For VP Eng who have to justify the AI bill

Prove your AI tools are worth it.
With a number your CFO can’t argue with.

Your AI seats renew whether you measure them or not. AgentLens measures whether Cursor, Copilot, Windsurf, Codex, and Claude Code are actually shipping value — then benchmarks your team against 100+ engineering orgs your size. Install a GitHub App, see your data in 10 minutes.

10-min setup
No source code stored
Engineers touch nothing

AgentLens

AI ROI · Q2 2026

Live · 2m ago

6-week AI adoption

+30 pts ↑

AI Tool ROI · Q2 2026

Live data
Time reclaimed+0 hrs/mo
Value of time$0/mo
AI tool spend$9,200/mo

Return on investment

↑ from 4.1× last quarter

0.0×
vs. 100+ orgs your size73rd percentile ↑
0th25th50th75th100th
Claude Code
68%
High
Cursor
41%
Medium
Copilot
22%
High
Windsurf
10%
Medium
Codex
6%
Medium

Commits analyzed this quarter

4,821 commits · 18 repos

GitHub-reviewed App(not OAuth — rate limits + enterprise trust)
95%+ detection accuracy · uncertainty always labeled
Your data is isolated — cryptographically separate per customer
14-day free trial · no credit card
The problem you’re already living with

You approved the spend. Now you have to defend it.

The only number you have on $200K of AI seats is Copilot’s acceptance rate — and you already know it’s meaningless. Accepting a suggestion isn’t the same as shipping value, and your CFO is going to ask anyway.

Cut the tool and you might be slowing the team down. Keep it and you might be lighting money on fire. Right now you’re guessing in front of the board.

Every vendor dashboard measures their tool in isolation. None tell you: is your org actually faster — and faster compared to whom?

Every renewal you auto-sign without this data is a bet you don’t have to make.

Copilot Dashboard

vendor view
Acceptance rate31%
Lines accepted12,847
Lines suggested41,412
Active users23

Last 28 days · 847 suggestions

“…and?” — your CFO, probably

The mechanism — no black boxes

How we know what AI actually wrote

No SDK. No agent. No code instrumentation. We read signals already in your Git history.

1

Connect

Install the AgentLens GitHub App and authorize the official Copilot Metrics API. Two clicks. We pull commit metadata and delivery metrics — never your source code, diffs, or commit message contents. First data lands in ~10 minutes.

GitHub App
Connected
10 min setup · no code changes
2

Detect — two paths, never averaged

Explicit (High): Co-Authored-By trailers + official Copilot API — 95%+ accurate. Heuristic (Medium/Low): statistical signals for tools that leave no trace. We never blend an explicit signal with a guess.

commit 3f9a1c2

Author: Priya Nair

Co-Authored-By: Claude

→ Claude Code · High confidence

commit 8b2e4f1…

→ heuristic · Medium confidence

never averaged · labeled honestly
3

Benchmark & export

We aggregate your metrics, strip every identifier, and place your team against 100+ orgs your size. Export a board-ready PDF ROI report in one click.

0thYou · 73rd ▲100th

100+ orgs · same headcount + stack

zero identifiers · PDF export

Early access results

What one team found in 30 days

anonymised early-access customer · results will vary

65-person Series B fintech eng org · London

Installed Nov 2025 · 30-day snapshot

Stack: Cursor + Claude Code + Copilot

+312 hrs

reclaimed / mo

$58,400

value of time / mo

6.3×

ROI on AI spend

Claude Code
68% adoptionHigh
Cursor
41% adoptionMedium
Copilot
22% adoptionHigh

“42nd percentile for cycle-time improvement — adopting fast, not converting to speed. That’s the kind of insight that changes your Q3 planning conversation.”

Key insight from AgentLens benchmark · first 30 days

linkedin.com
Alex Kim

Alex Kim

VP Engineering at Monzo

2d · 

6 months paying for Cursor, Copilot, and Claude Code. Couldn’t tell the CFO if any of it was worth it. 😬

Finally tried AgentLens. Setup took 9 minutes.

First insight: Copilot acceptance rate looked fine. But we were at the 42nd percentile for cycle-time improvement vs orgs our size. Adopting fast — not converting to speed.

…see more

#engineeringleadership #aicoding #devtools

127
23 comments8 reposts

anonymised · name + employer changed · results will vary

Want to see your org’s number? Book a 15-min demo →

I walked into Q3 planning with the AgentLens ROI export and the benchmark percentile. It was the first time the AI conversation wasn’t a debate — it was a number.

VP

VP Engineering

Series B fintech · ~120 engineers

🔒Source code never stored
🏢RLS-isolated per org
🗑GDPR · hard delete on cancel
🔐SOC 2 in progress

Built by engineers who ran platform teams. Backed by capital that lets us do this right.

0+

commits analyzed / mo

0

orgs in benchmark cohort

0.0×

avg measured ROI

<0 min

median time to first data

Dashboard preview

The four numbers your CFO actually asked for

Same engineers measured against their own pre-AI baseline — not a vendor’s cherry-picked cohort.

Time reclaimed

+312 hrs/mo

vs. pre-AI baseline

Value of time

$58,400/mo

at $187/eng·hr loaded

Tool spend

$9,200/mo

all AI seats

ROI

6.3×

↑ from 4.1× last quarter

Cycle time · AI-assisted vs. non-AI PRs

Days from first commit to merge · last 6 months

Tool breakdown

ToolAdoptionConfidence
Claude Code68%
High
Copilot22%
High
Cursor41%
Medium
Windsurf10%
Medium
Codex6%
Medium

The moat — no competitor gives you this

“Are we good, or just busy?”
Now you’ll know.

Your tools tell you what you did. AgentLens tells you how that compares to 100+ engineering orgs of your size — anonymously, both directions.

Example insight

“Your team ships AI-assisted code at the 73rd percentile for companies your size — but your cycle-time gains are only 42nd percentile. You’re adopting fast; you’re not yet converting adoption into speed.”

That’s actionable. Not a vanity number.

AI adoption distribution · 100+ engineering orgs · similar headcount

You · 73rd percentile
0th25th50th75th100th

73 of every 100 comparable engineering orgs have lower AI adoption than yours.

Aggregate-only.

We anonymize before anything enters the benchmark. Zero customer identifiers. Ever.

Both directions.

You're a row in everyone's benchmark. That's why the dataset is worth something.

Sized cohorts.

Compared to orgs your headcount and stack. Not a 5-person startup or a FAANG.

ROI calculator

What is your AI spend actually worth?

Rough estimate — AgentLens replaces this with measured data from your actual Git history.

50
5500
$180.0K/yr
$80K$350K
$9.0K/mo
$500$100K
2h
0.5h8h

Industry avg — AgentLens replaces this with your measured data

Your estimate

Hours reclaimed / month+433 hrs
Value of time reclaimed$48.7K/mo
AI tool spend$9.0K/mo
AgentLens Growth
$99/mo1.1% of spend
Estimated ROI5.4×

AgentLens pays for itself in 1 day of value reclaimed.

This is a rough estimate. AgentLens replaces it with measured data from your actual Git history — and shows whether your ROI is above or below the 73rd percentile for orgs your size.

Get the measured version — book a demo →

Pricing

Less than 1.5% of the AI spend you're already paying.

Your AI seats cost the same whether you measure them or not. AgentLens tells you if that spend is working — for a rounding error on the bill.

14-day free trial · 1 repo · no credit card
Every trial includes
Full benchmark on 1 repo
Board-ready PDF, watermark-free
No credit card · data deleted if you walk

Starter

≤ 30 engineers

Know if your AI spend is working.

$49/mo$69

Billed $588/yr

Pays for itself finding 3 unused seats · free from there

What you're measuring

$2–5K/mo

in AI tool seats / mo

~1.5%

of that to measure it

↑ this much of your AI bill, to know if all of it is worth keeping

One wrong renewal = $30K wasted. AgentLens annual = $588.

  • Claude Code · Copilot · Cursor · Windsurf · Codex
  • Benchmark percentile vs 100+ orgs
  • Monthly board-ready PDF export
  • High / Medium / Low confidence labels
  • Team-level only (privacy-first default)
  • 10-minute GitHub App install
Start free trial →

14 days free · no card needed

Most popular

Growth

≤ 100 engineers

Defend the AI budget at the board level.

$99/mo$149

Billed $1,188/yr

Pays for itself finding 5 unused seats · free from there

What you're measuring

$8–18K/mo

in AI tool seats / mo

~1%

of that to measure it

↑ this much of your AI bill, to know if all of it is worth keeping

One wrong renewal = $100K wasted. AgentLens annual = $1,188.

  • Everything in Starter
  • Weekly Slack ROI digest
  • Admin-only team breakdowns
  • Quarterly benchmark cohort update
  • Trend comparisons quarter-over-quarter
  • Priority support + onboarding call
Start free trial →

14 days free · no card needed

Scale

≤ 250 engineers

For orgs where the AI bill is a board-level line item.

$219/mo$299

Billed $2,628/yr

Pays for itself finding 11 unused seats · free from there

What you're measuring

$20–50K/mo

in AI tool seats / mo

~0.7%

of that to measure it

↑ this much of your AI bill, to know if all of it is worth keeping

One wrong renewal = $200K+ wasted. AgentLens annual = $2,628.

  • Everything in Growth
  • SSO + SAML
  • Audit log
  • Custom benchmark cohort filters
  • Dedicated customer success
  • SOC 2 report on request
Start free trial →

14 days free · no card needed

250+ engineers? Let's build a custom plan.

SSO · audit logs · custom cohorts · dedicated CS · SOC 2 on request

Talk to us →

Not convinced in 14 days? You owe us nothing.

Full data delete on cancel. No invoice. No awkward call to “review your options.”

No card to start
Cancel in 1 click
Data deleted in 30d

FAQ

The questions a skeptical VP actually asks

73
Ready in 10 minutes

Walk into Q3 planning with the answer, not a debate.

See your team’s AI ROI and your benchmark percentile against 100+ orgs. Source code never leaves your org.

No credit card · No SDK · Full data deletion on cancel