“If you're interested in an average response, then it's great because it's literally just a tool that is going to find the average best next word... But if you're using it for real work, that's terrible,” Joe says.
Joe and Jack run Forty Eight Point One, a brand and digital agency. Both are wary of AI as an oracle, and clear about where it earns its place.
One place it has: code review. Every pull request now runs through an AI check built on the agency's own standards. “There is a thorough check of every PR happening... That's never going back now,” Joe says.
The cost is starting to bite though. Token charges have joined the subscription fee, and the bill now rivals a salary. “The amount we're spending on tokens to run those sorts of things amounts to another team member,” Joe says.
Jack has resisted writing a formal AI policy. Early attempts stalled fast. “We were going to spend all of our time rewriting policies versus actually learning about this,” he says. He expects most agency AI policies will be written by AI, and forgotten within weeks.
His bigger worry sits with junior staff. Strip away critical thinking and the habit of wrestling with a problem, and something else is lost. “I actually think it will lead to more burnout, not less burnout,” Jack says.
Both keep coming back to the same point. Craft and human judgement decide whether the work is any good. Of creative work that starts and ends with a prompt, Joe puts it plainly: “That's not creative work, is it?”
Listen to the full conversation with Joe and Jack on Aperture: Agency AI.
