The state of your AI transformation.
A report on how a company really uses AI in engineering. Five questions, 25 checks, scored per team from the work itself.
Trusted by engineering teams
no new people
676 needed before AI · 1.9× per developer
- This is what the board sees.
- Split the performance, or the number lies.
- Both numbers are an average of four very different teams.
- So you open one team and ask why.
- And this is where the work actually is.
This is what the board sees.
350 engineers now deliver what 676 would have before AI, for $226,900 a month. Three numbers say whether that works: capacity added, $696 for each added engineer, and 48.6% of it on the roadmap. All correct. None of them says what to do next.
Split the performance, or the number lies.
A gain means nothing until you know the category. Here 51.8% of the spend built features and 31.2% kept old code running or fixed it. Only 48.6% answers to a named objective, and 28.7% to no stated outcome at all.
Both numbers are an average of four very different teams.
The Platform team pays $541 a month for each added engineer. The Data & ML team pays $1,082. Nobody decided this, and one company-wide program will not fix it.
So you open one team and ask why.
Good at everything except one thing: nobody tells them what to build. So the fastest team in the company keeps building things nobody asked for. Capacity and the scores are measured separately: this is what sits behind the number, not what caused it.
0 = not started · 5 = fully there
And this is where the work actually is.
When writing code gets cheap, the hard part is knowing what to build. For the Web app team, the row holding it down is objective context.
- Ticket clarity1.8
- Epic decomposition1.4
- Roadmap coverage2.2
- Objective context1.2
NextWrite down why each objective exists, not only what it delivers
- Spec readiness1.6
This is what the board sees.
350 engineers now deliver what 676 would have before AI, for $226,900 a month. Three numbers say whether that works: capacity added, $696 for each added engineer, and 48.6% of it on the roadmap. All correct. None of them says what to do next.
no new people
676 needed before AI · 1.9× per developer
Split the performance, or the number lies.
A gain means nothing until you know the category. Here 51.8% of the spend built features and 31.2% kept old code running or fixed it. Only 48.6% answers to a named objective, and 28.7% to no stated outcome at all.
- FeaturesNew value51.8%$117,534
- MaintenanceKeeping it running22.3%$50,599
- TestsCoverage11.9%$27,001
- DocsWritten down5.1%$11,572
- FixesBugs & slop8.9%$20,194
Both numbers are an average of four very different teams.
The Platform team pays $541 a month for each added engineer. The Data & ML team pays $1,082. Nobody decided this, and one company-wide program will not fix it.
- Platform team2.4×$541 each105 people, delivering like 251
- Mobile team1.8×$641 each82 people, delivering like 146
- Web app team1.7×$787 each93 people, delivering like 156
- Data & ML team1.8×$1,082 each70 people, delivering like 123
So you open one team and ask why.
Good at everything except one thing: nobody tells them what to build. So the fastest team in the company keeps building things nobody asked for. Capacity and the scores are measured separately: this is what sits behind the number, not what caused it.
0 = not started · 5 = fully there
- Capacity
- 1.7×
- AI spend
- $49,550
- Cost per engineer
- $787
And this is where the work actually is.
When writing code gets cheap, the hard part is knowing what to build. For the Web app team, the row holding it down is objective context.
- Ticket clarity1.8
- Epic decomposition1.4
- Roadmap coverage2.2
- Objective context1.2
NextWrite down why each objective exists, not only what it delivers
- Spec readiness1.6
Objective context at 1.2 is the row holding this down. The other four are between 1.4 and 2.2.
What a board actually receives.
Eight pages, built from the work itself. See exactly what it covers, then decide whether you want one about your own company.
The state of your AI transformation.
Five questions, 25 checks, scored per team from the work itself, never from a survey.
- Capacity
- +326none hired
- Cost
- $696each
- Roadmap
- 48.6%of the work
350 engineers · 4 teams · 27 repos
Last 30 days
An invented company, so you can see the shape of the answer before you send us a single repository.
Check any number yourself.
25 checks across 4 teams. Read a row for one team, a column for what is weak everywhere, and open any cell for the evidence behind the score and the next step that raises it.
Swipe for every question · tap a cell to open itFive bars in every cell · click a cell to open it
No number here is an average. Each question scores as its weakest check, and the overall score comes from how much each team improves, reported as capacity, cost and roadmap alignment. The evidence behind all 25 checks is in the method page of the example report. Every cell is a check you can run nightly: see all 25 →
A finished transformation looks like capacity you did not hire.
Every question here moves one number: how much one engineer delivers. It is measured the same way for everyone, so a company can be ranked against the ones below, whether its repositories are public or private.
What 100 engineers at each organization now deliver, on their public repositories. The company in this report is measured the same way, and ranked with them.
- #1OpenAI19.3× since Apr 2025+1,828per 100 real
- #2Microsoft4.9× since Apr 2025+390per 100 real
- #3Cloudflare4.1× since Apr 2025+309per 100 real
- #4Vercel3.3× since Apr 2025+231per 100 real
- #5Google3.0× since Apr 2025+196per 100 real
- #6Meta2.8× since Apr 2025+175per 100 real
- #7This report’s company1.9× · 350 engineers+93per 100 real
- #8UiPath1.2× since Apr 2025+20per 100 real
Same period, same measurement, 16× between top and bottom. The median company reached 3.31×. The difference is not talent, and it is not budget. Your own company is placed on this scale from your own repositories, public or private.
The same measurement on your own repositories, priced per engineer added, is AI ROI. What that capacity costs, and whether anyone asked for it, is token spend intelligence.
What a buyer actually asks.
The same 25 checks, read by the person writing the cheque.
- 45%from one team
Is the gain real, or is it a few people?
Platform team carries 45% of the added capacity with 105 of the 350 engineers. A strong result and a concentrated one. The report shows the split per team and per repository, before you price it.
- 51.4%off the roadmap
What happens if the AI budget is cut?
$226,900 a month is buying 326 engineers of capacity, and 51.4% of the work it funds is not on the roadmap. The report separates the spend that is working from the spend nobody asked for.
- +161already on payroll
What is the upside still inside the company?
Every team at Platform team’s rate would be 837 engineers of capacity instead of 676. That is 161 engineers you are already paying for and not yet getting.
Everyone says they use AI.
Almost nobody can show it.
We build this for your company, or for one you are buying. Read-only setup, under an hour, built from history you already have.