Don’t start your AI adoption with AI

I have been trying to make sense of what is actually happening with AI adoption in organisations.

Consultancies are training thousands of people to help clients implement it. Organisations are buying licences, running pilots and building agents. Some are mandating its use and measuring activity.

Much of this still appears to begin with the technology. Where can we use AI? What can we automate? Which workflows can we speed up?

Those are understandable questions, but they come too early.

The starting point should be the need we are trying to meet or the problem we are trying to solve. What does the customer, employee or citizen actually need? 

What would a materially better outcome look like? What currently prevents that outcome?

Only then should AI enter the conversation, as one of many possible capabilities.

This is why design thinking and systems thinking seem more important to me now, not less.

Design Thinking begins with empathy. It asks us to understand people’s needs, experiences and emotions before we design the solution. Systems thinking asks us to look beyond the visible workflow to the wider network of relationships, incentives, constraints and feedback effects in which the work sits.

The word “workflow” itself may be part of the problem. It suggests a flowchart: input, step, step, output. That makes it easy to imagine inserting AI into a box and declaring success.

But organisations are not collections of neat workflows. They are ecosystems. Change one part and the consequences appear elsewhere, sometimes in ways nobody anticipated.

An AI system may speed up one activity while increasing checking elsewhere. It may produce more reports, more options and more messages for other people to absorb. It may reduce cost locally while shifting risk, weakening judgment, or damaging the customer experience elsewhere.

I was reminded of a process redesign programme I managed in a retail bank some years ago. The substantial technology investment gave the work seriousness, resources and senior attention. Yet many of the benefits came from examining processes that had grown organically over time and no longer made sense.

That was valuable, but AI now gives us the opportunity to go further.

Sometimes the right answer will be to redesign a process. Sometimes it will be to remove it. Sometimes it will be to reconsider the whole proposition and find a fundamentally different way of meeting the need.

Amazon did not simply make the existing bookshop workflow more efficient. It helped create a different system for discovering, buying and distributing books.

The real opportunity with AI is not merely to automate what organisations already do. It is worth questioning whether they should be doing it that way at all.

Perhaps the first question in an AI initiative should not be, “Where can we use AI?”

It should be, “What human need are we trying to meet, and what is the best way to meet it now?”