Promethean for AdTech & Marketing Automation

AI ad-copy generation

Drafter producing ad creative variants per campaign. Schema constraint on prohibited terms (medical, financial); reviewer-gate on schema violations; chain shows generation-rate + reviewer engagement.

Audience-targeting classification

Classifier assigning users to behavioural segments. Closed-enum outputs; per-decision consent-state recorded; chain provides evidence for ePrivacy/GDPR audit.

Marketing chatbot

Drafter producing conversational responses. AI Act Art. 50 disclosure built into the spec; reviewer-gate on schema violations + sensitive-topic flags.

Email personalisation engine

Drafter + classifier combination personalising email content per recipient. Schema-bound output; deterministic fallback to safe-default content when model errors.

AdTech volumes can be hundreds of millions of decisions monthly. Does Promethean scale?

Yes, but not for every micro-decision. Enterprise tier carries a 10M entries/month fair-use allowance — that's a billing envelope, not measured peak throughput; production scaling is honest at Production tier today and would warrant onboarding-assist conversation at the 10M end. For micro-decisioning (ad-impression scoring), you'd typically not record every impression — you'd record the rule-application decisions (which audience-segment classifier ran, which targeting policy applied) at a higher granularity. Per-impression bidding doesn't typically need substrate-level audit; per-customer-decisioning does.

AI Act Art. 50 disclosure on chatbots: does Promethean enforce it?

Promethean enforces it via the spec. The promptTemplate in a chatbot spec can require the disclosure in the system prompt; the spec hash binds that requirement to every chat-response L12 entry. If you change the spec to remove the disclosure, the spec hash changes and the chain shows it. A regulator inspecting your chain can see exactly which version of the disclosure language was active across what time window.

DSA transparency for adtech vendors serving VLOPs — what's our exposure?

VLOPs (Very Large Online Platforms) are directly regulated under DSA Art. 26 + Art. 39 for advertising transparency. As a vendor serving VLOPs, you'll be asked for the data your customer needs to publish in their ad-repository. Promethean's chain provides per-decision evidence (which ad-targeting model ran, what targeting criteria applied, who saw the ad in aggregate) that flows into the VLOP's transparency reporting. You're not directly regulated by DSA unless you become a VLOP yourself, but your customers will demand evidence.

GDPR Art. 22 — does personalisation trigger it?

Depends. Pure content recommendation (which articles to show, which products to feature) is typically not legal-effects-bearing; the personalisation is presentation. But personalised pricing, personalised access to services, personalised credit-application gates — those DO trigger Art. 22. Post-Schufa, even scoring-style processing can fall under Art. 22 if it materially influences downstream decisions. Promethean's chain provides the per-decision evidence; your compliance team assesses whether Art. 22 applies per spec.

Most adtech vendors are using GenAI for creative production now. Does that change the regulatory surface?

Yes. AI Act Art. 50 applies — synthetic ad content must be machine-readably marked as AI-generated; users should be able to identify it. The substrate records each generation event with the spec + model identity; pair with your C2PA / similar marking pipeline to satisfy the machine-readability requirement. National consumer-protection authorities are increasingly auditing AI-generated advertising claims (especially in financial + medical sectors); the L12 chain is the audit-trail layer underneath.