Every asset: briefed, made, checked, approved. In a day, not a quarter.
You get an agentic team that makes the work, copy, design, web and motion, a check that reads every claim before a person does, and a reviewer who signs a page rather than a pile. Your approved claims sit inside the work, not in a spreadsheet you re-explain to every freelancer.
Medical devices · pharma · financial services. Download the brochure
Why claims, not tone, are the real AI risk in regulated marketing: read the argument. New to Design Catalyst? Start with the product page: everything there applies here.
Everyone who makes your content is guessing at your claims.
The brochure needs a graphic designer. The product page needs someone to design it and someone to build it. The white paper needs a medical writer. The congress stand and the campaign cuts need motion. At your size that is a mix: one or two people in-house, a freelancer or two, an agency when the budget allows.
Not one of them knows your approved claims. They know what you told them, in a brief, once. The claims themselves live in a spreadsheet you keep open in a second window, and you are the only person in the building who can say whether a sentence is allowed.
So you check everything before it goes anywhere, and what you miss, the reviewer catches. That is why review takes a day: your reviewer is reading on behalf of the whole supply chain, not reading what is new.
And the clock runs the whole time. Six weeks from brief to approved is normal, which means the piece arrives after the moment it was made for.
You are not the bottleneck. You are the only copy of the rules, and every piece has to pass through the only copy.
Watch the waiting disappear.
Two minutes on how a brochure gets made with only approved wording, checked before anyone reads it, and signed off on one screen, with every decision on record.
A team that makes it, a check that reads it, a record that proves it.
Your brand and your approved wording, codified into an AI designer your team briefs in plain language, in the tools you already sanction. Every asset type comes from the same library, and a complete page is built and published in minutes.
A quality layer runs before a person reads: colour, type, shape and language on every asset, and the approved wording where it applies. Nothing ships that breaks a rule.
A named reviewer decides on a page, and every version, check and approval is kept, under their own login and with a time that cannot be edited afterwards.
Five steps, and your reviewer appears in one.
Someone in marketing describes what they need, in plain language. No template, no ticket.
The piece is built from your brand rules and your approved wording, and the checks run on it before anyone reads it.
Marketing sends it to the named reviewer, who gets one email with one link.
The reviewer reads the piece as a page and chooses: approve, ask for a change in their own words, or comment. Ask for a change and the revision comes back in minutes, on the same record.
On approval the piece is published as a version: the final file, the copy the approver read, the source, and the checks that ran.
An email. A page. Three buttons.
Your compliance colleague signs in once with their own account. After that a review is one link, and the decision is recorded under their name.
Marketing has sent you a piece. Open it, read it, and decide. You need open nothing else.
Review and decide →Checks: passed · requested by marketing · waiting for your decision
A complete page built and published in minutesC1, on brand and on approved wording.
It shortens review by roughly seventy per cent.
The claims check fails a release on an incomplete, expired or forbidden claimC3.
Certified compliant with the regulation.
The system proves the known and flags the suspected; the reviewer decidesC4.
Every sentence arrives marked: proven, flagged, or forbidden.
The page arrives marked. Anything already approved carries its rule; anything claim-shaped that nobody has approved is put in front of the reviewer; anything forbidden stops the release outright.
Proven against your library, and tagged to the rule it came from, with its approver and its dates.
Nobody has approved this sentence. Approve it as a claim, reword it, or leave it as context. Yours to decide.
Blocks the release. Nothing ships until it is gone, whoever approves.
| # | Approved wording | Approver | Review by | Markets |
|---|---|---|---|---|
| C1 | a complete page built and published in minutes | Marketing director | 2027-03-12 | EU |
| C3 | the claims check fails a release on an incomplete, expired or forbidden claim | Compliance owner | 2027-03-12 | EU, UK |
| C4 | the system proves the known and flags the suspected; the reviewer decides | Compliance owner | 2027-03-12 | EU, UK |
Flagged for the reviewer: 1. Forbidden wording found: 1. Both named, with the sentence and the decision needed.
One library at group.
A layer in every market.
Group owns the wording that travels. Each market inherits it and adds what only applies there. A market can add; it cannot quietly contradict.
Local wording and what the local rules require.
Local language and local product terms, member state by member state.
Its own rulebook, said in its own words, on the same brand.
The local reviewer reads what is local and new, not the campaign from the beginning.
On your accounts, not ours.
The rule store, the record and the approvals sit in your own repository, under your own logins, your own retention rules and the region you choose. The model your team builds with is your own account with a vendor you have already assessed. We introduce no new platform, and after handover nothing routes through Catalys AI. Certifications belong to the providers you already use, and your security team obtains their current reports from each provider directly. There is a note for your security team that names every component, who operates it, where the data sits, and what leaves your organisation.
Four pages, written for security and third-party risk teams to read before a questionnaire is sent. Sent on request rather than published, so it arrives with the context your team needs.
Everything an auditor asks for is already written down.
Nothing here is assembled after the fact. It is the by-product of making the work in the first place.
- ›A claims library beside the brand rules: one rule per approved claim, carrying the approved wording verbatim, the approver role, the approval date, the review date, the jurisdiction and the variants that must never appear.
- ›Claims Control in place of Brand Evolution: the same inbox, review and release loop, run as change control in the vocabulary your quality system already uses.
- ›Two owners, two lanes: your compliance owner rules on claims, your brand lead rules on colour and type. Neither can approve the other’s files.
- ›A fourteenth check at every release: claims complete and in date, and no forbidden wording on any rendered surface. It runs in seconds and it cannot be skipped.
- ›Traceability you can hand to an auditor: who ruled, what, when, in a numbered decision log, the rule’s own history, the approval record and the release tag.
Why claims, not tone, are the real AI risk in regulated marketing, and how approving once and enforcing everywhere shortens the review.
The product this page extends: your own agentic AI studio, four weeks to build, yours forever.
Field Notes: three days running this loop on our own brand, including the part that broke.
Regulated teams, asked straight.
Does this replace our regulatory review?
No. It makes the review count once. Your regulatory or compliance owner approves a claim as a rule: the wording, the jurisdiction, the review date. From then on every asset the system renders can only use that wording, and a check fails the release if a forbidden variant slips in or a claim goes past its review date.
Where do the approved claims live?
In the same rule store as your colours and type, in a repository you own. One rule per claim, with the approver role, the approval date, the review date and the jurisdiction on it. Changing a claim is a new entry in its history, never an edit of the old one.
What happens when a claim reaches its review date?
The release check fails until the claim is re-approved with a new date or retired. Nothing renders with an out-of-date claim, and nobody has to remember the date: the system does.
Who approves a change, and how is it recorded?
Changes arrive as a request on a branch. The files touched decide who must approve: your compliance owner for claims, your brand lead for colour and type. The approver reads the rendered page or PDF, not a diff, and approves under their own login. The approval keeps identity and time and cannot be edited afterwards; the release is tagged with a version.
Does our reviewer have to learn a tool?
No. They get an email with one link, read the piece as a page, and choose one of three buttons: approve, ask for a change in their own words, or comment. They never open a terminal, a repository or a design tool. Signing in once with their own account is the whole setup.
What happens when the reviewer asks for a change?
The system revises the piece from what they wrote, re-renders what they read, runs the checks again and asks them once more, usually within minutes. The exchange stays on the record beside the piece, so the reason for a wording change is readable a year later.
Where does it run, and what leaves our organisation?
On your own accounts. The library, the record and the approvals sit in your own repository, under your own logins and retention rules; the model your team builds with is your own account with a vendor you have already assessed, in the region you choose. After handover nothing routes through Catalys AI. Ask us for the architecture and data-flow note.
Are you SOC 2 or ISO 27001 certified?
No, and we do not claim to be. We install on your infrastructure and introduce no platform of our own, so there is nothing new for your security team to certify. The providers you already use publish their own certifications and reports, and your team obtains the current ones from them directly. We complete your third-party due diligence for the engagement itself.
Can it sit alongside our document control system?
Yes. The rule store is the source for rendered marketing assets; your controlled-document system stays the record for the dossier. The release tag and the decision log give you the reference to cite in it. We map the two in the first week.