Intro
AI has made it trivial to look like a capable marketing agency. Anyone can wrap ChatGPT in a logo and sell "AI-powered content." The questions that follow are designed to tell the difference between an agency with a real system — AI for volume, humans for judgment — and one forwarding your brief to a chatbot. Ask them in a scoping call. A good agency will have crisp answers; a reseller will get vague.
1. AI + human workflow (6)
The core question: where does the AI stop and a human start?
Which specific AI tools do you use, and for what? Good answer: named tools mapped to tasks (e.g., enrichment, drafting, QA). Red flag: "proprietary AI" with no specifics.
What does a human review before anything ships? Good: a defined review gate on every client-facing asset. Red flag: "the AI handles it end to end."
Who is the senior person accountable for my account? Good: one named lead. Red flag: a rotating pool or "the platform."
Show me an asset the AI got wrong and a human caught. Good: a real example — proves the review gate exists. Red flag: "that doesn't happen."
What percentage of the work is AI-generated vs human-made? Good: an honest split with reasoning. Red flag: "100% AI" or "0% AI" — both are usually false.
How do you keep AI output on-brand? Good: brand inputs, examples, and human editing. Red flag: "we just prompt it with your website."
2. Data, privacy & compliance (5)
Who owns the data and content you produce for me? Good: you do, in writing. Red flag: hedging.
Is my data used to train any model? Good: a clear no, or explicit opt-out. Red flag: "it's all anonymized" hand-waving.
Can you work within my industry's regulations? Good: specific frameworks named. Red flag: "regulations aren't really our thing."
For regulated work, how do you prevent non-compliant content? Good: a review step against the actual rules. Red flag: "we're careful."
Where is my data stored and who can access it? Good: a straight answer. Red flag: "in the cloud."
3. Outcomes & accountability (6)
What outcome are you accountable for — and how is it measured? Good: pipeline, qualified leads, booked meetings. Red flag: impressions and "engagement."
What happens if you miss the target? Good: a defined remedy. Red flag: "marketing takes time."
Can I see a case study with real numbers? Good: specific metrics and context. Red flag: logos with no data.
What's your reporting cadence and format? Good: a regular, exportable report. Red flag: "we'll keep you posted."
How do you attribute results? Good: a clear, if imperfect, model. Red flag: taking credit for everything.
What's a realistic timeline to first results? Good: an honest range. Red flag: "instant results."
4. Pricing & terms (5)
How do you price — retainer, project, or outcome? Good: a model that matches the work. Red flag: opaque "it depends."
What's included and what costs extra? Good: a clear scope. Red flag: surprises later.
What's the minimum commitment? Good: a reasonable term tied to when results compound. Red flag: a long lock-in with no off-ramp.
What does it cost to leave — and do I keep the assets? Good: clean exit, you keep everything. Red flag: hostage data.
Are there per-seat or per-tool fees passed through? Good: transparency. Red flag: hidden markups.
5. The team (5)
Who actually does the work — employees, contractors, or AI alone? Good: a real team you can meet. Red flag: a black box.
What's the senior-to-junior ratio on my account? Good: senior oversight on strategy. Red flag: all juniors + a tool.
Can I meet the lead before signing? Good: yes. Red flag: "after onboarding."
What's your experience in my industry? Good: relevant proof. Red flag: "we work with everyone."
How do you stay current as AI tools change monthly? Good: a real process. Red flag: blank stare.
6. Tools & process (5)
Walk me through your workflow from brief to published asset. Good: a clear pipeline with a human gate. Red flag: "we just use AI."
How do you handle revisions? Good: a defined loop. Red flag: "the AI nails it first time."
What's your QA process? Good: checks before anything ships. Red flag: none.
How do you integrate with my existing stack? Good: named integrations. Red flag: "we'll figure it out."
What do you NOT use AI for? Good: a thoughtful answer (strategy, judgment, relationships). Red flag: "everything's AI."
7. Industry & specialization (5)
What kind of client gets the best results from you? Good: a specific ICP. Red flag: "anyone."
What's an engagement you turned down, and why? Good: shows standards. Red flag: "we never say no."
Do you specialize in any vertical? Good: a real niche (e.g., cannabis/regulated, B2B SaaS). Red flag: generalist with no edge.
How do you handle a category most platforms restrict (e.g., cannabis)? Good: specific compliance capability. Red flag: "we'd avoid that."
Why you over hiring in-house? Good: an honest cost/output comparison. Red flag: trash-talking in-house with no math.
Common questions
Ask what a human reviews before anything ships. The answer reveals whether there's a real quality gate or whether AI output goes straight to your audience unchecked.
Ask them to name the specific tools they use and what a human changes about the AI's output. Resellers describe a 'proprietary AI' and can't show where human judgment enters.
No. The strongest agencies use AI for volume — drafting, enrichment, sequencing — and keep humans on strategy, brand judgment, and final review. '100% AI' is a red flag, not a feature.
You should, in writing. Get data ownership and content rights confirmed in the contract, along with confirmation your data isn't used to train third-party models.
Expect a realistic range — typically weeks for outreach and content output, longer for compounding pipeline and AI-search visibility. Any promise of instant results is a warning sign.
Amir G'Nia
© 2026 Ad-Wise
AI at the volume. Humans on strategy.
