I've been thinking a lot about AI recently.
Not so much about whether it's coming to supply chain.
That's already happening.
I'm more interested in what happens when it gets here.
Because for all the conversation about artificial intelligence, automation and autonomous planning, there's a much simpler question sitting underneath it all.
What happens to the planner?
And more specifically:
What is a demand planner actually for?
It's a question I've found surprisingly difficult to answer.
One of the books that shaped how I thought about supply chain was The Machine That Changed the World.
I read it while studying supply chain at university.
Toyota.
Lean manufacturing.
Just-in-Time.
Pull rather than push.
Kanban cards moving through a factory, signalling that something had been consumed and needed replenishing.
The simplicity of it fascinated me.
Here was an incredibly sophisticated manufacturing philosophy built around some remarkably simple ideas.
Of course, planning technology didn't stop there.
MRP.
MRP II.
ERP.
Advanced planning systems.
Statistical forecasting.
Machine learning.
And now...
AI.
Each generation has given planners better tools.
More data.
More computing power.
Better forecasts.
Faster calculations.
More automation.
But AI feels slightly different.
Because we're no longer talking only about technology helping us calculate.
We're starting to talk about technology helping us interpret.
And increasingly...
decide.
Look at what some of today's planning technology can already do.
AI can identify patterns in demand.
Select forecasting models.
Analyse exceptions.
Help explain what has changed.
Generate scenarios.
Recommend actions.
Some platforms are already moving towards AI agents capable of taking action within predefined rules.
We're still a long way from every supply chain running autonomously.
But the direction is difficult to ignore.
Which brings me back to the question.
What is the planner for?
There is an obvious answer.
Judgement.
The machine produces the forecast.
The planner applies experience.
That sounds reassuring.
There's just one problem.
Humans aren't always very good at it.
Research into judgemental forecast adjustments has repeatedly shown that planners can improve statistical forecasts...
…but they can also make them worse.
Think about how easily it happens.
"Sales think this will be higher."
"This customer always orders more."
"Last month was unusual."
"I just don't believe that number."
Sometimes there's valuable information behind those statements.
Sometimes there isn't.
And that's where I think AI could force demand planning to confront something uncomfortable.
Perhaps the future planner isn't someone who constantly overrides the machine.
Perhaps they're someone who knows when not to.
Imagine the forecast says 12,450 units.
The planner reviews it.
Checks the assumptions.
Looks at what's happening with customers.
Understands what Sales knows.
Considers whether something genuinely structural has changed.
And then...
Does nothing.
No override.
No adjustment.
No intervention simply to demonstrate that the forecast has been "reviewed".
That might actually be excellent planning.
The value wasn't in changing the number.
The value was in knowing that it didn't need changing.
Now there's another part of this that I find fascinating.
Because AI isn't going to arrive equally everywhere.
Imagine two demand planners a few years from now.
Same job title.
Completely different working day.
One works for a huge multinational.
They arrive in the morning and their planning system has already processed thousands of demand signals.
Forecasts have been updated.
Exceptions identified.
Supply constraints considered.
Scenarios generated.
Potential actions recommended.
Their screen effectively says:
Here are the five things that need your attention today.
Now imagine another planner working for a smaller manufacturer.
Export the sales history.
Open Excel.
Copy.
Paste.
Update formulas.
Check the forecast.
Email Sales.
Update the production plan.
Same profession.
Completely different job.
And I think the gap between those two worlds could become enormous.
Except...
I'm not completely convinced it will.
Because something else is happening.
The planner in the smaller business now has access to technology that would have been almost unimaginable a few years ago.
They can give an AI tool a dataset and ask questions about it.
Build formulas.
Analyse patterns.
Find anomalies.
Write queries.
Explore forecasting approaches.
Create scenarios.
Automate repetitive work.
Things that could once have required specialist analysts, developers or expensive software are becoming accessible to someone sitting at a laptop.
So perhaps AI widens the gap.
Perhaps it closes it.
Maybe it does both.
Large organisations have enormous advantages.
Data.
Infrastructure.
Integrated systems.
Investment.
Specialist teams.
But smaller businesses have something powerful too.
Speed.
They can experiment.
They can change processes.
They can sometimes adopt new ways of working without navigating years of transformation programmes.
I genuinely don't know which advantage will prove more important.
And that's where I keep coming back to Toyota.
Kanban wasn't powerful because someone invented a clever card.
The card worked because of the operating philosophy surrounding it.
The process.
The discipline.
The thinking.
AI won't be any different.
You can put incredibly sophisticated technology on top of poor master data, weak processes, constant forecast overrides and planning meetings where nobody actually makes a decision.
You'll certainly have more technology.
I'm not convinced you'll have better planning.
AI can analyse bad data remarkably quickly. It's still bad data.
You might simply have found a much faster way of doing the wrong things.
Maybe, then, we're asking the wrong question.
Perhaps it isn't:
Will AI replace demand planners?
I'm increasingly bored by that question.
The more interesting one is:
What should demand planners stop doing?
Because if AI can take away hours spent collecting data, cleaning spreadsheets, producing reports, identifying exceptions and preparing analysis...
What do we do with the time it gives back?
That's the opportunity.
Less time producing the forecast.
More time understanding it.
Less time reporting what happened.
More time exploring what might happen next.
Less time touching every SKU.
More time understanding which ones actually require attention.
Less time preparing meetings.
More time helping the business make decisions.
That doesn't sound like the end of demand planning to me.
It sounds like the job getting closer to what it was supposed to be in the first place.
The tools of planning have been changing for decades.
Cards became calculations.
Calculations became systems.
Systems became algorithms.
Algorithms are becoming intelligent.
And perhaps eventually some of those systems will make decisions without us.
But underneath all of that technology sits the same problem planners have always been trying to solve.
The future is uncertain.
Someone still has to help the business prepare for it.
Maybe that's what a demand planner is for.
Theme
Evolution
Signal
The tools of planning have changed for decades. The responsibility of the planner hasn't.
Reflection
AI will undoubtedly change demand planning.
Some of the work planners do today will disappear.
Some will become automated.
Some will become dramatically easier.
And new responsibilities will emerge that we probably haven't even imagined yet.
But perhaps the biggest opportunity isn't better forecasting.
It's giving planners the time and capability to become better decision-makers.
The future of planning isn't simply human or machine.
It's understanding what each should be trusted to do.
Question
If AI removed half of the work from your planning role tomorrow...
What would you do with the time it gave back?
The Demand Signal
Lessons from the front line of supply chain.