Let’s be honest about what’s underneath AI resistance.
It’s not technophobia. Most founders who’ve built software businesses are not afraid of new technology. It’s something more specific, and actually quite reasonable.
You’ve seen the hype cycles before. You’ve watched tools get adopted across your industry with breathless enthusiasm, only to discover that the ROI was questionable, the implementation was painful, and the main beneficiaries were the vendors. You’ve built a business on judgment, not trends. You’ve learned to be sceptical of things that promise to change everything.
And now you’re watching every conference, every newsletter, every LinkedIn feed fill up with AI content, most of it superficial, much of it plainly wrong, and something in you is saying: wait.
That instinct is not wrong. But it may be leading you to the wrong conclusion.
What You’re Actually Resisting
Tim Barker was honest with the Business of Software Europe 2026 audience about his own version of AI scepticism. Not about whether AI is real, but about the quality of what’s being said about it.
“When you spend your time online, there’s a lot of people talking about AI, but how you implement it in an organisation is a bit hard to find. You’ve got so many people that are just giving you magic prompts that you should be using. It’s not really the answer.”
He’s right. Most AI content is either breathless promotion or useless generality. The 30-prompt lists. The “AI will replace your whole team” takes. The case studies that leave out everything that went wrong. The consultants who’ve never built anything with it telling you how to transform with it.
Resisting that is rational. The mistake is when the resistance extends beyond the noise to the underlying reality.
Because the underlying reality, built by practitioners actually running businesses on this, is considerably less hype-dependent and considerably more durable than the content ecosystem around it suggests.
The Legitimate Concerns, Addressed
Concern #1: “What if it goes wrong and damages our brand or our customer relationships?”
This is the right question, and the answer isn’t “don’t use AI.” It’s “build guardians.”
Tim runs four AI guardian agents whose sole job is to check the work of his other agents before anything goes external. Brand and messaging consistency. Trust implications. Information security. Financial commitments. Nothing reaches a customer or a prospect without passing through at least one check that isn’t the same agent that produced it.
“Move at the speed of trust,” he told the BoS Europe audience. “I’m not at a point where I want agents interacting with prospects and customers directly, until I’ve got their tone of voice to a point where I’m happy with it.”
The answer to “what if it goes wrong” is governance, built in from day one, not bolted on after something breaks.
Concern #2: “What if we become dependent on a vendor and they change pricing or go down?”
This one came up directly from the floor at BoS Europe. An audience member asked Tim: does your business grind to a halt when Anthropic’s servers are down?
“Anthropic or ChatGPT down is the new ‘internet’s down,'” he said. “Nothing happens when they’re down. We’re just reviewing specs that AI produced for us.”
On pricing: Claude’s tokens cost more from 1pm onwards, when the US market is awake. Tim’s adaptation? Schedule the heavy compute work for mornings. The economic model is changing. Surge pricing is coming, just as it did with cloud infrastructure before it. The response is operational adjustment, not abstention.
The organisations that avoided cloud infrastructure because of vendor dependency concerns aren’t looking clever today.
Concern #3: “We don’t have the engineering capacity to build AI systems properly.”
Tim’s response to this one is practical: you probably don’t need as much engineering as you think, and the AI tools available to non-engineers have changed the calculus significantly.
He used Google Stitch, which he calls “the Figma of AI,” to design an app from a written specification, then fed that into Claude Code to build it. A few hours of work, no dedicated engineering resource.
The barrier to building AI-augmented systems is lower than it’s been at any point in the history of building software. That cuts both ways: it lowers the barrier for you, and it lowers the barrier for everyone competing with you.
The Risk You’re Not Accounting For
Founders who resist AI tend to be running a specific mental calculation: the risk of getting AI wrong versus the benefit of getting it right.
What that calculation systematically underweights is the risk of standing still.
“The cost of production is now collapsing relatively compared to how it used to be,” Tim said. “The bar is continually getting reset on how you raise and what progress you need to have.”
This is the part that’s hardest to see from the outside. Seed-stage businesses are now arriving at the milestones that used to require a Series B, because AI has removed the engineering headcount that used to be the rate-limiting step. The companies doing this aren’t announcing it loudly. They’re just shipping.
If you’re in a market where a competitor figures out a 3-5x productivity multiplier before you do, the compounding effect of that advantage over 18-24 months is not recoverable with a catch-up programme. You can’t retro-compound a feedback loop.
The Difference Between Caution and Paralysis
Tim is not reckless about AI. He lost $2,000 because he let an agent operate without the right financial constraints. He talks openly about the mistakes he made and the governance gaps he’s had to close.
He does not advocate for moving fast and breaking things. He advocates for moving at the speed of trust, which means starting, building carefully, establishing what the system can do reliably, and expanding autonomy from there.
“You still need governance structures. But once you’ve got that governance in place, over time, you can apply different levels of depth of governance in some areas where you can autopilot, and others where you want to keep very close to it.”
That’s not an argument for abandoning caution. It’s an argument for applying caution in a way that still produces movement, rather than using caution as a reason not to start.
The founders who will have the worst of both worlds are the ones who wait until the risk of not acting becomes undeniable, and then try to implement fast, without the governance infrastructure, making the exact expensive mistakes that careful early movers already worked through.
The Right Question Isn’t “Should We Use AI?”
It’s “what’s our governance model for how we use it?”
That’s the question worth spending time on. And that’s the conversation the Build Your AI Company Operating System workshop is designed to have starting 20 May, with practitioners who’ve built the guardrails, made the mistakes, and can show you what a trustworthy AI operating model actually looks like.
Tim Barker spoke at Business of Software Europe 2026. He is the founder of Attain IP, a legal tech business applying AI to patents and intellectual property. He previously served as CMO at Salesforce Europe and as CEO of a 600-person digital healthcare company.