The public,
before you
go public.
A model of how millions of people will respond. Ask it before you speak, launch or decide. Then watch it get graded by what people actually do.
Every decision that meets the public is a guess.
Until now.
Campaigns test ads after they’re made. Companies learn how an announcement landed after it’s out. Governments find out a message failed after it has. The answer always arrives too late to use.
Three ways to know what people think.
Ask them.
Polls and focus groups. Weeks per answer, a few thousand people, stale on arrival.
Simulate them.
AI populations answer in minutes, but they are checked against surveys: what people say.
Simulate, then prove.
Omni answers in minutes, tests the answer on real people in days, and is scored on what they do.
What people say is not what they do.
When researchers asked AI twins the same questions as real people, the answers missed by about 12 points on average, and missed most for the groups least like the people who build AI. A model of the public is only as good as the reality it is held to.
Ask the public anything.
Will voters in this district believe this ad?
How will patients read this risk statement?
What will employees hear in this announcement?
Which price change will customers forgive?
How will this town take the new policy?
Minutes.
A model of the people you care about reacts to your message, segment by segment, with honest uncertainty.
Days.
When it matters, the answer is tested on real people before you commit, and the two are compared side by side.
Every time.
Every result, right or wrong, teaches the model. It gets sharper with every question anyone asks.
Graded
in public.
We publish every forecast before the result is known. And every miss after. Trust in a model of the public should be earned in public.
First
Politics.
Because every November the answer key is published. Hundreds of races, graded by real votes. No better place to prove a model of the public, and no harder audience to please.
Then
Everyone else.
Health and pharma. Public health. Financial services. Corporate communications. Any organization that has to speak to the public and can’t afford to be wrong.
Built for both sides of every argument.
Your work stays yours.
Every client works in a sealed workspace. Nothing you test trains a model anyone else uses.
No shortcuts on data.
Consented panels and permitted sources only. Nothing harvested from people who never agreed to it.
Bias, measured and shown.
Every result reports where the model leans, by party, race and education, so you can see it before you trust it.
No fakes. Ever.
Omni tests messages. It never imitates a real person’s face or voice.
Know the response
before you run.
We’re opening Omni to a small group of agencies and campaigns for the 2027 races.
Request access