Someone made the case in LinkedIn this week that AI is eating the agency deliverable, and they’re not wrong. Call me old fashioned (and I’m sure some of you will), but my response was simple – quality AI output is only as good as the prompts and rigour you put into it. Truly knowing your sh*t still wins.
AI is really good at producing a passable version of most deliverables (meeting minutes, copy, code, strategy frameworks, content, audits) in a fraction of the time. For us experienced practitioners, that’s genuinely useful. Things that used to take hours, take minutes. You can test more, iterate faster, cover more ground. But is “passable” good enough?
AI doesn’t know what it doesn’t know. It will produce confident, well-structured output that can be completely wrong. Only people who have ‘seen and done things’ for a while know enough to recognise when something sounds plausible but simply isn’t right. And when challenging an AI, they know when to keep pushing back even when it argues that it’s right.
If you’ve been hiding behind processes, templates, and hours, AI isn’t going to save you. It will expose you. You’re now producing low-value work faster that you’ve ever done before, that no-one will thank you for.
This is the judgement gap, and it’s the one thing that doesn’t (yet) show up in an AI playbook. Knowing your sh*t means:
- Knowing which questions to ask before the brief is written.
- Knowing when a technically correct answer is commercially useless.
- Knowing that the client’s real problem isn’t what they’ve described.
Knowing these things comes from experience, pattern recognition, failures and genuine domain knowledge. AI can’t replicate it, and the gap between practitioners who have it, and those who don’t, is about to get whole a lot more visible.
If you’re ‘the client,’ this matters. Agencies and consultants are increasingly producing AI-assisted work. Some of it will be excellent. Some of it will be polished, confident, and horribly wrong. If you don’t have the internal capability, or the governance framework, to evaluate what you’re being given, you won’t know the difference until it’s expensive. Ask Colgate.
Yes, AI has made it easier to produce work. But, it hasn’t made it easier to know whether the work is any good. That still requires a human. The question worth asking of your agency, your consultants, and honestly, yourself, is whether there is genuine hard-won judgment behind the output.
