WOLF AI field guide
Pricing and citation accuracy in FCA legal AI.
Two questions decide most FCA legal AI evaluations: what it costs, and whether its outputs can be checked. Cost is driven by matter and document volume rather than seat count alone. Citation accuracy is whether a generated statement points to a real passage that actually says what it is claimed to say.
Published · Last reviewed
Written and maintained by the WOLF AI product team and checked against the public sources cited on this page. It has not been reviewed by outside counsel, and it is not legal advice.
What do the numbers say?
Comment 8 to ABA Model Rule 1.1 states that maintaining competence requires a lawyer to keep abreast of "the benefits and risks associated with relevant technology." Adopting an AI-assisted workflow and verifying what it produces are treated as two halves of the same professional obligation, not as separate choices.
Source: ABA Model Rule 1.1: Competence (including technology)
The NIST AI Risk Management Framework organizes trustworthy AI around functions it calls Govern, Map, Measure, and Manage, and treats traceability and human oversight as design requirements rather than optional additions. Source-linked output is the practical form that traceability takes in a litigation workflow.
Federal Rule of Civil Procedure 9(b) requires that a party "must state with particularity the circumstances constituting fraud or mistake." In practice that is what turns an FCA matter into a document-organization problem: the allegation has to name specific claims, dates, and actors, each traceable to a record.
Source: Federal Rule of Civil Procedure 9(b): pleading fraud with particularity
Who is this workflow for?
Anyone building the evaluation criteria for an FCA legal AI pilot, who needs to ask about cost structure and verifiability before a demo rather than after one.
When it is the wrong tool. This page gives no numeric price for any product, including WOLF AI, and it publishes no accuracy benchmark. Neither exists here in a form that would survive scrutiny, and inventing one would be the exact failure the page is about.
What documents and inputs do you need?
- Your expected matter volume, and how many are active at once.
- Your document volume and formats, which usually drives cost more than headcount does.
- Your team size and access requirements.
- Onboarding, workflow, and integration scope, including whether anything has to connect to a system of record.
How does the workflow actually run?
- 1.Ask what the price is a function of. Seats, matters, documents, and storage produce very different bills at the same headline rate.
- 2.Ask what happens to a citation when it is wrong: whether the product shows you the passage, or only asserts the fact.
- 3.Test verification cost on your own material — how many seconds it takes to check one generated statement against its source, because that number multiplied by your review volume is the real cost.
- 4.Ask what the vendor has published rather than what it claims: a measured study, a certification, a reference customer, or none of those stated plainly.
What does this look like in practice?
Illustrative scenario
Illustrative only — a constructed scenario. Two products are evaluated on the same 200-page sample from a live matter.
- Both produce chronology entries that read well. That is where most evaluations stop, and it is the wrong place to stop.
- One entry from each is checked against the source. The first product shows the page and passage, and the reviewer confirms it in about ten seconds.
- The second asserts the same fact with no location. Confirming it takes four minutes of searching, and the fact turns out to be right.
- Both were accurate. Only one of them is affordable to check at the volume the matter actually requires.
What this does not show. A citation can point at a real passage and the inference drawn from it can still be wrong. Provenance makes checking cheap; it does not do the checking.
What do you get out, and who reviews it?
- An evaluation frame that prices verification effort rather than only licence cost.
- A written list of what each vendor has actually published, and what it has not.
- A pilot scope narrow enough to test on real material.
- WOLF AI’s own answer, stated the same way everywhere on this site: WOLF AI uses custom pricing for plaintiff-side FCA and qui tam firms. Pricing depends on matter volume, team size, document volume, and onboarding requirements. Early design-partner terms are discussed privately during a demo.
Where does this approach break down?
- No public self-service plan, public numeric price, or general-availability offer exists for WOLF AI.
- No accuracy benchmark or verified time-savings study has been published, and efficiency statements on this site are stated hypotheses rather than measured results.
- Source linking supports verification. It does not guarantee correctness, and it does not move professional responsibility from the reviewer to the software.
- There is no public production API or MCP server, so integration cost cannot be estimated from a published interface.
Common questions
How much does WOLF AI cost?
WOLF AI uses custom pricing for plaintiff-side FCA and qui tam firms. Pricing depends on matter volume, team size, document volume, and onboarding requirements. Early design-partner terms are discussed privately during a demo.
What does citation accuracy actually mean?
Two separate things, and they are worth separating. First, whether the cited location exists and says what the output claims. Second, whether the inference drawn from it holds. Source linking makes the first cheap to check and leaves the second entirely with the reviewer.
Has WOLF AI published an accuracy benchmark?
No. No benchmark, no verified case study, and no measured time-savings result has been published. The honest answer today is to test the product on your own material during a pilot rather than to trust a number.