Show Notes
On the 8th of October, Gareth and Callum went to The Robots Are Coming, the AI event hosted by Agency Hackers in London, and came away with far more than one episode could hold. In this episode they cut it down to the three takeaways that matter most for an independent agency: what AI does to the margin model and who ends up with the benefit, why AI transformation stalls without named ownership and carved-out time, and the data assets agencies are already sitting on and barely using. Credit to Ian Harris and the team for organising a great event.
In this episode
- Mossy O’Mahoney of Teamwork.com on how agencies are planning resource now – the next hire gets discussed as a role, then somebody asks whether it needs to be a person at all, or whether an agent sits in that slot alongside the team
- His view that 30% of junior work in agencies will end up being done by AI – Gareth thought that was low, and most of the clients he has put the figure to have come back with half or more. Plus what that opens up for the next generation of agency people, who learn the craft by doing the brunt work in the engine room
- David Barker at Inspire KBB on the client report that used to take an hour and now takes fifteen minutes – the AI produces it in three, and a human spends the rest of the time checking it and sending it. That gap is where the whole margin question starts
- The decision almost nobody is making consciously – bank the saved time as margin, hand it back to the client as a lower price, or spend it on controlled scope creep and give them more for the same fee. Gareth’s fear is that agencies are people-pleasers by nature and will give it away by default
- Why claiming that margin means a conversation most owners would rather avoid, and why hiding the saving is the worse option – although not as bad as simply not knowing. Measure it, and measure agents and people separately for now
- The ownership problem – the prodigi.team research that three quarters of agencies name lack of resource as a barrier to transformation, how AI ends up as a side job or a champions group, and why the momentum fades back into client delivery within weeks when nobody carves out the time
- Why one project finished properly beats five started – governed at the outset, owned through the build and again through the upgrades, then measured at the end so the effort turns into margin. What you deploy is not broken a few months later, it is just out of date
- The third takeaway, the data assets you already own and treat as information rather than assets – transcripts, timesheets, Slack, email, cred and pitch decks, policies, processes, frameworks, people and roles, this year’s goals. Context is the difference between AI that understands your business and AI that understands Reddit
- What that unlocks in practice – an early warning system on unhappy clients built from email and meeting sentiment, the upsell a client has raised in three meetings that nobody picked up, and quoting new work from what the last job actually took rather than a guess. Callum’s line to close: your people need food, water and sleep, and AI needs data
