AI, jobs and the adoption gap

I have been thinking about two recent essays by David Oks — Why I’m not worried about AI job loss and Why ATMs didn’t kill bank teller jobs, but the iPhone did. They are useful because they look past the usual extremes i.e. AI changing nothing and AI wiping out work overnight.

Oks makes two arguments that I think business leaders should take seriously. First, labour substitution is harder than it looks. An AI model may perform a task well, but that does not mean a business can remove the human role around it. Real organisations run on approvals, legacy systems, local knowledge, customer expectations, exceptions, politics and trust. Those frictions slow replacement and slow change.

Second, the biggest shifts often come from new operating models, not direct task automation. His ATM and iPhone example makes the point well. The ATM automated part of the teller’s work, but branches and tellers still had a role and infact employment of tellers actually grew as they moved more into sales and relationship roles. Years later, mobile banking changed the whole model. Only then did the labour impact happen because the mobile app could increase convenience much further than the ATM but more importantly it could take on almost all the rest of the teller role.

That aligns closely to what I see in the market. The immediate issue for most organisations is not mass job loss. It is whether teams can adopt AI in a practical, governed and repeatable way. The gap between AI capability and business value is still very wide. The businesses that close it fastest will be in the strongest position - but it sure ain't easy and actually requires more people.

Businesses capability is needed

My view is that in this phase of AI, most organisations should focus less on dramatic replacement stories and more on capability building. AI value appears when teams can use it confidently in the flow of work, with the right context, controls and support. That is a much bigger management challenge than a technology one.

This is also why I think adoption matters so much. If human-plus-AI is the dominant model for now, then the near-term opportunity is to make people better, faster and more consistent with AI. Over time, that learning creates the foundation for bigger workflow redesigns - but it is currently unclear what those are.

How Theta Assist fits

We built Theta Assist for exactly this stage of the market. Most organisations don't need more AI hype. They need a practical way to help people adopt AI safely, share what works, and use it inside the tools and information they already rely on.

That is the role Theta Assist plays: turning scattered experimentation into structured adoption. If teams have secure access, shared assistants, business context and clear governance, AI becomes easier to trust and easier to use. That is how organisations step from curiosity to capability.

A real question for 2026

The most useful question right now is how an organisation can learn to use AI well at scale? The firms that answer that question first will get the strongest results, whether the next step is productivity gains, service improvements, new offerings or deeper operating model change.

That is the lens I would encourage leaders to use. Stay clear-eyed about the long-term shift. Stay practical about the short-term work. And focus on adoption, because that is what turns AI from a headline into a real advantage.

‍