Take the Engineering Leader Benchmark Q4 2026

Engineering Leader Benchmark Q4 2026Powered by Cloudflare

Website

1Your teamWhat's your role?

Five quick details first, so we can compare you with teams like yours.

2Your teamWhat kind of company do you work for?

Choose one.

Start-up (before Series B) Scale-up (Series B or later) Established company with its own engineering team Agency, consultancy or software house Public sector or non-profit

3Your teamHow many engineers are on the team, including contractors?

Choose one.

AI todayHow long has your team used coding agents in production work?

Agents edit files and run commands for you. Autocomplete alone doesn't count.

AI todayHow do most engineers on the team work with AI today?

Pick what describes most of the team, not the most advanced person.

Autocomplete in the editor They ask a chat assistant and paste the code in An agent edits files, and they approve most steps An agent finishes a task on its own, then they review it Agents start from tickets, schedules or CI and open pull requests themselves

AI todayRoughly what share of the code merged last quarter was written by an AI model?

Your best estimate of merged code that a model generated and a person kept largely as it was.

AI todayWhat does the team spend on AI tools per engineer per month, all in?

Seats, licences and API usage together, divided by the number of engineers.

Your toolsWhich tools does the team use to prototype or turn designs into code?

Choose up to three.

Ways of workingHow does a typical task reach an agent?

Choose one.

The ticket text or a prompt written on the spot A rewritten prompt with some background A short plan the engineer writes before starting A reviewed plan with constraints, split into small steps

Ways of workingWhere does an agent find your codebase's conventions?

Choose one.

Nowhere written down, engineers explain them each time In READMEs and wiki pages written for people In a context file per repository, such as AGENTS.md In context files that are updated in pull requests, like code

Ways of workingHow does the team share its agent setup: prompts, commands, skills and hooks?

Choose one.

Everyone keeps their own People share what works in chat It's stored in the repository It's stored, reviewed and has a named owner

VerificationBefore an agent-written change merges, what does a second person do?

Choose one.

Nothing, the author merges it Skims the diff Reads the diff carefully Reads and runs it, with extra care for data, auth and payments

VerificationWhen are tests written for agent-generated code?

Choose one.

Rarely By the same agent, after the code By a person or a separate agent, after the code Before the code, as the spec the agent has to meet

VerificationCan an engineer merge a change they can't explain?

Choose one.

Yes, it happens regularly It happens now and then Rarely, reviewers usually push back No, it's an explicit review rule

Platform and guardrailsWhere do agents run code and commands?

Choose one.

On laptops, with the engineer's full access On laptops, with some commands blocked In containers or cloud development environments In isolated sandboxes per task, with scoped credentials

Platform and guardrailsDoes every pull request get its own preview environment?

Choose one.

No, changes are tested on a shared staging environment For some repositories For most repositories Yes, every pull request deploys a preview automatically

Platform and guardrailsHow do your applications and internal tools reach AI models?

Choose one.

With API keys in each app or on laptops With shared team keys per provider Through one internal proxy or gateway Through a gateway with logging, rate limits, cost tracking and fallbacks We don't call models from our own code

Platform and guardrailsWhich gateway do your model calls go through?

Choose up to three.

MeasurementWhich delivery metrics does the team track?

Choose one.

None consistently Velocity or story points Lead time or deployment frequency Lead time, deployment frequency, change failure rate and recovery time

MeasurementDo you know what AI tools and model usage cost per team?

Choose one.

No Only the total invoice Per tool Per team or product, reviewed every month

MeasurementDo you track bugs and incidents caused by AI-generated code?

Choose one.

No Only anecdotally When it comes up in a post-mortem Yes, they're tagged and reviewed

Team and hiringHow do candidates show how they work with AI in your hiring process?

Choose one.

They don't, we don't assess it We ask about it in conversation In an exercise or take-home where AI is allowed In a live session in a real codebase with their own agent, scored on written criteria

Team and hiringHow do you check fundamentals without AI?

Choose one.

We don't check them separately Algorithm or whiteboard questions A take-home assignment A live session in a real codebase with AI tools switched off

Team and hiringHow do your current engineers learn to work with agents?

Choose one.

On their own Through occasional demos or talks Through pairing and shared guidelines Through a structured programme with time set aside and clear expectations per role

Results so farSince the team started using agents, how has the amount you ship changed?

Choose one.

Results so farAnd the quality of what reaches production?

Think of bugs, incidents and rollbacks.

Results so farWhere is the bottleneck in your delivery now?

Choose one.

InfrastructureWhere does most of your production run?

Choose up to three.

InfrastructureRoughly what is your monthly cloud and hosting bill?

Choose one.

InfrastructureDoes your product offer AI features to customers?

Choose one.

No, and none planned We're exploring or prototyping Yes, built on one model provider Yes, built on several model providers

InfrastructureWhat are your infrastructure priorities for the next twelve months?

Choose up to two.

Lowering cloud costs, including data transfer Faster performance for users in more regions Security: bots, DDoS and access for people and agents Moving off legacy infrastructure Running AI features or agents in production Data residency and compliance No major changes planned

The next twelve monthsHow do you expect engineering headcount to change in the next six months?

Choose one.

The next twelve monthsWhat limits your team most right now?

Choose one.

Finding senior engineers Budget Review and verification capacity Legacy code and technical debt Security or compliance approval for AI tools Unclear product priorities

Your resultWhat is one thing about working with agents your team hasn't solved yet?

Optional. One or two sentences is plenty.

Your resultLast step: where should we send the report?

Your result appears on the next screen. We use your email for the report and, only if you ask, to set up a call.

First name Business email Company (optional, never shown)

Send me the full report when it's published in January 2027. I'd like a 30-minute call to talk through my result.

We store your answers, first name and email, and publish aggregates only, never a group under five teams. Privacy policy