AI Enablement Assessment

See how your team actually uses AI.

Most engineering leaders suspect their team isn't getting full value out of AI tools. Our assessment can measure exactly that. Our quiet telemetry sprint powered by Pheebs, the open-source measurement tool we created, resulting in an executive debrief and a concrete plan for full AI proficiency.

Built on Pheebs, our open-source toolClaude Code, Cursor, and Codex30-day telemetry sprintAI harness feature usageAI quality behaviorsModel spend, priced in dollars

Schedule a call

Why measure

You're paying for AI.
Is it paying you back?

Between seats, tokens, and engineering hours, AI is already a real line item. Without measurement you can't see which part of that spend compounds and which part leaks. The assessment turns suspicion into a baseline, including an opportunity-savings estimate in dollars.

Idle capability

Licenses and features your team barely touches. Paid for, never adopted, and invisible until someone measures adoption.

Model misuse

Frontier-priced models doing routine work, and small models failing tasks that needed a bigger one. Both burn money. Pheebs sizes the work in every session against the model that ran, and the gap becomes a dollar number.

Unverified output

AI code accepted without tests or pushback turns into rework. The cost lands weeks later, in review time and bug hunts.

Structural gaps

AI harnesses offer tooling for teams to make the best out of them. What are the gaps in your team and codebase?

30 days of passive telemetry

3 coding agents covered: Claude Code, Cursor, and Codex

24 AI practices measured per engineer

$ opportunity savings in model usage, sized in dollars

How the service runs

A quiet sprint, light on your team.

01

Instrument

We set up Pheebs, the measurement tool we built, on the AI tools your engineers already use. One CLI, minutes per machine. No source code and no file paths leave the building, and prompt text is never stored.
02

Measure

Engineers work normally for about 30 days while passive telemetry captures how AI actually gets used. Nothing changes about how anyone works.
03

Debrief

We walk your leadership through the baseline: adoption, quality habits, and spend, with a written report your leads keep.
04

Close the gaps

Every finding maps to a fix: scaffolding to ship, coaching to run, habits to spread. The next window shows whether it moved.

The report

What the debrief covers.

Not a score. A baseline your leads can act on, view by view.

Which AI capabilities your engineers actually use: models, reusable skills, MCP connections, verification, context habits, and multi-agent workflows
Skill and command adoption: what gets reused, what got tried once and dropped
Tool intent and token spend, mapped to teams and repos
Model fit: the work each session needed against the model that ran, with the savings opportunity priced in dollars
Quality habits: how often AI output gets verified, questioned, and refined before it ships
AI changes that shipped with no verification and no pushback, as a share of lines and by repo
Per-engineer profiles and a team heatmap that separate structural gaps from coaching gaps
A prioritized enablement plan: what to ship, who to coach, and who should demo

Sample views from the report:

Practice adoption funnel Team · trailing 4 weeks Illustrative data

Models 67% recurring

Artifacts 50% recurring

MCP 50% recurring

Context management 33% recurring

Orchestration 33% recurring

Evals 0% recurring

One bar per competency, split by how many engineers haven't acted on it, acted on it once, or acted on it week after week. Unobserved and Adopted are the competencies to be intentional about.

Practice heatmap Team · trailing 4 weeks Illustrative data

Every engineer against every competency. A cold column, like Evals here, means the team is missing the setup and practice for it: that fix is structural. A cold row, like Kevin's, calls for more strongly for coaching.

Engineer view Pam · Trailing 4 weeks Illustrative data

Models Recurring 67%

Artifacts Recurring 33%

MCP Recurring 100%

Evals Adopted 20%

Context management Recurring 50%

Orchestration Recurring 40%

Coverage index 57%

The per-engineer view. Each row shows how much of a competency has become habit, and the coverage index sums it up in one number you can track over time.

Signals by engineer Team · trailing 4 weeks Illustrative data

Engineer	Verification	Pushback	Refine : repair	Wholesale	Model-fit
Michael	39%	9%	1.1 : 1	41%	58%
Dwight	55%	17%	1.6 : 1	26%	63%
Jim	44%	12%	1.2 : 1	38%	51%
Pam	71%	22%	2.4 : 1	15%	35%
Angela	62%	15%	1.9 : 1	21%	66%
Kevin	26%	4%	0.7 : 1	55%	41%
Team median	50%	14%	1.4 : 1	33%	55%

The five judgement signals, per engineer. One weak number is a coaching conversation. A weak column across the whole team is a structural gap. Model-fit is the one column with a price on it.

Skill reuse & connectors Team · trailing 4 weeks Illustrative data

Skill reuse, share of engineers invoking each

/review 78%

/release-notes 56%

pr-describe 44%

db-migrate 22%

Connectors in use, share of engineers using each

Linear r/w 67%

GitHub r/w 56%

Postgres r 33%

Sentry r 22%

Which shared skills and connectors the team actually reaches for. Ship a new one, then watch this view in the next window to see whether it caught on.

Dollar savings Team · trailing 30 days Illustrative data

Savings opportunity $9,960 of $32,400 list-price spend API list-price equivalent, estimated upper bound

Models used 67 · 19 · 12 · 2%

Work, as sized 37 · 25 · 33 · 5%

Coverage complete 1,912incomplete 131no telemetry 58unpriced 14 Decisions and figures come from complete sessions only.

The gap between the two bars is the opportunity: half the frontier sessions were sized for a large model or below.

After the assessment

We don't hand you a report and leave.

Have us move the needle with you and we can help fill the gaps the assessment surfaces.

Structural fixes

Shared gaps get scaffolding: eval harnesses, MCP configuration, context files, and kit skills shipped into your repos.

The enablement program

Individual gaps get coaching: group sessions, a structured curriculum, and pairing, run the way we run our own AI program.

Embedded engineers

The fastest lever. Eversynced engineers join your team and raise the bar from inside. Our AI practice comes with them.

Explore Staff Augmentation

Frequently asked questions

What leaders ask before we start.

Is this surveillance of my engineers?

No. Pheebs reads interaction patterns, not content. No source code, no file paths, and prompt text is never stored. The debrief is about habits and gaps, and anything flagged per engineer is handled by your leads the way any code review is.

What does setup require from my team?

One CLI install per machine, a few minutes each. Pheebs registers hooks in the agent's own config and runs in the background. No code access, no IDE changes, and engineers work exactly as they did before.

Why does it take 30 days?

Habits don't show up in a week. The model reads recurring behavior over a trailing window, so a practice counts when it shows up week after week, not when it fires once. Around 30 days is the shortest window that separates real habits from one-off tries.

Where does the dollar savings number come from?

From model fit. Pheebs sizes the work in every complete session, from a one-file change to open-ended design, and compares it with the model class that ran. Sessions that ran above what the work needed are priced at API list rates, and that gap is the number: an estimated upper bound, from your team's real sessions. Nothing gets routed or switched. The report prices the gap, and your team decides.

Can we keep Pheebs after the assessment?

Yes. Pheebs is open source, so you can keep the telemetry running and build your own reporting on top of it. Teams hire us for our reporting and enablement work.

Know where your team stands.

One call to scope it. Thirty days of quiet telemetry. One debrief your leadership will actually use.