Organizations are operating blind.
Companies can see how much AI each person uses: seats, prompts, tokens and acceptance rates. They can't see who knows when to lean on AI, when to check it, and when not to reach for it at all. That's where the risk sits, and where the value is.
Book a demoPiloting with engineers this fall.
Built first for engineering, where AI agents are already doing real work, with more roles to follow. DiscernIt is opening a small number of design partnerships for 2027.
How it works
A realistic simulation.
Engineers work alongside an AI coding agent, deciding which of its changes are safe to ship and what to do when something breaks.
Measured on what people do.
DiscernIt looks at behavior in the simulation (what gets accepted, what gets checked, what gets through), not self-report.
Repeated to show what changed.
Results show where to invest, how to delegate and which teams already set the bar, and measuring again shows whether it worked. Each engineer gets a private write-up of their patterns, with suggestions for getting more out of working with AI.
Start standard, then make it yours.
1. Prove it holds up.
One standard simulation, scored independently, to show the measurement is reliable.
2. Fit it to your work.
Scenarios built from your own work on the same frame, so results still compare.
3. Set norms from evidence.
Teams decide how they work with AI based on what good judgment looks like in practice.
On the side of the people doing the work.
The people being measured get as much out of it as the people leading them, starting with a private read on their own patterns and something concrete to try next.
Founder
DiscernIt exists because Sophia Romee held two jobs at once. At College Board she founded the GenAI Studio, leading a 25-person team that built assessment simulations measuring hard-to-observe skills, while also running AI integration across the organization's 2,000 staff. The hard part was never the tool, it was that organizations default to how they have always done things and have no way to tell who inside them has a better answer. The team includes experts in measurement, learning science and AI engineering.
Chat with Sophia