On 15 July 2026, Intapp and Litera each announced that a single AI agent would now run the firm. Intapp made Celeste generally available, calling it an expert AI coworker for the business of the firm. Litera relaunched the whole company around “one platform, one agent, one dataset,” with its Lito agent spanning the practice of law. Same day, same claim, opposite halves of the building.
That collision is the reason this pairing lands on shortlists that would have looked absurd two years ago. Intapp has never drafted a document. Litera has never cleared a conflict. What changed is that both now sell the firm a single agent and a single dataset, and a firm can only anchor its agent in one place first.
The routing question
Ask where the agent needs to sit to be useful on day one.
Intapp anchors at the point where the firm decides whether to take the work at all. Celeste ships with playbooks for origination, business development, intake, conflicts, pricing, fundraising and lateral hiring, running on Intapp’s own systems of record — DealCloud, Conflicts, Intake and Walls — with a no-code builder for the workflows Intapp did not write. Ethical walls and need-to-know restrictions apply to the agent exactly as they apply to a human, and every action lands in an audit trail. BakerHostetler is the named early adopter on intake and conflicts; the private capital firm Hg is the named adopter on deal screening.
Litera anchors at the point where the document is written. Lito now spans five connected capabilities — Draft, Review, Negotiate, Protect and Close — inside Microsoft Word, Outlook, Google Workspace, browsers, iOS, iManage and NetDocuments. Litera then pushes toward the business side from there: Foundation 365, its institutional-knowledge and CRM layer built on Microsoft Dynamics 365, is deployed across five of the ten largest firms in the world.
So the two products converge on exactly one seam — firm CRM, experience data and business development — and nowhere else. Everything either side does outside that seam, the other does not do at all.
Where Intapp wins
Governance is the product, not a setting. Intapp Walls policies defined once enforce across Harvey, Microsoft Copilot and Celeste, with cloud Walls enforcement in Harvey reaching GA in July 2026. That means the firm governs AI it did not build, which is the position Litera cannot occupy because Litera is the AI.
The financials are public and they are checkable. Intapp closed fiscal 2026 on 30 June with $590.5 million of total ARR, $495.7 million of it cloud, up 29% year over year and now 84% of the total. Cloud net revenue retention was 123%. Litera is private, held by Hg since 2019, and publishes none of this.
Conflicts and intake are one system with one record. Intapp Conflicts runs the search and the disposition; Intake carries the risk questionnaire, the approval chain and the AML and KYC steps. That single record is the artifact a client’s outside counsel audit asks for, and reconstructing it from two systems after the fact is a project.
Private capital is a real second market. DealCloud serves more than 1,700 private capital and investment firms. If the buyer is a fund rather than a firm, Litera has nothing to sell them.
Where Litera wins
The agent is already inside the application lawyers use. Lito runs in Word, not in a portal a lawyer has to remember exists. Litera reports 70% of active Litera One accounts using Lito, against a base of more than 15,000 customers, over a million daily users, and 10 million document comparisons a month.
Generative output is optional. Litera’s own product page states that when work is high-stakes the deterministic engines run and return the same result every time, and that generative AI “can be switched off entirely” with the rules-based engines still working. For a firm whose risk committee has not yet cleared LLM output on client documents, that is a deployable configuration rather than a promise.
Coverage of the drafting chain is already built. Precedent search in Litera Create shipped ahead of the 18 August 2026 announcement, Litera for Windows reaches GA on 24 August 2026, the Lito Manage Changes workflow in Word is dated late September, and precedent-grounded rewrites plus a Due Diligence Agent joining Lito to Kira are dated Q4 2026.
Penetration is close to total. 99% of the Am Law 100 and 99% logo retention. Intapp counts 96 of the Am Law 100. In practice the firm already owns Litera, which changes the buying motion from a new vendor to a line item.
Pricing: the agent is the upsell vs the agent rides along
Neither vendor publishes a price, so the useful comparison is the shape of the bill rather than the number.
Intapp’s management described Celeste’s model on the fiscal Q4 2026 earnings call as a platform fee plus a usage fee, sold as an addition to existing lines — DealCloud with Celeste, compliance with Celeste, Time with Celeste. AI bookings were over 20% of net new bookings in that quarter, roughly double the prior quarter, and management called monetization early. Read that plainly: Celeste is the thing you buy next, and its cost moves with how much the agent works.
Litera went the other way first. On 27 October 2025 it made Lito available at no additional charge inside Draft Base, Draft Pro, Draft Advanced and Kira, for new and existing customers alike. The July 2026 relaunch published no new packaging, so as of today that inclusion is the last stated position. Litera’s cost therefore moves with seats and module count, not with agent activity.
For absolute magnitude, the only figures available are third-party transaction data, and they are thin enough that you should treat them as bands rather than prices:
Intapp
Litera
Published price
None
None
Stated model
Platform fee plus usage fee (Celeste)
Lito included in named products since Oct 2025
Third-party contract data
DealCloud: about $85K low, $505K average, $1.43M high per year — from fewer than ten tracked deals
Enterprise average $58,847/yr across 47 tracked customers; SMB average $2,010/yr across 7
Disclosed client mix
897 clients above $100K ARR; 142 above $1M
Not disclosed
The DealCloud sample is small enough to mislead on its own, but Intapp’s own disclosure corroborates the altitude: 897 clients paying more than $100,000 a year, out of $590.5 million of ARR. Litera’s tracked enterprise contracts cluster an order of magnitude below that, which reflects a genuinely different sale — Litera is bought module by module by firms that already own three of them, Intapp is bought as a platform.
Watch-outs
Intapp’s agent has nothing to ground itself in if the data is not there. Celeste runs on the firm’s systems of record. A firm without DealCloud, Conflicts or Intake in production is buying a data migration first and an agent second — budget the sequence, and do not let a Celeste business case carry the platform implementation inside it.
Litera’s roadmap is mid-cycle this week. ILTACON runs 23–27 August 2026 in Nashville and Litera has staged its next Lito version for it, with Litera for Windows GA on 24 August and two further milestones dated late September and Q4. Signing before the show means signing against a demo, not a shipped product. Ask for the GA date in the contract for anything you are buying on.
“One dataset” is a claim about the vendor’s products, not about the firm’s. Both agents are one agent over the vendor’s own estate. A firm running Litera for documents and Intapp for intake still has two datasets and two agents, and neither vendor’s messaging changes that. Decide which half you want consolidated and accept that the other half stays separate.
The seam is the only place a bake-off makes sense. Foundation 365 against DealCloud for firm CRM, experience and business development is a real head-to-head. Running a full evaluation across the whole footprint wastes a quarter comparing products that do not overlap.
Verdict
Pick Intapp when the constraint is risk and admission of work — conflicts volume, intake turnaround, ethical walls, outside counsel guideline terms — or when the buyer is a private capital firm rather than a law firm. Pick it also when the firm is deploying multiple AI vendors and needs one place to define who may see what, because Walls enforcing inside Harvey and Copilot is a control point no document vendor offers.
Pick Litera when the constraint is the document and the lawyers’ own hours: drafting, comparison, redline triage, due diligence, closing books. Pick it also when the risk committee has not cleared generative output, since the deterministic engines run with generative AI turned off. And pick it when budget is the real gate, because the agent is already included in products the firm most likely licenses today.
If the firm needs both, sequence rather than choose. Anchor on whichever half currently loses the most hours, and run the other on its existing footprint for a year. The seam to arbitrate is firm CRM and experience data — Foundation 365 against DealCloud — and it is the only bake-off worth staffing.
Pick neither when the work in question is the brief. Neither product researches law or drafts an argument; that is Harvey or Legora — see Harvey vs Legora. If the job is contract lifecycle for an in-house team rather than a firm, it is Ironclad — see best CLM platforms. And if the firm cannot say how many matters it cleared last quarter or how many pitches it ran, neither agent has a dataset to be grounded in, and the first purchase is matter management discipline, not AI.
On 15 July 2026, Intapp and Litera each announced that a single AI agent would now run the firm. Intapp made Celeste generally available, calling it an expert AI coworker for the business of the firm. Litera relaunched the whole company around “one platform, one agent, one dataset,” with its Lito agent spanning the practice of law. Same day, same claim, opposite halves of the building.
That collision is the reason this pairing lands on shortlists that would have looked absurd two years ago. Intapp has never drafted a document. Litera has never cleared a conflict. What changed is that both now sell the firm a single agent and a single dataset, and a firm can only anchor its agent in one place first.
The routing question
Ask where the agent needs to sit to be useful on day one.
Intapp anchors at the point where the firm decides whether to take the work at all. Celeste ships with playbooks for origination, business development, intake, conflicts, pricing, fundraising and lateral hiring, running on Intapp’s own systems of record — DealCloud, Conflicts, Intake and Walls — with a no-code builder for the workflows Intapp did not write. Ethical walls and need-to-know restrictions apply to the agent exactly as they apply to a human, and every action lands in an audit trail. BakerHostetler is the named early adopter on intake and conflicts; the private capital firm Hg is the named adopter on deal screening.
Litera anchors at the point where the document is written. Lito now spans five connected capabilities — Draft, Review, Negotiate, Protect and Close — inside Microsoft Word, Outlook, Google Workspace, browsers, iOS, iManage and NetDocuments. Litera then pushes toward the business side from there: Foundation 365, its institutional-knowledge and CRM layer built on Microsoft Dynamics 365, is deployed across five of the ten largest firms in the world.
So the two products converge on exactly one seam — firm CRM, experience data and business development — and nowhere else. Everything either side does outside that seam, the other does not do at all.
Where Intapp wins
Where Litera wins
Pricing: the agent is the upsell vs the agent rides along
Neither vendor publishes a price, so the useful comparison is the shape of the bill rather than the number.
Intapp’s management described Celeste’s model on the fiscal Q4 2026 earnings call as a platform fee plus a usage fee, sold as an addition to existing lines — DealCloud with Celeste, compliance with Celeste, Time with Celeste. AI bookings were over 20% of net new bookings in that quarter, roughly double the prior quarter, and management called monetization early. Read that plainly: Celeste is the thing you buy next, and its cost moves with how much the agent works.
Litera went the other way first. On 27 October 2025 it made Lito available at no additional charge inside Draft Base, Draft Pro, Draft Advanced and Kira, for new and existing customers alike. The July 2026 relaunch published no new packaging, so as of today that inclusion is the last stated position. Litera’s cost therefore moves with seats and module count, not with agent activity.
For absolute magnitude, the only figures available are third-party transaction data, and they are thin enough that you should treat them as bands rather than prices:
The DealCloud sample is small enough to mislead on its own, but Intapp’s own disclosure corroborates the altitude: 897 clients paying more than $100,000 a year, out of $590.5 million of ARR. Litera’s tracked enterprise contracts cluster an order of magnitude below that, which reflects a genuinely different sale — Litera is bought module by module by firms that already own three of them, Intapp is bought as a platform.
Watch-outs
Verdict