Research the audiences that matter, at scale.
Fortune 500 CEOs. Members of Congress. Institutional investors. Radiant builds artificial societies of the people your decisions depend on, grounded in anonymised real human behaviour. Survey, interview and convene thousands of personas at once, in hours, not weeks.
Would you pay 10% more for a supplier that runs on clean energy?
- Definitely
- Probably
- Neither willing nor unwilling
- Probably not
- Definitely not
Ask the questions you couldn’t ask before.
Reach the audiences that matter by the hundred, not the five you’d get on expert calls. Test plans too sensitive to show anyone. And get answers while the decision is still open, not weeks later.
Which of these three enterprise pricing tiers would Fortune 500 CEOs approve, and which one stalls in procurement?
Fortune 500 CEOs
What will each analyst ask on the earnings call, in their own words, before a single number is public?
Global Institutional Investors
How do Democrats and Republicans in Congress read the same announcement, and where do their readings diverge?
119th Congress
If the price rises 18% in January, which shoppers switch to the store brand first, and why?
UK General Population
What do you want to know?
How would you prefer to pay for an AI agent that resolves your team’s engineering tickets?
In a sentence, what would make a per-ticket price feel fair to you?
What would stop you signing in the first year?
How would you prefer to pay for an AI agent that resolves your team’s engineering tickets?
Paying per resolved ticket leads at 38%, with per seat close behind at 31%. Hourly and flat pricing barely register.
The split follows sector. Finance leaders hold on to seats because they can forecast the bill, while technology founders and engineering heads want to pay only for work that ships.
You chose per seat over per resolved ticket. What would make paying per ticket work for you?
Thinking...A definition of ‘resolved’ that my team signs off, not yours. If the agent closes a ticket and we reopen it a week later, I’ve paid twice. Give me a thirty-day reopen window and a cap on the monthly bill, and I’d move to per ticket tomorrow.
The survey split between paying per resolved ticket and per seat. What would make you trust a per-ticket price?
A definition of ‘resolved’ we sign off, and a thirty-day reopen window. Otherwise I’m paying twice.
@BI want per ticket. I only pay when the work is done, and my board gets that.
@APer seat is easier to forecast. Founders get punished for spend that jumps month to month.
@DEngineers will game whatever you count. Count tickets and you’ll get smaller tickets.
@CWe can’t sign anything without a cap. We budget a year ahead.
@EThen cap it. Per ticket, with a ceiling, and I’d switch from seats.
Who should decide that a ticket is done: your team, the agent or the vendor?
Our team. The agent marking its own work is the whole problem.
@CWhoever reopens it. If it comes back within thirty days, it was never resolved.
An audit trail we can check, so it isn’t a matter of opinion.
And if your team and the vendor disagree about a ticket, who settles it?
A monthly review with a named owner on each side. Disputes can’t sit on the invoice.
@DDefault to us. We’re the ones paying.
If there were a monthly cap, how would you want it set?
From last quarter’s volume, plus a margin, reviewed twice a year.
@DLow enough that the board never has to hear about it.
@AA hard ceiling. A soft one is just a forecast.
How would you want to try per-ticket pricing before committing?
One team, one quarter, both prices side by side.
@EOn real tickets, not a demo backlog.
The room moved to paying per resolved ticket once two conditions were on the table: a monthly cap and an agreed definition of ‘resolved’. The warning from the Senior Software Engineer stands: a per-ticket price invites smaller tickets.
Every participant tied trust in a per-ticket price to who decides a ticket is done. Without an agreed definition, they expect to pay twice for reopened work.
“A definition of ‘resolved’ we sign off, and a thirty-day reopen window. Otherwise I’m paying twice.”
The investor and the university judge a price by whether they can budget for it a year ahead. A monthly cap answers both.
“We can’t sign anything without a cap. We budget a year ahead.”
A ticket the two sides disagree on should be settled in a monthly review, by a named owner on each side, before it reaches the invoice.
“A monthly review with a named owner on each side. Disputes can’t sit on the invoice.”
Before anyone signs, the room wants one team to run both prices side by side for a quarter, on its own backlog rather than a demo.
“On real tickets, not a demo backlog.”
Combine these insights with what I learnt from the survey and the interview
Per-ticket pricing wins, on two conditions
You asked how US tech leaders would want to pay for an AI agent that resolves their team’s engineering tickets. The survey, the interview and the focus group give the same answer: per resolved ticket, as long as two things are settled first. Everyone has to agree what counts as ‘resolved’, and the monthly bill needs a ceiling.
What each method adds
- Survey. Per resolved ticket is the first choice (38%), with per seat close behind (31%). Finance is the exception: nearly half its leaders want seats, so they can forecast spend.
- Interview. The VP Engineering would leave seats for a thirty-day reopen window and a cap on the monthly bill.
- Focus group. With both on the table, four of the five chose per ticket. The engineer’s warning stands: count tickets and you’ll get smaller tickets.
“Give me a thirty-day reopen window and a cap on the monthly bill, and I’d move to per ticket tomorrow.”
VP Engineering, Loomfield LabsInterviewWhat should we research next?
Combine these insights with what I learnt from the survey and the interview
Inside Radiant
An artificial society is 20 to 3,500 personas grounded in anonymised real human behaviour and connected in realistic networks, to model how opinions form in groups.
Design and run survey experiments across a whole society at once, and see not just what each persona selected, but why.
Pick any persona from the results and follow up one to one. Ask why they answered as they did, then keep asking.
Put personas in a room and let them argue it out. They push back, persuade and win each other round, so opinions shift as they do in real groups. See where the room splits, and who changed whose mind.
Export all the raw data, or ask our research agent: frontier AI, tuned for research analysis. It reads across every survey, interview and focus group in the project and gathers charts and quotes for your reports.