The connective tissue: Why AI needs people

Article appeared in the Courier Mail on on 8 September 2026.

The biggest mistake many organisations make with AI is aiming their programmes at the wrong target. The instinct is to find a discrete task, automate it and bank the savings. It makes for a tidy business case, but it rarely reflects the way work actually gets done.

The problem is that productivity doesn’t come from making individual tasks faster. It comes from improving the flow of work across people, systems and decisions from beginning to end. Speed up one step without changing everything around it and you’ve simply moved the bottleneck elsewhere. In fact, sometimes you’ve made the overall process less efficient by flooding downstream teams with work they still can’t process any faster.

Many organisations are asking the wrong question. Rather than asking, ‘Which tasks can we automate?’ they should be asking, ‘How should this process work end to end, and where does AI genuinely improve it?’.

It also helps to be honest about what today’s AI is. It is remarkably capable within a narrow boundary, but far less capable outside it. It often can’t distinguish between an answer that is correct and one that is simply convincing, because it lacks the context to make that judgement. That is not a flaw waiting to be patched in the next model - for now it is a defining characteristic, and it is the whole reason the workflow question matters.

AI is enormously useful when someone who understands the work sits alongside it, but left to run a process on its own, it can produce outputs that are locally right, but globally wrong.

That’s why the role of people changes rather than disappears.

We’re often told that AI removes routine work so people can focus on ‘higher-value, creative work’. It sounds great in theory, but I think it misses the point. The value a person brings to an AI-driven process is not that they are more creative than the machine, but that they possess the context the machine cannot see. For example, what ‘good’ means in a specific setting, which edge case quietly breaks the process three steps downstream, what a regulator will question as output. This context is held in the knowledge and subject matter expertise which people have, and it is the connective tissue that lets one step in a flow actually connect to the next.

At a recent lunch, one attendee spoke about the growing importance of Business Analysts: people who understand not just the process itself, but the intended outcome. I think that’s exactly right. Organisations will increasingly need people who can connect business objectives, technology and operational reality.

This has important consequences for AI investment. If organisations assume their experienced operators are no longer needed because AI is handling the work, they risk removing the very people who understand how the process functions. The result isn’t better automation. It’s faster failure at scale.

The organisations seeing the greatest productivity gains from AI aren’t simply automating more tasks. They’re redesigning how work flows.

You can’t bolt AI onto a process designed around human strengths and weaknesses and expect transformational results. The work is to redesign the flow itself, identifying which steps should merge, which should disappear and where people should remain responsible for judgement and oversight. Real value comes from understanding where this judgment now needs to sit, as it may be at points that did not exist before.

That’s why organisations should invest in the connective tissue - the people who understand how work really gets done. It’s their knowledge that makes AI-driven processes both safe and effective. They are an asset to build on, not a cost the technology lets you remove.

Ultimately, the decision isn’t how much of the organisation to automate. It’s whether to treat AI as task-level tools bolted onto today's processes, or as an opportunity to redesign how work flows end to end.

The first approach is easier to justify and easier to measure, but it will almost always underdeliver. The second is more difficult, but it’s where the real value lies.

Before investing in the next AI tool, ask two questions: Which end-to-end process are we trying to improve, and who holds the knowledge that makes that process work? Get those right and AI becomes genuinely transformative. Get them wrong and you’ve simply made the wrong things happen faster.