are good news for you. Each release has its own dated, stable page, back to April 2025.
· Transform, in Glean’s words: “an early preview of what we’re building.”

• Two Glean pages disagree on connectors: 250+ and 275+
• Dated per-release notes you can cite, back to April 2025
• No published prices, plan names or seat model
• Glean Transform is “Coming soon,” not available

Sierra Agents that act

Sierra began in customer experience, across chat, voice, email and messaging, priced on outcomes. It has pushed well past that. Horizon, announced July 16, targets goals that run for days, weeks or months, such as loan origination and prior authorization. September added multimodal agents on the 10th, AIUC-1 certification on the 17th and Ghostwriter on the 28th, now a named product rather than a demo. On price, say it plainly: there’s no rate card. Outcome pricing exists only as prose. No per-resolution rate, no platform fee, no minimum, no billing term. On September 29 Sierra joined the OpenAI Marketplace. That’s still not a rate. But it’s the first published way anyone can actually pay Sierra. Two numbers you’ve probably seen aren’t Sierra’s. Neither 40% of the Fortune 50 nor a roughly $15.8B valuation appears anywhere on its site. On data, Sierra publishes one good sentence and nothing more.
· Sierra’s data wording: “Your data is only used as you instruct, and is never shared with other customers.”
· sierra.ai/pricing returns a 404, read October 1, 2026.

• Joined the OpenAI Marketplace September 29, 2026
• No rate card: sierra.ai/pricing is a 404
• Data used only as instructed, never shared between customers
• The 40% and $15.8B figures are not first-party

OpenEvidence Clinical evidence

A medical answer engine grounded in licensed clinical literature, free to clinicians and deployed across health systems. Start with the contradiction, because it decides whether you can use this at all. The privacy policy says user questions and prompts may be sold or monetized as a commercial data feed, and that user inputs aren’t covered by HIPAA. The vendor’s April 25, 2025 announcement says it’s fully HIPAA compliant. Both are first-party. Both are live. The announcement records what OpenEvidence said in 2025. The policy is the live page, so treat it as governing until OpenEvidence confirms otherwise in writing. Ask for the business associate agreement, which isn’t published. The rest is confirmable. NEJM Group content agreement, February 19, 2025. JAMA Network, June 5, 2025. DeepConsult and a $210M round at a $3.5B valuation, July 15, 2025. Nothing on the announcements index carries a 2026 date, so don’t call that funding recent.
· The policy wording: “Your user data, including your questions/prompts, may be sold to third parties or otherwise monetized as a commercial data feed.”

• Privacy policy: questions and prompts may be sold as a data feed
• Same policy: user inputs are not covered by HIPAA
• April 25, 2025 announcement: fully HIPAA compliant
• Free at the point of use: no pricing page or enterprise plan

Read the live page before the launch post

A dated article is authoritative for what was announced and when. A live page is authoritative for what is true today. When they disagree, the live page governs the present and the article governs the record. Keep your own dated copy of the live page, because it carries no history.

Get the price in writing

Here a written quote is usually the only price that exists. Six of the eight vendors publish nothing usable. Cognition’s list stops at Enterprise, where Agent Compute Units bill at an unpublished order-form rate. And tell a broken pricing page from a deliberate one before you draw a conclusion.

Ask which surface it arrives on

Hebbia arrives inside Claude, ChatGPT and Cursor over MCP. Harvey arrives inside Microsoft 365 Copilot. Each time, ask whose permission model, data terms and audit trail apply there. And ask whether the embedding vendor’s model terms match the ones the model maker publishes.

Rented intelligence

Many verticals wrap OpenAI or Anthropic. Confirm the domain layer is real before you pay specialist prices for someone else’s model.

Vendor comparisons name competitors now

Glean’s August 26 post claims it “saves 81% on token costs and is preferred 78% of the time versus Claude Cowork.” Those are Glean’s figures about a named rival, published in Glean’s own marketing. Treat any head-to-head number from a vendor as a hypothesis to test on your corpus. Stay just as skeptical when a competitor answers with numbers of its own.

The terms that need a human with a browser

Four Glean pages decide Glean’s data terms, and a tool can’t open any of them. glean.com/legal/dpa, glean.com/legal/ai-addendum and glean.com/legal/security-standard are blocked to automated access by Glean’s own robots.txt. trust.glean.com builds its content in the browser, so anything else gets metadata and nothing more. Those four pages are where the answer sits on whether Glean trains on your data, how long it keeps it and when it deletes it. No other page on Glean’s estate answers those questions. Two more trust centers behave the same way, Sierra’s and Abridge’s, and Abridge keeps its model cards, network diagram, pentest summary and security whitepaper behind an access request on top. So put this on a person’s calendar, not in a backlog. Open the four Glean pages. Ask Glean for the Limited Retention Addendum its own legal index lists. Request the gated documents from Sierra and Abridge. It’s an hour of someone’s time and it decides whether the contract is signable.

Announced is not available

Five things on this map have been announced without becoming buyable, and the day counts are getting hard to defend. Harvey II, introduced August 18, 2026 with no availability language of any kind and no pricing: forty-four days, and the September changelog still doesn’t mention it. Hebbia Max, “rolling out to a small set of firms first” since July 30: sixty-three days, now marketed as a flagship with no general-availability statement and no retraction of the gating line. Glean agent identity, in public beta since July 15 and still beta-bannered in documentation updated September 30: seventy-eight days. Glean Transform, launched with fanfare on August 26, listed “Coming soon,” called an early preview by Glean itself, and still a 404 at its own product address: thirty-six days. And Harvey and Everlaw, who said on September 3 that they “are building” an MCP integration: twenty-eight days, same wording. None of these belongs in a dated rollout plan. If one of them is load-bearing in yours, ask for a date in writing and read “soon” as a no.

Consolidation, at speed

At least eight deals across these eight vendors since July 10, 2026. In the seventeen days to July 27: Cognition bought TierZero and welcomed The Interaction Company, Harvey bought Benchmark, Sierra bought Takeoff, Abridge took on the Altrina team. On August 13, Cursor acquired Firetiger, which builds agents that work on software once it reaches production. No terms were disclosed. On August 14, Cursor announced that it “has officially been acquired by SpaceX,” completing a process begun in April. Again no valuation and no terms. On September 10, the Dioxus team joined Cognition, no terms. Money moved too. Harvey took a strategic investment from Goldman Sachs and J.P. Morgan on July 28 with no new valuation published. Cognition announced a Series E on September 8 it says raised over $2B at a $48B valuation. One of these eight vendors is now owned by a launch company. That sharpens the point. It doesn’t soften it. The specialist you sign with may not merely be a different company in a year. It may sit inside a group whose priorities have nothing to do with your codebase.

A missing price is not a discontinued product

Cursor’s pricing page publishes an entry price per group rather than a price per tier: one “$20 / mo.” over Pro, Pro+ and Ultra, one “$40 / user / mo.” over Standard and Premium. Those cards carry no placeholder and no loading state. So a missing per-tier price is a layout choice, and nothing is failing to render. It isn’t a discontinuation notice either. Cursor Start and Cursor for iPad aren’t listed at all. Whether Start was withdrawn or is geo-gated to visitors in India can’t be determined from outside India, so leave it unconfirmed. Hebbia is a different case and shouldn’t be filed with Cursor. hebbia.com/pricing returns “Failed to load page data from Contentful.” and no price at all. That page is broken, not minimal, and you can’t even tell whether a price was ever published there. Now look at the field. Harvey, Sierra, Glean, Hebbia, Abridge and OpenEvidence publish no usable price between them. Harvey’s and Sierra’s pricing addresses are 404s. Cognition is the only vendor here with a full published list. And neither cursor.com/pricing nor cursor.com/enterprise states a seat minimum for Teams or Enterprise, so that’s a question too. Three habits follow. Don’t diagnose a vendor’s page from what you expected to see on it. Tell a broken page apart from a deliberate one. And get any tier you intend to buy confirmed in a written quote, because on this map a written quote is usually the only price that exists.