← About this survey

AI Spend & Return

Open for submissions

Twelve questions, four to six minutes. Three are required: total FTE, engineering FTE and total AI spend over the last full calendar month. Everything else is optional. Nothing is shown for any cut with fewer than eight organisations.

Opens
13 October 2026
Closes
30 November 2026
Contributors
Not yet reportable

What this round asks

  1. 1.
    How many FTE did your company employ last full calendar month?Total FTE · FTE · definition
  2. 2.
    How many of those worked in engineering, including QA, data and devops?Engineering FTE · FTE · definition
  3. 3.
    What was your total AI spend last full calendar month, in EUR?Monthly AI spend · EUR / month · definition
  4. 4.
    How much of that went to AI running inside your product, rather than internal tooling?Product AI spend · EUR / month · definition
  5. 5.
    What is the most one developer may spend on AI tools per month, in EUR? Enter 0 if there is no limit.AI spend cap per developer · EUR / month · definition
  6. 6.
    How was your AI budget for the coming year arrived at?
  7. 7.
    Where does AI spend sit in your P&L?
  8. 8.
    What is your policy on AI tools staff buy themselves?
  9. 9.
    Do you measure the productivity gain from AI in engineering?
  10. 10.
    How much time do you reckon AI saves per engineer, as a percentage?Claimed time saving per engineer · % · definition
  11. 11.
    What share of the code you merged last full calendar month was predominantly written by AI, as a percentage?AI-written share of merged code · % · definition
  12. 12.
    How many roles fell vacant in the past 12 months and were deliberately left unfilled because AI absorbs the work?Roles deliberately left unfilled · count · definition

Results: /results/ai-spend-2026-h2 — a permanent address; the figures appear there once this round is published.

What you give

One submission per company, roughly ten minutes. You can save a draft and come back to it. Every numeric question shows its exact definition beside the input, so you are never guessing what is being asked.

What you get

The peer distribution for every question — p25, median and p75 for numbers, and the spread across options for choices. Filter it down by HQ country, ARR band, headcount, engineering headcount and funding type to compare against companies that actually resemble yours.

Any slice matching fewer than eight companies comes back suppressed rather than as a number. That is deliberate: it is what stops the benchmark being used to work out any single company's figures, including yours.

Take part

Create an account with your work email and the survey is waiting for you. Once we've checked your answers, the results for this survey unlock for your company.