Private for each lawyer
Individual reports are coaching instruments. They are not legal-quality determinations, public rankings, or employment-performance evaluations.
For firms already using Harvey
You bought Harvey. Now prove it's working.
Every lawyer gets an AI coach. The firm gets proof.
We turn Harvey usage data into a clear picture of adoption, private coaching for each lawyer, and evidence leadership can use for renewals, ROI decisions, and the next phase of the firm's AI program.
Developed through real-world analysis inside large law firms, covering thousands of lawyer prompts.
Harvey coaching report · JulySample · fictional lawyer
M. Calloway
Partner, Corporate · Monthly report, with history from February
78Prompts
18Active days
61Legal-work turns
Monthly activity
You, next to the firm
| Measure | You | Partners |
|---|---|---|
| Active days in July | 18 | 9 average |
| Prompts citing a source | 58% | 71% median |
| Threads with a follow-up | 31% | 19% median |
Prompt recommendation
Your prompt"summarize the key risks in this agreement"
Recommended rewrite"We represent the buyer. Summarize the top risks in this agreement for the client's GC, as a short bullet list with clause references."
The rewrite adds context, audience, and output format, so the first response is usable instead of generic.
Task+ Context+ Audience+ OutputDoc attached
ROI evidence for leadership
Leadership sees adoption, capability patterns, and movement by role and practice group, the evidence needed for seat allocation and renewal decisions.
Secure transfer after scoping
We agree the minimum data, access rules, retention, and approved transfer method in writing before anything moves. Read the process.
I.
Replace the anecdotes with evidence.
Your firm made the Harvey investment, and Copilot, ChatGPT, and Claude are probably in the building too. The invoices arrive every month. The evidence usually does not.
- AdoptionA minority of lawyers become regular users while paid seats sit untouched.
- Prompt qualityMissing context, audience, and format make the first response disappoint, so the lawyer quietly goes back to the old way.
- TrainingA lunch-and-learn costs billable hours and fades quickly because it is disconnected from each lawyer's actual work.
Who actually uses the tools you pay for?
Are they using them well?
What changed since last month?
Firm Harvey ScorecardSample · illustrative firm
154 licensed seats, one month
Weekly users (18)Tried it (44)Never opened it (92)
| Question the Scorecard answers | Example finding |
|---|---|
| Seats never used this month | 92 of 154 |
| Practice group furthest ahead | Corporate |
| Share of prompts missing key context | 63% |
| Month-over-month engaged users | +12% |
Tools covered
Harvey
Microsoft Copilot
ChatGPT
Codex
Claude
II.
The AI-enabled law firm is the destination
Buying Harvey is one decision inside a much larger operating model. Firms that turn AI into client-visible value align six layers: data, general-purpose AI, legal platforms, practice tools, custom applications, and the people and policies that make all of it usable.
- 01An NDA comes back in minutes, with exceptions ready for lawyer judgment.
- 02A brief meets the other side's strongest argument before it is filed.
- 03Firm knowledge becomes reusable across matters without bypassing security.
III.
Two ways to start, based on the data your firm is ready to use.
Complimentary · metadata only
Adoption Snapshot
A first look at whether the licenses are being used. No prompt text is required.
- Active users and usage frequency
- Feature usage, cohorts, and practice-group patterns
- Movement across the available reporting period
Separately scoped · prompt access
Prompt Coaching Assessment
A deeper analysis that produces coaching-quality findings from lawyers' actual usage.
- Private, identifiable coaching reports for participating lawyers
- Prompt patterns and capability gaps for firm leadership
- Written rules for access, recipients, retention, and deletion
The Adoption Snapshot and Prompt Coaching Assessment are different scopes. We do not request prompt text for a metadata-only snapshot. See how the data boundary is documented.
IV.
How prompt coaching works
a.
Agree the data boundary
- Document the fields, visibility rules, retention period, and secure transfer method.
- Use a Harvey export or read-only API access. Nothing is installed in your document management system.
b.
Analyze every substantive prompt
- Score context, task, audience, instructions, output format, and verifiability.
- Compare firm patterns month over month, by role and practice group.
c.
Coach every lawyer
- Give each lawyer private examples, rewrites, and one experiment for the next 30 days.
- Give leadership the firm Scorecard with adoption and capability patterns.
Private coaching, built from each lawyer's actual work.
V.
Inside a coaching report
Five pages, personal to each lawyer, generated from their own work. Here is what the four core sections do. All examples below are fictional.
Section oneYou, next to the firm
- Compare activity with the firm and peers in the same role.
- Keep the report private. It is never a leaderboard or employment-performance evaluation.
Active days: 18 (partners' average: 9)
Prompts citing a source: 58% (median: 71%)
Prompts citing a source: 58% (median: 71%)
Section twoYour prompting patterns
- Review every substantive prompt on six elements.
- Show the lawyer's strongest and leanest prompts side by side.
Context ✓ · Task ✓ · Audience ○ · Instructions ○ · Output ✓ · Verifiability ○
Section threePrompt recommendations
- Rewrite real prompts with the smallest useful change, before and after.
- Teach one pattern the lawyer can reuse on next week's work.
Their promptsummarize the key risks in this agreement
Recommended rewriteI represent the buyer in this asset purchase. Review the agreement and list each material risk with a severity rating and the specific change we should request. Format as a table I can forward to the client.
Why it helps: role, stakes, and format give the tool clearer instructions for producing a stronger draft for lawyer review.
Section fourOne experiment for the next 30 days
- Choose one habit and one unused capability to test.
- Show next month whether the behavior changed.
This month: name the intended reader on drafting tasks.
Untried capability: batch document review across a deal set.
Untried capability: batch document review across a deal set.
VI.
A clear path from evidence to improvement
Step 1 · See the work
Read the sample
Five-page PDF, fictional lawyer, real format
- Review the full five-page coaching report
- See the analysis, prompt recommendations, and benchmarks before sharing any data
- Circulate it to your innovation committee
Step 2 · Measure adoption
Get the Adoption Snapshot
Complimentary · metadata only
- See adoption by role and practice group
- Find unused seats and feature-utilization gaps
- Take evidence, not anecdotes, into your renewal
Start with firm details only. Do not attach an export.
Request your snapshotStep 3 · Improve for 90 days
Coach every lawyer
Monthly program · scope and price confirmed in writing
- A personal coaching report for every lawyer, every month
- The firm Scorecard for leadership, with trend lines
- Built around Harvey; Copilot, ChatGPT, Codex, and Claude covered where you run them
VII.
Who's behind it
Automated Legal is run by Robbie Allen. He has spent more than 25 years on both sides of enterprise technology: writing the code and making the executive decisions around it.
That experience shapes the work here. Robbie understands the engineering underneath the tools and the evidence leaders need before renewing a platform or funding the next build.
- From code to CEOThirteen years at Cisco, ending as its youngest Distinguished Engineer, followed by founding and leading Automated Insights.
- AI before the boomStarted StatSheet in 2007. It became one of the first commercial generative AI companies and produced 10 million personalized Yahoo Fantasy Football recaps a week.
- Three acquisitionsAutomated Insights, Infinia ML, and Bionic Health all went through acquisitions.
- Inside law firmsDeveloped this coaching method through months of Harvey usage analysis, thousands of lawyer prompts, and reports delivered directly to lawyers.
VIII.
Questions law firms ask first
Do you see client-confidential information?
The complimentary Adoption Snapshot uses metadata and does not require prompt text. Prompt Coaching Assessments require separately scoped access because prompts can contain matter references or confidential information.
Before any transfer, the engagement documents identify the fields, authorized recipients, storage and processing services, access controls, retention period, deletion process, and any approved zero-data-retention configuration. Do not send exports by email. Read the data-handling process.
Which tools do you cover?
Harvey first. That's where the analysis is deepest: prompt-level scoring, feature utilization, and per-lawyer coaching from the actual usage data. If your firm also runs Copilot, ChatGPT Enterprise, Codex, or Claude, we fold them into the firm Scorecard where the vendor's export supports it, and we tell you exactly what your stack supports before you sign anything.
Is this training?
No live training sessions. Each participating lawyer receives personalized, actionable advice based on their own work and can read it on their own time.
Are you affiliated with Harvey or the other vendors?
No. We are independent and vendor-neutral. We analyze the tools; we don't sell them. If your usage data says a tool isn't earning its seat cost, the report will say so.
What does it cost?
The sample report and metadata-only Adoption Snapshot are complimentary. Prompt Coaching Assessments and ongoing coaching are separately scoped based on the lawyer cohort, tools, and data required. We confirm the scope and price in writing before any usage data moves.
Replace the anecdotes with evidence.
Tell us where the firm is today. The form prepares a request email without collecting or uploading confidential information.