Agency Chiefs Want 'AI Upskilling.' What They Actually Need Is a Hard Reset on Basic Competence.

Here we go again: it’s autumn in New York, ad agency execs are back from their overpriced Hamptons rentals, and the trade press is lapping up another round of ‘what will you change this year?’ from the same warmed-over leadership deck. Digitas says they want AI upskilling. Partners + Napier wants more time for so-called ‘problem-solving.’ Pereira O’Dell and Walrus are all parroting the same future-forward, posturing nonsense. It’s a parade of LinkedIn-approved buzzwords with zero real consequence for clients or outcomes.

Let me translate: ‘AI upskilling’ means, ‘We’re reading ChatGPT prompts from a Notion doc and calling it innovation.’ The real problem nobody at these agencies wants to admit? Their teams couldn’t ship a functional landing page without a 16-step Figma review, a standup with five project managers, and three different plugins that break on Safari. The last time I saw a Digitas site, it scored a 42 on Lighthouse—so forgive me if I don’t buy the whole ‘AI-powered future’ narrative.

Here’s another classic from this crop: ‘We want to make more time for problem-solving.’ Translation: ‘We’re drowning in meetings about meetings, so maybe if we say we’re working on important stuff, no one will notice we haven’t actually shipped anything in months.’ Ask any engineer who’s had to bolt a new tracking pixel onto one of these agency projects: the real bottleneck isn’t lack of creative thinking, it’s a culture obsessed with process theater and CYA documentation.

You want to fix something? Stop letting your interns run LLM prompt engineering off a YouTube course and start with mandatory code reviews—real ones, not the ‘looks good to me’ Slack emoji. Fire half your project managers and hire two actual backend devs. And tell your clients the truth: AI can’t save you from your own inability to ship on time and on budget. That’s the uncomfortable recommendation: less training, more actual building. The agencies who do this might ship something that works by this time next year. The rest will still be stuck in the same status meeting, talking about what they’ll change ‘over the next 12 months.’