For decades, one of the promises of technology has been remarkably simple: make work faster and, for the most part, it has.
Paper forms became web forms, filing cabinets became databases and spreadsheets became platforms. Manual processes became workflows and information that once took days to find can now be retrieved in seconds.
These changes have created enormous value… but there is a problem.
We changed the tools far more often than we changed the work.
Consider a typical approval process:
Twenty years ago, someone might have completed a form, passed it to their manager, sent it to finance for approval and then waited for someone else to process it.
Today, that same process might happen entirely inside a digital platform.
The form is online, notifications are automatic and approvals happen with a click. Everyone can see the status in real time.
It is undoubtedly faster but underneath the technology, the process may be almost identical.
Request → Manager approval → Finance approval → Processing
We digitised the process but we didn’t necessarily ask whether the process still made sense. This pattern extends far beyond approvals though.
Reports became dashboards, meetings moved online, spreadsheets became applications and manual handovers became system integrations.
Again and again, technology has helped us perform existing work more efficiently. But… efficiency and effectiveness aren’t the same thing.
Making something easier to do doesn’t necessarily mean it needs to be done.
The processes we inherit
The problem is that most unnecessary work didn’t start out being unnecessary.
Processes usually exist for a reason - an additional approval might have been introduced because something went wrong, a monthly report might exist because, at some point, senior leaders couldn’t easily access the information themselves. A spreadsheet might have been created because two systems couldn’t talk to each other.
At the time, each of those decisions probably made sense… but organisations change.
People change, systems are replaced, teams are restructured and the original problem sometimes disappears completely - the process itself often doesn’t.
Over time, another approval gets added, another report is requested, another meeting appears in the calendar and another spreadsheet is created to fill the gap between two systems.
Eventually, nobody is entirely sure why some of it exists - we just know that it does.
Organisations are very good at creating processes. We’re considerably less good at getting rid of them.
This isn’t necessarily because anyone is doing anything wrong either - removing something carries risk and keeping it usually does not.
Nobody gets questioned because they continued producing the monthly report - stop producing it and suddenly someone might want to know why.
So the work survives and this is where AI creates an interesting problem.
For years, there was a natural barrier to automation: it was difficult and often expensive.
Automating a process might require a new system, development work, integrations, consultants or a lengthy business case. That meant there had to be enough value in automating something to justify the investment.
But that barrier is disappearing - AI can already summarise documents, analyse information, generate reports, classify requests, draft responses and perform tasks that previously required significant amounts of human effort.
Increasingly, if we can describe a piece of work, there’s a reasonable chance we can automate at least some of it.
That sounds like an enormous opportunity… and it is, but it also makes it incredibly easy to ask the wrong question:
How can we use AI to automate this?
When perhaps we should be asking:
Why are we doing this in the first place?
Because if a report no longer needs to exist, producing it automatically isn’t transformation, if an approval adds no meaningful control, making it instantaneous isn’t transformation either.
We’ve simply made unnecessary work cheaper.
And AI could make us extraordinarily efficient at doing things we don’t need to do.
Start with the work, not the technology
Perhaps then, the biggest opportunity with AI isn’t automation at all - it’s giving us a reason to look again at how work actually happens.
If we’re going to spend time identifying processes that AI can automate, we should probably spend some of that time asking why those processes exist in the first place.
- What outcome is this trying to achieve?
- What would happen if we stopped doing it?
- Is it managing a genuine risk or simply following a process we’ve inherited?
- Could we achieve the same outcome in a completely different way?
Answering these naturally leads to the key question:
Could AI help us do it better?
This sounds obvious, but it represents quite a different way of thinking about technology.
For years, digital transformation has often started with the technology. A new platform arrives, a capability becomes available and organisations look for places to use it.
AI makes that temptation even greater because the technology can do so much but starting with the capability inevitably shapes the question.
If we start by asking “What can we automate?”, we’ll find things to automate.
If we start by asking “What are we actually trying to achieve?”, we might arrive somewhere completely different.
A monthly report might not need to become an AI-generated monthly report… perhaps people should simply be able to ask questions of the information when they need it.
An approval process might not need faster approvals… Perhaps the decision could be made by the person doing the work, with AI helping them understand the rules, risks and information they need.
A meeting that exists largely to exchange information might not need an AI-generated summary… Perhaps the meeting doesn’t need to happen.
That’s a much more interesting use of AI - but not because we’re using it to remove people from processes or squeeze another few percentage points of efficiency from the work we’re already doing, but because it gives us an opportunity to rethink some of the assumptions those processes were built around.
Technology has always been good at helping us do things faster - AI will make us faster still.
But speed isn’t particularly useful if we’re running in the wrong direction.
The biggest productivity gains from AI may not come from doing our existing work faster.
They may come from discovering how much of that work we no longer need to do.
