Don't outsource your core competence to AI

I've had a couple of conversations with business leaders recently that left me uneasy.

One company had already started working with a vendor that has a product that promised to do big chunks of the core work of this organisation with AI.

The other company had been approached by an AI vendor with an offer: we'll build you the thing. Fast, cheap and all we ask is that we keep the IP in what we build.

One leader was treating it as an obvious win - the other was more dubious. I don't think it's a win.

I do understand the appeal. Everyone is racing to adopt AI, nobody wants to be the one who was left behind, and the board wants to know what you're doing about it. An offer like that solves your AI problem overnight, at little cost, and gives you something to say. Turning it down means going slower and spending more money - which is a difficult argument to win at the moment.

But it's worth going back to that old adage: don't outsource your core competence.

The two conversations were different situations, but they end up with the same problem - the loss of the ability to differentiate in the market.

In the situation where you are using an off-the-shelf tool to do what is your core competence, you get forced into selling what the product offers. And every product suggestion you make or change request you pay for helps someone else to get better at what used to be your secret sauce.

In the situation where a vendor offers to build the tool for you (and keep what they learn), they will learn your rules, your edge cases, your two decades of knowing which exceptions matter and which don't - encoded, owned by them, and sellable. At best, you end up reselling a generic version of what used to be the value you represented.

In either case, you stop being able to be different from everyone else in the market. You are now just a feeder into your supplier's roadmap and you are beholden to what they choose to add to their product. Ask yourself, after three years of delivering your core competence through someone else's product, can you take back control?

Which raises the obvious question: how do you tell what's core? I don't have a clear answer, but I think it's about understanding what makes you different from everyone else. It boils down to what your sales people say when asked why the customers should buy from you rather than your competitor.

None of this means doing everything yourself from first principles. Take us at NewOrbit Ltd: designing and writing software is part of our core competence. I could probably - soon - find an AI business that would take my requirements and hand back working code. I won't. But we use AI tools extensively. We write the standards and guidelines those tools work to, we design the processes, we choose which models do which work, and we make the judgement calls. That's the distinction I'd draw. The tools are tools. The control stays with us.

You absolutely should look at how you can use AI to improve your core business, that is not negotiable. It's about retaining control. My friend and business partner Sean Worthington has written a great article about it.

At which point there's a fair challenge coming: isn't hiring a company like mine outsourcing your core competence?

It's a reasonable question and I'd rather answer it than let it sit there. The systems we build are owned by our customers, entirely - the code is theirs, and the proprietary knowledge they share with us remains their property. We're not accumulating their differentiator into something we can sell on to the next client. We're helping them execute their own value proposition better. That's the line, and it's the same line I'd apply to any vendor, including an AI one: what do they walk away with, and what have I got left?

So this isn't really an argument about AI at all. It's the same question businesses have always had to answer, just arriving faster and dressed up more attractively than usual. What is it that makes you worth choosing - and who are you letting near it?

Footnote: Some business are in the situation where their core competence is already something that generic AI can just do. That's a different conversation and not one you should run away from.