Find where AI enters the work
Ask the agency to name each stage where it uses AI, the kind of tool involved, and what comes out of that stage. Research summaries, mood boards, mockups, copy, and final artwork carry different consequences. A broad claim that the agency is “AI-powered” tells you little about who shaped the result.
Then ask how audience research and positioning become a design brief before concepts are judged. If an early image board is involved, our guide to turning an AI mood board into a branding brief shows the questions that should survive the transition from inspiration to requirements.
Set boundaries for your inputs
Find out what client information, unpublished plans, reference files, and customer material could enter a tool. Ask who has access, which settings or retention terms apply, and whether your project can use a restricted or no-AI route. Have the agency explain exceptions before the brief contains sensitive material, and record the agreed boundary in the scope.
Keep creative judgment visible
Ask who decides which direction moves forward, who checks for lookalike work and visual mistakes, and how feedback changes the design. A polished generated concept still needs to answer the brief and work at the sizes and places your brand will use it. Request a sample review path, including who can reject or redraw an idea.
A fast image is only as useful as the decision behind it.
Make the answers comparable
Send the same brief and ten questions to each shortlisted agency. Ask for one representative project walkthrough and the evidence behind each answer. A useful response identifies a stage, a responsible person, a review decision, and a deliverable; where rights or sensitive claims matter, confirm the answer with the appropriate specialist.
| Decision area | Questions to ask | What a useful answer names |
|---|---|---|
| Use and inputs | 1. Where do you use AI? 2. What client inputs reach each tool? | Stages, tool types, inputs, and the boundary between exploration and final work. |
| Data and limits | 3. Who can see, retain, or reuse those inputs? 4. Can you support a restricted or no-AI route? | Access, applicable settings and terms, exceptions, and a workable alternative. |
| Judgment and quality | 5. Who chooses and approves the direction? 6. How do you check originality and errors? | Named people, research, similarity review, corrections, and approval gates. |
| Disclosure and claims | 7. When do you disclose AI involvement? 8. How do you verify factual and visual claims? | Agreed disclosure points, source checks, and the owner of final review. |
| Handoff and change | 9. What files, rights, licenses, and records do we receive? 10. What if a tool, term, or license changes? | Written deliverables, permissions to confirm, editable files, and a change path. |
Compare the answers with finished work, not just a presentation about process. The Hoss Straps project shows an identity used across print and products. Ask a prospective agency to trace a similar application from the brief through approval, source files, and final production requirements; a portfolio image alone cannot explain those decisions.
Inspect the handoff
Before signing, confirm the approved logo versions, editable masters, usage guidance, formats, licenses, and production files in the scope. Ask what records document AI involvement and who handles a correction after delivery. NIST’s AI Risk Management Framework offers a broader way to think about assigned responsibility and ongoing review; it is not a substitute for your project agreement.
When to bring us in
Bring us in when your identity needs a clear brief, accountable review, and files that work beyond the concept deck. Left Hand Design can scope those decisions and deliverables through our logo design and brand identity services.
