Ask a business leader where artificial intelligence could create value and one answer tends to appear quickly.
Efficiency.
Could AI automate this process?
Could it reduce the number of hours people spend doing that?
Could we handle more customer enquiries without adding employees?
Could we produce the same output with fewer resources?
These are sensible questions.
Cost reduction is tangible. If a process requires 1,000 hours of human effort and AI can reliably reduce that to 500, it’s possible to put a number against the improvement.
The business case can be relatively straightforward.
And given the capabilities now emerging, I suspect there will be significant opportunities to remove repetitive cognitive work from businesses.
But there is a danger in allowing efficiency to become synonymous with AI value.
Because the economic impact of a technology isn’t limited to making existing activities cheaper.
Sometimes the much bigger opportunity comes from making previously uneconomic activities possible.
Or helping people produce better outcomes.
Or allowing a company to serve customers differently.
Or increasing the capacity of a constrained part of the organisation.
Or enabling entirely new products and services.
If we only ask how AI can reduce the cost of the business we already have, we may miss what AI allows the business to become.
Cost savings are easier to see#
There is a reason efficiency dominates the conversation.
It’s measurable.
Imagine ten employees each spend five hours a week producing a particular type of report.
We can estimate the annual cost of that activity.
If AI reduces the work by 50%, we can estimate the theoretical saving.
Reality will usually be messier than the spreadsheet, but at least there is a visible starting point.
Now imagine AI helps the same employees produce better reports.
How much is “better” worth?
Perhaps decisions improve.
Perhaps customers receive more useful advice.
Perhaps employees identify opportunities they previously missed.
Perhaps nothing economically meaningful changes at all.
The causal chain becomes harder to prove.
This creates a bias.
Businesses can naturally gravitate towards the AI opportunities whose value is easiest to calculate rather than those capable of creating the most value.
Cost is visible.
Opportunity often isn’t.
That doesn’t make efficiency projects bad investments.
It means we should be careful not to confuse measurability with importance.
The same productivity gain can create different value#
Suppose AI allows a team to complete its work 20% faster.
What is that worth?
The obvious calculation is labour.
If the company can produce the same output with fewer hours, there is an efficiency benefit.
But that’s only one possible outcome.
Imagine instead that the team is a constraint on growth.
There is more demand for its work than it has capacity to handle.
Now completing work 20% faster doesn’t necessarily mean reducing the team.
It might mean serving 20% more customers.
The underlying technological improvement is identical.
The economic outcome is completely different.
In one business, productivity becomes cost reduction.
In another, it becomes additional capacity.
And if additional capacity can be sold profitably, the revenue and margin created could be substantially more valuable than the labour saving.
This is why businesses need to separate what AI improves from how the business captures the improvement.
AI might save an hour.
What happens to the hour is a business decision.
Productivity is not the value. It’s a capability the business has to convert into value.
That conversion matters.
Revenue deserves as much attention as cost#
Imagine a salesperson with 100 potential accounts.
Today, they can properly research ten.
AI makes that research dramatically faster, allowing them to understand 50.
We could measure the hours saved preparing for each account.
But that might be the least interesting number.
What if better research means the salesperson identifies stronger opportunities?
What if they enter conversations with more relevant information?
What if they personalise outreach more effectively?
What if the result is a small improvement in conversion?
Now AI is influencing revenue.
The same principle can apply across a business.
AI might help customers find the right product. It might help a company identify customers likely to leave, improve recommendations, make personalised communication economical, identify sales opportunities hidden inside existing information or allow a company to economically serve smaller accounts.
None of these automatically produces revenue.
Technology doesn’t get to skip the difficult work of implementation, customer behaviour and measurement.
But they illustrate why asking only “How many hours does this save?” can produce an incomplete view of the opportunity.
Sometimes the more important question is:
What becomes commercially possible because this now takes fewer hours?
Better decisions create value without removing work#
There is another category I’ve written about recently: decisions.
Suppose AI gives a manager better information before an important decision.
The decision still takes the same amount of time.
Nobody is removed from the process.
There may be no productivity saving whatsoever.
But if the decision improves, the economic impact can still be substantial.
A better pricing decision could improve margin.
A better purchasing decision could reduce waste.
A better prioritisation decision could direct sales effort towards stronger opportunities.
A better customer decision could prevent a valuable account from leaving.
A better operational decision could prevent an expensive problem.
The AI hasn’t made the organisation more efficient in the conventional sense.
It has helped it become more effective.
That distinction is important.
Efficiency asks:
How can we achieve the same outcome using fewer resources?
Effectiveness asks:
How can we achieve a better outcome with the resources we have?
AI can potentially influence both.
A business strategy focused entirely on the first is therefore looking at only part of the economic opportunity.
Scarce expertise creates another opportunity#
Some of the most valuable people inside a business are valuable precisely because their knowledge is scarce.
They know the product.
The customers.
The market.
The systems.
The exceptions.
The strange things that happen only occasionally but matter enormously when they do.
Other people depend on them. Questions accumulate. Work waits. Decisions get escalated.
The obvious AI interpretation is:
Can we automate the expert?
A more interesting question is often:
Can we increase the reach of the expert?
Imagine an experienced employee whose knowledge helps 20 colleagues today.
If AI can make some of that knowledge accessible to 200 people — while preserving appropriate human oversight — the economic effect isn’t necessarily reducing the cost of the expert.
The organisation has increased the availability of something previously scarce.
The expert might spend less time repeatedly answering routine questions and more time dealing with the genuinely difficult ones.
AI hasn’t eliminated expertise.
It has potentially increased its leverage.
This connects to a broader idea I’ve written about before.
Businesses contain valuable capabilities that are rationed because humans have finite time: analysis, research, personalisation, software development, specialist knowledge and attention.
AI may allow businesses to apply some of those capabilities much more widely.
And once something becomes less scarce, the opportunity isn’t always to spend less on it.
Sometimes it’s to use much more of it.
Customer experiences are shaped by economics#
Think about how many customer experiences exist in their current form because providing something better would cost too much.
Customers are grouped into segments because individual personalisation is expensive.
Smaller accounts receive less attention because human service is expensive.
Support operates through queues because immediate expertise is expensive.
Businesses send standardised communications because producing something individually relevant for every person is expensive.
These aren’t necessarily bad decisions.
They’re rational responses to economics.
But if AI changes those economics, some assumptions deserve revisiting.
What happens if individual research becomes cheap?
If natural-language interaction becomes embedded into products?
If every customer can receive an explanation adapted to their circumstances?
If a small account can economically receive a level of service previously reserved for a large one?
Not all of this will work.
Customers won’t necessarily want AI involved in every interaction.
Poor automation can destroy experiences rather than improve them.
Trust matters.
Accuracy matters.
Human relationships matter.
But the opportunity is larger than reducing the cost of today’s customer service.
It includes asking whether the customer experience itself can be redesigned.
New capabilities are harder to put into a spreadsheet#
The hardest AI opportunities to value may be things the business doesn’t currently do.
There is no existing cost to reduce.
No employee hours to save.
No current process to benchmark.
Suppose AI allows a company to create a new service.
Or enter a market that previously required expertise it couldn’t economically provide.
Or analyse information it has collected for years but never had the capacity to use.
Or build small pieces of software that would previously never have justified development.
What is the ROI?
Before the capability exists, that can be difficult to answer.
This is a familiar problem with technological change.
If we evaluate a new technology entirely through the economics of the old way of working, we can systematically undervalue what becomes possible when the constraint changes.
The internet wasn’t valuable merely because it reduced the cost of sending letters.
Smartphones weren’t valuable merely because they allowed people to make telephone calls more efficiently.
Cloud computing wasn’t valuable merely because companies could run the same software on somebody else’s servers.
Each ultimately enabled products, behaviours and business models that were difficult to see if the analysis remained anchored to what existed before.
AI may prove similar.
We don’t yet know the scale or shape of its eventual economic impact, so comparisons with previous technological shifts should be treated cautiously.
But the underlying principle is useful:
A technology’s largest value can emerge from what it enables, not merely what it replaces.
Cost reduction still matters#
There is a risk of overcorrecting here.
Saving money matters.
For many businesses, particularly those operating on thin margins, an AI system that reliably removes significant cost could be far more valuable than an ambitious attempt to create something new.
Some processes genuinely should be automated.
Some repetitive work adds little value.
Some organisations carry costs technology can sensibly remove.
There is nothing strategically inferior about an efficiency opportunity simply because it is an efficiency opportunity.
The point is not:
Stop using AI to reduce costs.
It’s:
Don’t stop looking once you’ve found the cost savings.
A useful view of AI opportunity should include several different forms of value: cost, revenue, margin, capacity, decision quality, risk, customer value, speed and new capability.
Different businesses will find value in different places.
The important thing is that the search is broad enough to find it.
Ask what becomes possible#
AI will almost certainly continue to be sold to businesses through productivity.
Save time.
Automate work.
Do more with less.
Those messages are easy to understand because they fit neatly into the business we already know.
And some of the resulting opportunities will be substantial.
But leadership teams should add another question whenever they encounter a meaningful improvement in AI capability.
Not only:
What could this allow us to do more cheaply?
Ask:
What could this allow us to do that wasn’t previously economical, practical or possible?
That question opens a different set of opportunities.
Perhaps the answer is serving more customers.
Making better decisions.
Giving more employees access to scarce expertise.
Personalising something previously standardised.
Creating more capacity.
Building something new.
Or doing valuable work that the business has never performed because human intelligence was too expensive to apply to it.
The biggest AI opportunity inside a company may still turn out to be cost reduction.
But we shouldn’t assume the answer before looking. As I’ve argued before, businesses should start with where value exists rather than with a list of AI use cases, and understand the economics of the opportunity before deciding where the technology belongs.
Because if AI really does make useful intelligence cheaper and more abundant, its economic consequence won’t only be that businesses can do today’s work with fewer resources.
It will be that businesses can do things tomorrow that don’t make economic sense today.