Every vendor gives you a dashboard. None of them answers the question.
Somebody is going to ask how the company's AI spend is actually being used. Per-tool dashboards can't answer it — and they were never going to.
Four tools, four truths
Copilot, Cursor, OpenAI, and Claude each report different numbers at different granularities. Adding them up in a spreadsheet is a quarterly ritual that's stale by Tuesday.
Seat ≠ person
Teams share logins and API keys. Vendor dashboards assume one seat is one human, so shared accounts silently undercount your real adoption.
Graded homework
Every vendor measuring its own product's impact has the same credibility problem. No vendor will ever tell you a competitor's tool is used better.
Connect. Backfill. Score.
Self-serve from the first click: no call, no pilot, no CSV wrangling.
- 01
Install the agent
Run the Revealyst Agent on your machine. It reads your local Claude Code sessions and pushes only aggregates with a device token — never prompt content. Nothing runs in the background.
- 02
Summarize
Your usage is normalized on-device onto one metrics model, with an attribution-confidence tag on every record telling you exactly what the data supports.
- 03
Score
Adoption, Fluency, and Efficiency compute from versioned definitions you can inspect — every number tied to the exact formula that produced it, so a 78 is something you can trace.
Three numbers that answer the board's question.
These scores measure adoption and usage sophistication — a leading indicator of where AI value can come from, not a measure of realized productivity. Every one is computed from a versioned definition you can inspect: you can see which formula produced which number, and history recomputes when definitions improve.
The question every CTO is being asked: are we still stuck in pilots, or actually getting leverage?
Your connected tools already hold the usage half of the answer. The AI Maturity model reads them into one board-level level — Dormant through Amplified — across three measured axes. The levels are modeled and directional, not a certified grade; naming what they can't show is the point.
Placed on three measured axes
How widely AI is used — the share of known people who are active, plus how many distinct features are in play.
How sophisticated the use is — agentic work, multiple features in a day, and parallel agent runs.
How steady the habit is — whether people show up week after week rather than in bursts.
What we don't measure — and won't guess
A maturity level is a leading indicator of usage sophistication, not a business-outcome number. Where telemetry can't honestly support a figure — or shouldn't — we name the gap instead of inventing one.
- Shadow AI
- Usage on personal accounts and unconnected tools that never reaches us.
- ROI and time saved
- A dollar return or hours-saved figure attributed to AI use.
- Per-person quality or ranking
- A scoreboard rating individuals by code quality or output.
- Governance & training maturity
- A rung for policy, guardrails, and enablement maturity.
Numbers you can defend, because we refuse to invent them.
Every metric carries an attribution-confidence tag: the granularity the data honestly supports. When it only supports key-level truth, that's what you see. Revealyst never fabricates per-user numbers — a gap is shown as a gap, not a guess.
“You think 12 people use AI. The pattern says it's more.”
Shared-account detection flags round-the-clock seats and outlier volume — so you learn adoption is undercounted, sharing is violating vendor ToS, and shared credentials are an exposure.
Built to pass the works-council test.
Scoring people is near the EU AI Act line even without reading content — so privacy is architecture here, not a settings page.
Pseudonymized, team-level by default
Individual identities appear only if an org admin explicitly changes the visibility mode — never silently. Individual self-view is the free Personal mode, where you are your own data subject.
No prompt content. Ever.
Scores use only behavioral signals the vendor APIs already expose — acceptance rates, engaged days, feature breadth. Nothing your people type is read.
No extension, no proxy
Ingestion is an on-device agent you run yourself — it reads your local AI-tool logs and pushes only aggregates. We rejected browser extensions outright — that's monitoring, and it's not the product.
Compliance guidance included
DPIA template, works-council notification note, and AI Act checklist ship inside the product. Built for EU buyers, not retrofitted.
Start with your own score.
Revealyst is free forever for individuals. Connect your own Claude Code usage, get your own Adoption and Fluency scores, and — if you choose — share the card.
Example card — scores are measured from real AI-tool usage, not self-reported.
The score card is opt-in and shows exactly one thing: the label you chose and your featured score. No email, no employer, no history. Revoke the link any time.
Curious how you compare? Opt into anonymized benchmarks to help build the published comparison set. Next to your scores you'll see a clearly-labeled modeled estimate in the meantime — verified figures replace it only once confirmed against a primary source.
Test your AI fluencyThe cheapest answer in the category, on purpose.
Value scales with headcount, so pricing is per tracked user — an identity-resolved person with real usage in the period. Unresolved keys and shared accounts are surfaced, never billed.
Billing is handled by Paddle as merchant of record — sales tax and VAT are collected and remitted for you, worldwide.
The board is going to ask. Answer with numbers.
Who's using AI, how well, and whether usage is keeping pace with the spend — measured neutrally across every tool you run, in minutes.
Get your first score — free