AI can make a bad process look like a great investment.
A workflow takes three days. AI reduces it to thirty seconds.
The productivity case looks obvious.
But speed tells us nothing about whether the work still makes sense.
That is the risk I see in many AI conversations. Companies start with the business as it exists today — current processes, roles, approvals and products — and ask where AI can fit.
I think that question comes too late.
Before deciding what AI should automate, improve or replace, leadership should challenge the assumptions that created the work in the first place.
The video episode is scheduled on YouTube. The player will become available when the episode is public. The YouTube player loads only after you choose to play it.
Watch on YouTube →
THE CONSTRAINT MAY HAVE DISAPPEARED. THE PROCESS STAYED.
Most processes were not designed badly.
They were designed around the reality of the time.
Information was difficult to access. Systems could not communicate. Decisions needed several people because no one had the full picture. Customisation was expensive because experts had to do the work manually.
Then technology changed.
The original constraint weakened or disappeared.
But the process often survived.
That is how a company ends up applying AI to a structure built around a problem that no longer exists.
A weekly information-sharing meeting is a simple example. Years ago it may have been the only practical way for Finance, Sales and Operations to assemble the full picture.
Today the information may already be available through shared systems, dashboards and AI.
Adding an AI meeting assistant might make the meeting more efficient.
The more important question is whether that meeting should still exist in the same form.
AI can make bad design extremely efficient.
That is exactly why the logic needs to be challenged before the work is changed.
FOUR QUESTIONS CHANGE THE CONVERSATION
I use four questions to separate what deserves improvement from what deserves rethinking.
1. What problem was this designed to solve?
Go back to the origin.
Why was the approval added? Why was that team created? Why does information move through those steps?
If nobody remembers, that is useful information in itself.
Understanding the original problem gives you something against which to judge the current design.
2. Does that problem still exist?
If it does, keep solving it.
Old does not automatically mean wrong.
But if the underlying constraint has disappeared, leadership should question whether the solution built around it still deserves to survive.
Improving something is not always progress.
Sometimes the right move is to remove it.
3. What is possible now?
This is where AI becomes more commercially interesting.
A lot of companies stop at: How much time can we save? How much work can we automate?
Those questions matter. But AI can also change the economics of what the company offers.
Custom work may become scalable.
A service delivered periodically might become continuous.
Expert knowledge that previously had to be delivered one client at a time could become accessible to a much larger market.
A customer segment that was previously too expensive to serve may suddenly become viable.
That changes the conversation from productivity to customer value, growth and new revenue.
4. If we designed it today, would we build it the same way?
This is the hardest one.
Ignore the current org chart for a moment.
Start with the customer outcome, the information available today and the capabilities AI now gives the business.
Would you create the same approvals?
Would work move through the same departments?
Would people and AI divide the work in the same way?
Would you even sell the same thing?
Then compare that design with the business that exists today.
The gap is where the real transformation conversation begins.
THIS IS BIGGER THAN PROCESS AUTOMATION
There are two directions this thinking can take.
One is Rethink the Work.
You discover that an existing process, role or workflow should be redesigned because the assumptions underneath it have changed.
The other is Rethink What You Can Sell.
You discover that AI has removed a constraint that previously limited what the company could offer, who it could serve or how it could create revenue.
That second opportunity is often overlooked.
The same technology that removes cost from today's business may make tomorrow's offer possible.
WHERE I WOULD START
Not with a list of fifty AI use cases.
Take one important part of the business that people understand well.
Put the current way of working in front of the team and run the four questions against it.
Then make a decision.
If the existing logic still makes sense, improve it.
If the logic no longer makes sense but the future model is unclear, rethink it.
If you already know what the new model should be, rebuild around it.
Only then decide where AI belongs.
That sequence becomes even more important as AI moves from assisting people to taking actions itself.
An AI agent can execute a bad rule perfectly.
Giving outdated logic more autonomy does not make it better.
THE QUESTION WORTH ASKING
“Where can we use AI?” will produce plenty of answers.
I think leadership should ask something harder:
What no longer makes sense because AI exists?
That question can lead somewhere very different.
Sometimes the answer is better technology.
Sometimes it is a redesigned process.
And sometimes the opportunity is much larger: a different way of creating value and revenue.
That is why I would challenge the logic before changing the work.