AI delivery
How do agencies QA AI-written client deliverables?
Review AI drafts against written client rules and an 8-check list: goal, facts, voice, claims, audience, format, links and approval. Send misses back with the rule that broke, then add that rule to the client file. Rewrites shrink because the AI gets the rule next time.
Why reviews turn into rewrites. Reviewers apply rules that only live in their heads. The AI never heard them, so the reviewer rewrites the draft. Net gain: 0.
The 8 checks.
- Goal: does it serve the brief?
- Facts: is every claim true?
- Voice: does it sound like the client?
- Claims: any banned or risky claims?
- Audience: is it for the right reader?
- Format: right length and layout?
- Links: do they work and go to the right place?
- Approval: who signs off?
Match review to risk. Spot check internal drafts. Run the full list on client work. Get client sign-off on pricing and legal claims.
Close the loop. Every miss becomes a new line in the client context file. That is the step most agencies skip.
FAQ
- Who should review AI work?
- The account owner, against the checklist. Escalate pricing and legal claims to the client.
- How long should an AI QA checklist be?
- Short. 8 checks gets used. A 30-point list gets skipped by Friday.
- Is there a free checklist?
- Yes. The AGL Agency Capacity connector returns the 8-check approval list, with extra checks by service line.