Straight answers, including the awkward ones.

A dress rehearsal for something you're about to say. You describe the situation in a few sentences. PRISM builds an AI-simulated audience from the people who will actually hear it — your customers, your staff, residents, the press, the regulator — and lets them react to the real material. A panel of AI reviewers in named seats then argues about what those reactions mean, and a reviewer whose only job is to break the draft attacks the conclusion.

What comes back is a Decision Report: the recommendation first, then the reasoning behind it, laid out so you can check it. The report is ready about a minute after a short guided conversation.

No. It is a simulation. PRISM models how audiences are likely to react to what you're about to say. It does not foretell events, it is not a poll of real people, and it is not a substitute for your judgment.

That distinction is built into the report rather than buried in a disclaimer. Every finding states the observable event that would reverse it. The confidence reading can only be lowered, never raised. And afterwards PRISM asks you what actually happened, so a briefing can be scored against reality instead of remembered as right.

You can, and it will answer in seconds, fluently, with exactly the same confidence whether it is right or badly wrong. Nothing behind that answer can be checked, and nobody ever finds out whether it held up.

PRISM was built for the opposite job. Each finding points at the specific simulated reactions it rests on. Where the reviewers read the evidence differently, both readings are printed rather than averaged away. The confidence reading is computed from what the evidence supports and can only move down. And the finished report is sealed when it is delivered — its fingerprint recorded — so it can't be quietly rewritten later to look right.

Presaga Crisis is for reacting: something happened — a recall, an outage, a story about to break — and you have to respond. Presaga Campaign is for initiating: a launch, a price change, a policy, a campaign line you're about to put in front of people. Presaga Pulse is for a conversation already happening without you — you point it at the thread and ask where it's heading, and whether stepping in helps or hurts.

If you're not sure, start a briefing and describe your situation. The guided conversation works out which one fits and tells you what it picked. There's no penalty for coming through the wrong door.

From the stakeholder groups you describe — customers, employees, residents, journalists, regulators, whoever is actually involved. PRISM adds the real public bodies your situation draws in rather than inventing them: when a briefing needs an institution you didn't name, it sends a place and a role to a web-search provider, never your content, and writes down what it looked up. It also reads the live pages you point it at, and records exactly when it read them.

The audience is simulated, not sampled. It is not a group of real people, it carries no margin of error, and a finding from it should never be quoted as though it came from a survey. What it gives you is a structured rehearsal you can interrogate — with every reaction on the record and every exposure logged, so PRISM can say what each simulated person had seen before it spoke.

No. Crisis and Campaign will draft candidate statements from what you've told them — usually two or three, each taking a different approach. Every sentence is built only from facts you stated: PRISM will not invent a commitment, an offer, a number, a discount, or an apology you didn't describe.

You pick the candidate closest to your instinct, edit it freely, and accept it. Nothing is tested until you accept it. You can also skip drafting entirely and test the decision itself — “how would people react to each of my options?” — with no statement at all.

A Decision Report. It opens with The Call — the single recommendation the evidence supports, written to be shared as it stands with the people who decide. Then it shows its work: findings with the simulated reactions behind them; where the reviewers disagreed, printed rather than smoothed over; what would change the call, as observable events you can watch for; recommended actions, each with an owner and a timeframe; a confidence reading with its reasons attached; and your original input reproduced exactly as you submitted it.

Each product has its own sections for its own question, but every report is built the same way — learn to read one and you can read them all.

Five ways. Read it in the app. Publish it on your own branded client portal at your own subdomain, where invited people sign in and see only the projects you grant them. Download it as a PDF. Or send a secure link. There is also an audio briefing you can listen to on the way in.

Consultants and agencies use the portal to deliver reports to their clients under their own brand.

Your account, documents and reports are hosted and stored in Canada, encrypted at rest. On the standard service the AI model that processes your content runs outside Canada; if your rules require in-country processing end to end, we run PRISM for you on dedicated Canadian hardware as a private deployment.

The AI providers involved are named in our privacy policy, together with what leaves and where it goes. Content is encrypted before it is written to the database or a backup, so a stolen file is unreadable — and our Security page states plainly what that protects and what it doesn't.

No. Our privacy policy states it in its own words: “Your document is not used to train or fine-tune AI models.” The same policy names the AI providers our engine sends content to.

We link to the policy rather than paraphrasing it here, because the version you can hold us to is the policy itself.

Yes, yourself, from your settings — an individual briefing, a project, or the entire account. Account deletion is confirmed with your password — and your two-factor code if you have turned it on — then removes your content and uploaded files. Financial records are anonymized and kept only as required for accounting.

Until you delete something, we keep it. We don't quietly age your work out, and we don't repurpose it.

Yes — as a private deployment. That can mean dedicated Canadian hardware, an isolated environment in a region you specify, or PRISM installed inside your own data center. Outbound capabilities such as web research can be disabled for a private deployment; when one is off, the report says so rather than guessing.

These are scoped with you rather than switched on. Tell us the rule you have to satisfy and we'll tell you what meeting it takes. See Deployment & sovereignty.

One credit = one full briefing. That's the whole report, every delivery format included — there is no report-type selector and no per-report charge. Today that's current pricing.

For Presaga Crisis and Presaga Campaign, a follow-up re-test of the same material is priced in tokens; once the situation has genuinely changed, that's a new briefing. For Presaga Pulse, every re-read is priced in tokens as a follow-up — however much the conversation has moved, a re-read is never a new credit. You're charged when the report is delivered; if a run fails, what you were charged goes back automatically.

No. There are monthly plans and one-time packs, and both grant credits and tokens. Your currency is set automatically from your region.

Every figure on this site is read from the same catalog the app bills from, at the moment you load the page — nothing here is a price typed from memory. The pricing page is the live list.

It can be, and the report is built on that assumption. The confidence reading can only be lowered, so a genuinely contested situation earns a low number with its reasons beside it instead of false comfort — that is the system working, not failing. Where the reviewers split, you read both readings.

Every finding names the observable, datable event that would prove it wrong. Those become your watch-list for the first days after you act. And afterwards PRISM asks you what actually happened, which turns each briefing into a claim that can be checked — and, over time, into a record of where it was right and where it wasn't.

That's the point of the format. Each finding names the simulated reactions it rests on, so you can follow any conclusion back to what produced it. A claim that can't be traced to evidence is dropped and counted, never smoothed into the prose. Your original input is reproduced in the report exactly as submitted, so anyone reading it can see what the scrutiny actually examined.

The report is sealed the moment it's delivered and its fingerprint recorded, so it can't be edited afterwards to match how things turned out.

Create an account, click Start a new briefing, and describe the situation in a few sentences of your own words. A short guided conversation fills in the rest — usually two or three exchanges, mostly by clicking rather than typing — and a readiness check tells you whether there's enough for a trustworthy result before you spend anything.

Then it runs, and the report is ready about a minute later. If your situation is already a live conversation somewhere, paste the link: that's what makes Pulse different from asking any AI what it thinks.

Presaga is founded and run by Craig Bondy, and is built and operated in Canada. He writes on the company blog under his own name.

About Presaga covers who builds it and what it's for. Our philosophy is the standard we hold ourselves to, published so anyone can hold us to it.