How it works

One scan. Three readings. Every finding cited.

A scan reads your message the way each of your audiences will. It returns three readings — Exposure (what could cost you), Clearance (what must pass) and Reception (what will land) — and behind every finding: the mechanism, the source, the confidence, and the fix.

The loop

Scan. Fix. Re-scan. Publish with the record.

Add your message
A campaign, an announcement, a policy, an email. Paste it or upload it.
Choose your lenses
Pick the audiences who will actually read it. Each lens is one audience, resolved from the Corpus.
Read the findings
Three readings, every finding pinned to the exact phrase, explained and cited.
Fix and re-scan
Apply the rewrites, run it again. The re-scan is your sign-off record.

The three building blocks

Three building blocks turn a library into a foundation.

Block 01

Lenses

A lens is one audience, resolved: the verified research from the Corpus describing how that audience actually reads. Whether identity is held individually or collectively, whether people move toward gain or away from risk, how much effort the wording demands, how much trust the sender starts with. Scans run through lenses, never through an average.

Block 02

Collections

Expert-curated sets of Corpus content for one industry or theme. A collection puts the relevant profiles, briefs and films in one place, and it is the same content the scan reads, so your team is never working from a different source than the check.

Block 03

Kits

Everything for a use case in one box: the verified resources and assets, ready to deploy to a team. A kit is how a function starts working from shared evidence without a rollout project.

See what the Corpus holds →

The three readings

Risk, sign-off and reach are three different questions.

One scan covers all three. Open any reading to see what it returns, and the worked finding behind it.

“What will this cost us?”

For · Corporate affairs · Communications · Legal · Brand safety

Nobody budgets for the disaster that did not happen. Exposure finds the element that becomes the headline before it is permanent, and shows you why it misfires, for whom, with the source behind the call.

Framings that alienateA phrase that reads as dismissal or exclusion under a specific, citable mechanism.

Symbols and timingImagery, gestures, dates and colours that carry meaning you did not put there.

Claims that invite scrutinyWording likely to draw regulator or press attention in your sector.

Red linesThe small set of things that must never ship, held as hard rules rather than model judgements.

Exposure finding · pinnedIllustrative

“A fresh start — leave the old ways behind you.”

How it read: For first-generation audiences, “leave the old ways behind” reads as a demand to discard heritage. The campaign becomes the story.
The mechanism: Conflicts with bicultural identity integration: belonging is additive rather than substitutive for much of this audience.
The change: “A fresh start — bring what matters with you.” Keeps the promise, removes the demand.

Benet-Martínez & Haritatos (2005) · confidence: high · reviewed lens · rewrite attached

What separates a finding from an opinion.

Frontier AI does the scanning and structures the evidence. Mosaica governs everything the scan delivers: the source citations, the criteria, the confidence thresholds, the red lines, and the expert review behind every source.

That governance is why a finding can be interrogated. It names a mechanism, cites the entry it comes from, states how confident we are, and proposes a change. You can disagree with any of those four things on the evidence. You cannot do that with a verdict that arrives as a number.

Confidence is published, not implied. Where the literature is strong, findings are rated high. Where it is thinner or more contested, they are rated moderate and say so. A tool that is equally certain about everything is telling you nothing about anything.

The rule that governs everything

We identify mechanisms, never traits.

We never say a group believes something. We say a framing conflicts with a documented mechanism — one more prevalent in a cohort, which varies by generation, acculturation and context, and which is always cited.

A mechanism is testable, citable and defensible. A stereotype is none of those. The test: if a reviewer from that community would wince, it is a stereotype and it does not ship. If they would ask for the citation, it is a mechanism.

The same scan, in your sector’s vocabulary.

Collections and lenses are curated per domain, so findings arrive against the stakes you actually carry, from sources that know your field.

Healthcare

Screening campaigns, patient letters, consent materials, public-health messaging.

Government

Service notices, compliance letters, emergency communication, campaign material.

Financial services

Product campaigns, disclosures, hardship communication, member letters.

Advertising & media

Creative testing before spend, localisation, pitch defence, brand safety.

Workplace & HR

Change announcements, policy rollouts, inclusion programs, training kits.

All industries

Every sector we hold collections and lenses for, with a worked example in each.

See all

See it run on your message.

Bring one real message going live this month. We will read it with you and you keep the findings.