Ascenda / Scaleup Benchmark
How are scaleups actually working with AI agents?
Once an engineering organisation grows past a handful of teams, agent use stops being one habit and becomes many. Some teams run several agents at once, some carry the review load on two people, some work late because the agents do.
We’re inviting engineering organisations of roughly 16 to 200 people to help build the benchmark, and to see their own teams in it first.
Founding teams include
What founding organisations get
Your teams first, then the comparison.
01
Your organisation, by team
See concurrency, steering, review and rework for each team that reaches ten contributing engineers, and for the organisation as a whole.
02
Early access to Ascenda Flow
Each engineer sees their own working patterns on their Mac, free. Using Flow doesn't require contributing to the benchmark.
03
The comparison, as it develops
As more scaleups contribute, you'll see how your teams compare with organisations of a similar size and tool mix. We're building this comparison, and it will show how many organisations stand behind it.
04
Founding cohort access
Receive early findings and help decide which questions Ascenda investigates next.
What we’re investigating
Questions that only appear at scale.
These are measured from the AI tools themselves, from the first day. Which ways of working hold up needs longer, and comes after.
Concurrency: How many agent sessions does each engineer run at once, and does it differ between teams in the same company? When people run more, does more get done, or does more of their time go to supervising?
Who carries the review load: As agent output grows, is checking it shared across a team, or does it fall on a few senior engineers? Does that change as the organisation adds people?
Steering: How often do engineers correct or redirect an agent, and does that change as a team gets more experienced with them?
Long sessions and context: How long do agent sessions run before context limits set in, and how do teams deal with it?
Out-of-hours work: Is agent work moving into evenings and weekends across time zones, because agents can keep going when people stop?
Rework inside agent sessions: How often does agent work get undone before it leaves the session, and does that change with concurrency or session length?
Tool mix: Which agents and editors do engineers use day to day, and do teams that standardise on one work differently from teams that mix several?
How it works
One setup call, then engineers use their tools as usual.
1
Agree the scope
A setup call with you and whoever owns developer tooling. You choose which teams to invite first.
2
Engineers connect their tools
Each engineer installs Ascenda's open-source collectors for the tools they use, one command each. Flow can import existing local history, so the first read covers weeks already worked.
3
Each engineer decides
Consent is per person, in one question that lists exactly what it shares. Engineers can take part back at any time.
4
Read it by team
Results arrive in aggregate. Any group under ten people is withheld.
The wall
Aggregate only. Never an individual.
Flow reads your AI tools’ local history and live activity on your Mac. What it sends is counts, timings and tool events, never your prompts, code or model outputs. Your journal and reflections stay on your Mac unless you turn on sync.
Every figure covers at least ten people and is withheld below that, so a team of eight is read as part of a larger group. Today every engineer opts in.
Read the published rules →What an organisation never receives
- Any individual, at any level of detail
- A performance, productivity or quality score for a person or a team
- Who joined, who did not, or who left
- Source code, prompts or model outputs
Apply
Apply for the founding cohort.
For engineering organisations of roughly 16 to 200 people where engineers use coding agents daily. We select organisations across different sizes, tools and ways of working, so comparisons are useful. Smaller team? The founding benchmark is for teams of 2 to 15.
Want a 30-day Report on one cohort instead? Book a Study.