I’m speaking at AI Evening in London on 4 September.
The event runs from 18:00 to 22:00 and brings together people from Canva, HeyGen, Lovable, Gamma, Fotor, buildher labs, GCreators.me and RakLaw. Comerix is one of the partners. The billing is inspiration, innovation and impact. My contribution will be slightly less comfortable than that sounds.
The pattern I keep seeing
I’ve spent 20 years building ecommerce businesses and the last year building Comerix AI. In that time I’ve sat down with a lot of companies who have adopted AI properly. AI in the CRM. AI in the helpdesk. AI in the content workflow. Output went up. Response times came down. Everyone is busier and producing more.
And almost nothing changed about what gets decided, or when, or by whom.
That’s the gap. Not capability. The models are good enough for most commercial decisions a mid-sized business makes in a week. The gap is that nobody has wired AI into the moment a decision actually gets made. So it sits in the reporting layer, summarising what already happened, while the person who needed the answer made the call on instinct 3 days earlier.
Why the gap exists
Three reasons, in my experience.
AI arrived through the tools layer. It was bought function by function, by the people who owned those functions, and each one optimised their own throughput. Nobody owned the decisions that cut across them.
Decisions don’t live in systems. They happen in a call, a Slack thread, a corridor conversation, a gut feeling at 11pm. Your systems find out afterwards, if at all.
The signal that would inform those decisions sits in the messiest places you have. Sales calls. Email threads. Support tickets. Meeting recordings. It’s the least structured data in your business and it’s where the actual intelligence is.
Recording is not deciding
This is the line I keep coming back to, and it’s the one I’ll build the talk around.
CRMs record what happened. The unsolved problem is deciding what should happen next.
Recording what happened is a solved problem and has been for 20 years. The unsolved problem is deciding what should happen next, at the moment it matters, using everything the business already knows but hasn’t organised.
A dashboard that tells you a deal went quiet in week 6 isn’t intelligence. Something that tells you in week 2 that this deal looks like the last 4 you lost, and names the reason, is.
What I’ll cover on the night
What it takes to move AI out of the dashboard and into the decision. Where teams get it wrong. What I’ve learned building Comerix, including the parts that didn’t work. And the uncomfortable question of which decisions you would actually be willing to hand over.
If you’re in London
Come and find me. I’d rather have 5 proper conversations than collect a stack of business cards.
And if you’re not going but you have a view: what’s the one decision in your business you’d hand to AI tomorrow if you trusted it enough? Send it over. I’ll work the best answers into the talk.