Every AI output that ships from this team passes through the same four-step review loop. It's the difference between us and a prompt-based design service. Here it is, in plain English.
Step 1, Brand check
Open the brand kit. Compare the output side by side. Colours, type, voice, spacing system, photo treatment, iconography. If anything is off the kit, mark it. The first few times you do this it takes 10 minutes per output. After 50 outputs it takes 30 seconds.
Step 2, Brief check
Read the founder's actual brief one more time. Not the cleaned-up version the agent saw. The original. Did the output answer the question, or did it answer a slightly different question that was easier? AI is exceptionally good at answering adjacent questions. The senior designer's job is to notice when it has.
AI doesn't know the difference between the question you asked and the question that's easier to answer. The senior designer does.
Step 3, Craft check
Would I sign my name to this? If the answer is anything other than 'yes', iterate. AI outputs at 90% craft are common. AI outputs at 100% craft are rare without human refinement. The closing 10% is where the team earns its value.
Step 4, Trade-off check
The output is 100% on-brand and on-brief. What did we sacrifice? Maybe the layout is safer than it could be. Maybe the type is more conventional than the brand allows. Make a deliberate decision: ship the safe version, or push back through one more iteration to get the more interesting one. Both are valid. The point is to make the call consciously, not by default.
Four steps. Two minutes per output once you've internalised them. The team's signature is the review loop, not the agent.
