The Real Question: What Problem Are You Actually Buying?

Before you can answer whether CoCounsel is worth the money, you have to answer a harder question: what problem are you actually trying to solve? Because the honest answer from the research is that CoCounsel solves one problem very well — and it's probably not yours.

CoCounsel's documented strengths are case-law research, deposition prep, and large-scale document review. Its deepest features map to Westlaw Brief Builder, Tabular Analysis for up to 10,000 documents, and Deep Research — all document- and research-centric workflows. Even Thomson Reuters' own criminal-law page describes discovery document review, timeline building, and motion drafting, with no mention of bodycam video, interrogation footage, jail calls, or 911 audio. That's not a knock on the product; it's just not built for what lands in a defense discovery dump.

And what lands in that dump is enormous. Public defenders receive hundreds of hours of digital discovery each year — bodycam, custodial interrogations, jail calls — and Kansas City police alone generate 350+ hours of footage in a single day. Reviewing that material is the bottleneck, and it's a media problem, not a research problem. No source describing CoCounsel — including the vendor's own materials — documents timestamped video or audio review with citable transcripts. When you evaluate it against defense discovery, you're evaluating by absence.

The value question also resolves conditionally. Multiple independent assessments converge on the same verdict: CoCounsel is worth it when two things are true — you already run on Westlaw, and your work leans on case-law research and large-scale document review. If neither is true, you're paying for a research platform plus a sales cycle to get an AI layer you could buy more cheaply and transparently elsewhere. The pricing reinforces that mismatch: no published rates, no self-serve trial, third-party estimates running from roughly $225/user/month for CoCounsel Core to ~$492/user/month for the full stack, plus a Westlaw subscription — with some full configurations estimated near $3,000/seat.

One more thing to weigh before you sign anything. The foundation of any evidence-review tool is transcription, and independent research shows AI transcription misattributes speakers and drops dialogue — a Northeastern study of 176 bodycam videos found AI averaged 3.3 participants per transcript versus 4.8 in human-edited versions, omitting roughly 25,000 words across the sample. As Calli Schroeder of the Electronic Privacy Information Center put it, a 2% error "could completely change who is at fault and what occurred in an encounter." Any tool you trust with evidence — CoCounsel or otherwise — needs a verification workflow where AI output is marked for human review and every claim traces back to the exact source moment.

That's the standard we built Evidence Copilot against: a grounded evidence-review workspace where every AI answer cites the exact second of video or page of the PDF, and nothing is applied without attorney review. Whatever you buy, buy it for the problem you actually have. If your problem is 40 hours of bodycam and a jail call that contradicts the police report, research AI isn't the answer — grounded evidence review is.

What CoCounsel Costs and What the ROI Numbers Really Say

Ask Thomson Reuters what CoCounsel costs and you won't get a number — you'll get a sales call. The vendor publishes no rates on its product pages, and third-party reporting confirms there's no free trial or self-serve evaluation either. For a solo defender or a PD office weighing line items, that opacity is itself a data point.

What the estimates say is sobering. Pricing analyses put CoCounsel Core around $225/user/month, with full stacks estimated near $3,000/seat once Westlaw is bundled in. A Thomson Reuters configurator snapshot for a five-attorney firm showed CoCounsel Legal at roughly $492/user/month — and that's before the Westlaw subscription most of its value depends on.

Then there are the ROI numbers, and they deserve careful handling:

• 76% improved research quality and a 33% time reduction — from a Forrester study Thomson Reuters itself commissioned, per its own marketing page.
• "Up to 63%" faster discovery review and timelines "79% faster" — unverified vendor claims on the criminal-law page.
• The "$60,000 in recovered billable time" figure is the vendor's own extrapolation at a $250 billing rate.

None of these are independent benchmarks. As one assessment bluntly notes, any single number you see online about CoCounsel is an unverified third-party estimate.

The accuracy record matters too. A 2024 Stanford study cited by The Legal Prompts found Westlaw's AI hallucinated at a 33% rate — nearly double Lexis+ AI's 17%. That's the same AI family CoCounsel sits on top of.

For defenders, the honest math is this: CoCounsel's strengths are research and document review, not bodycam footage, jail calls, or interrogation video. If your bottleneck is the discovery dump itself, a research platform at ~$400+/user/month won't move it. Tools like Evidence Copilot publish their pricing up front — $39/user/month Solo, no sales wall — because defense buyers deserve to run that comparison without booking a demo.

The vendor's own framing tells you who this is for: firms already standardized on Westlaw. Everyone else is paying for a sales cycle on top of a research platform.

The Transcription Problem Every AI Evidence Tool Must Answer

Before you trust any AI tool with discovery, ask one question: what happens when the transcript gets a speaker wrong? A recent Northeastern University study of AI bodycam transcription shows how much rides on that answer — and why human-review workflow should be the first thing you evaluate, not the last.

Researchers analyzed 176 bodycam videos spanning roughly 23 hours across 73 incidents. The AI captured most words correctly, but it frequently misattributed statements to the wrong speakers. Its transcripts averaged 3.3 participants per incident versus 4.8 in human-edited versions, and the AI versions came up about 25,000 words short across the sample. In a multi-officer encounter, a transcript that drops participants is a transcript that drops testimony.

Calli Schroeder of the Electronic Privacy Information Center put the stakes plainly: an AI transcript "may be 98% accurate, but that 2% could completely change who is at fault and what occurred in an encounter." For defense work, that 2% is often the whole case — who said what during a custodial interrogation, whether consent was actually given on the bodycam, what the jail call actually contains.

So when you evaluate any AI evidence tool — CoCounsel, JusticeText, or a transcription-first product — make the verification workflow your core criterion:

• Can you see and correct the transcript, word by word, with speaker labels you can rename?
• Does every AI answer cite back to the exact timestamp or page, so you can verify against the original?
• Are AI outputs clearly marked as pending human review, with originals kept immutable?
• Is there an audit trail of what was reviewed, redacted, or changed?

The Northeastern researchers concluded that human review remains essential before AI transcripts are used in criminal cases. Shellie Solomon of Justice & Security Strategies framed it the way most experienced defenders already work: treat AI as "the investigative assistant who prepares the file," while humans decide what the details mean.

This is also where CoCounsel's fit gets shaky for defense work. Its documented strengths are case-law research, deposition prep, and large-scale document review — the vendor's own criminal-law page describes discovery document review and motion drafting, with no mention of bodycam, interrogation footage, or jail-call transcription. And the 2024 Stanford study cited in industry analysis found Westlaw's AI at a 33% hallucination rate, versus 17% for Lexis+ AI — worth knowing when the tool is drafting your research memos.

For defenders drowning in digital discovery — public defenders receive hundreds of hours of footage and call audio each year — the right tool is one built around verification from the ground up. Evidence Copilot, for example, marks every AI output for human review, cites answers to the exact second of video or page of PDF, and keeps originals immutable. Whatever you choose, hold every vendor to the Northeastern standard: the AI prepares the file, and a human signs off on it.

A Practical Evaluation Checklist for Defense Evidence-Review AI

Before you spend an hour on a vendor demo, run your candidate tool through the same gauntlet you'd use on a new investigator: does it do the actual work, can you verify it, and what does it really cost. For defense evidence review, five questions separate usable tools from expensive research platforms.

Does it transcribe your actual evidence? Bodycam, interrogation footage, jail calls, and 911 audio are the bulk of modern discovery — public defenders receive hundreds of hours of digital discovery each year, and Kansas City police alone generate 350+ hours of footage daily. Yet CoCounsel's documented features are research- and document-centric: Brief Builder, Tabular Analysis for up to 10,000 documents, deposition prep. No source describes it producing timestamped, citable transcripts of bodycam or interrogation video. If the tool can't diarize speakers with word-level timestamps, it isn't reviewing your evidence — it's summarizing your documents.

Does every answer cite the exact source moment? This is the verification question, and it matters more than any feature list. A Northeastern study of 176 bodycam videos found AI transcripts averaged 3.3 participants per transcript versus 4.8 in human-edited versions — frequent speaker misattribution — plus roughly 25,000 fewer words across the sample. As EPIC's Calli Schroeder put it, an AI summary may be 98% accurate, but that 2% "could completely change who is at fault." If an answer can't click through to the exact second of video or page of the report, you can't verify it, and an unverifiable answer is worthless in a suppression hearing.

Is AI output marked for human review? The Northeastern researchers concluded human review remains essential before AI transcripts are used in criminal cases. Any tool that presents AI conclusions as finished work fails this test. Evidence Copilot, for example, marks every AI answer as pending human review, files AI-suggested redactions and sync proposals as review cards rather than auto-applying them, and never makes legal conclusions — cross-exam drafts are suggested questions tied to quoted sources, saved as attorney-review notes.

Is pricing published? CoCounsel publishes no pricing; third-party estimates run from ~$225/user/month for CoCounsel Core to ~$492/user/month for the full product, plus a Westlaw subscription, with no free trial or self-serve evaluation. The Legal Prompts' broader finding: legal AI pricing is "deliberately opaque." A tool you can't price or trial before a sales call is a tool you're evaluating blind. Evidence Copilot publishes its pricing — Solo at $39/user/month, Firm at $99/user/month, PD Office at $750/attorney/year — with no feature tiers and hard caps on metered overage.

Can you evaluate it yourself? If the answer to any question above is no, you have your answer about the tool. If the evidence-review half of your workflow is the gap, a grounded workspace built for defense discovery — timestamped transcripts, citations to the exact video moment, and published pricing instead of a sales cycle — is the alternative worth a self-serve trial before you commit to anything.