How to Use AI to Improve EBITDA: From Productivity to Profit
- Paul Bensley
- Aug 13
- 13 min read
Updated: Aug 27
AI is increasingly being discussed in boardrooms as a productivity opportunity. Businesses are investing in tools, encouraging employees to use AI and measuring how many hours are being saved as a result. But using AI to improve EBITDA requires something more than simply demonstrating productivity gains.
If AI reduces a task from three hours to thirty minutes, the business has created two and a half hours of capacity. Unless something happens with that capacity, the salary is still being paid, the employee is still employed and the operating cost appearing in the P&L hasn't changed.
There is an important distinction between AI efficiency and financial improvement, and I think it is one that businesses need to understand much better.
Saving time is relatively easy to demonstrate. Turning that time into additional revenue, better margins or a structurally lower cost base requires management action.
If we want AI to improve EBITDA, we need to start thinking beyond productivity.
Start With EBITDA, Not AI
One of the mistakes I see businesses making with AI is starting with the technology and then searching for applications. A new tool becomes available, someone demonstrates what it can do and the organisation starts looking for places to use it.
I would reverse that thinking.
Start with the P&L.
Where are you losing revenue? Where is gross margin leaking? Which costs are increasing? Where are people spending time on activities that don't require their full capability? Where are customers leaving? Where are sales opportunities being missed? Where is poor information causing poor decisions?
Once you start there, AI becomes a means of improving a business rather than an objective in itself.
For most businesses, there are three broad ways AI can influence EBITDA:
Grow revenue. Improve gross margin. Reduce or avoid operating cost.
There is a fourth benefit that sits between them: creating additional organisational capacity. That can ultimately affect all three, but only if management decides what to do with the capacity created.
This is why I think every meaningful AI initiative should eventually be capable of drawing a line from the technology to the P&L.
AI Intervention → Operational Improvement → Financial Outcome → EBITDA
If you can't explain that connection, you may still have a worthwhile experiment, but you probably don't yet have an EBITDA improvement.
Revenue Is Often the More Interesting Opportunity
Most conversations about AI begin with cost reduction. That is understandable because efficiency is relatively easy to see and measure. If a task takes four hours today and one hour tomorrow, the improvement is obvious.
But for many businesses, particularly those with relatively high gross margins, using AI to improve revenue could be considerably more valuable.
Imagine a £50 million business operating at a 30% gross margin. Finding another £1 million of profitable revenue potentially creates £300,000 of additional gross profit before considering any incremental costs required to service it.
Compare that with trying to remove £300,000 from an already stretched operating cost base.
Both improve EBITDA, but they are very different conversations.
This is where commercial leaders should start looking much more carefully at how AI can support the sales process.
A salesperson managing 100 customers probably doesn't have enough time to analyse every account properly. Which customers are buying less than they did twelve months ago? Which product categories have disappeared from their purchasing pattern? Which customers buy one product from you but logically should be buying another? Which quotations haven't been followed up? Which accounts have characteristics similar to your most profitable customers?
The information often already exists somewhere in the business. The problem is the amount of human effort required to analyse it continuously.
That is exactly the type of problem AI can help with.
Instead of asking a salesperson to look through hundreds of transactions and identify patterns, AI can help surface the accounts that deserve human attention. The salesperson still decides what to do, but they spend more of their time acting on opportunities and less of it trying to find them.
That is a very different use of AI from writing emails faster.
Look for Revenue Leakage Before Chasing New Revenue
One of the things I have learned in commercial roles is that growth isn't always about finding new customers. There is often a surprising amount of revenue already leaking out of the existing customer base.
Customers gradually reduce their spending. Product categories disappear from orders. Quotations aren't followed up. Contracted customers buy outside the agreement.
Salespeople focus on the customers shouting loudest rather than those with the greatest opportunity.
Individually, these movements can be difficult to notice.
Collectively, they can represent a significant amount of money.
AI is particularly useful where the business has enough data for patterns to exist but too much data for people to review it manually.
Rather than asking a sales manager to inspect thousands of customer and product combinations, you can start asking much more useful questions. Which customers have reduced their purchasing frequency? Which customers are buying materially less than comparable accounts? Which product combinations normally appear together but are missing from particular customers? Which customers are showing behaviour associated with previous customer losses?
AI doesn't need to make the commercial decision.
It needs to tell the salesperson where to look.
That is often enough to create value.
Gross Margin May Be an Even Bigger Opportunity
Revenue naturally gets attention because it appears at the top of the P&L, but I think AI has enormous potential further down.
Margin leakage is rarely one large problem. It tends to be hundreds or thousands of small decisions accumulating across the business.
A salesperson gives slightly too much discount. A customer remains on an old price. Freight is absorbed when it shouldn't be. A low-margin product becomes a larger part of the sales mix. A customer who looks valuable by revenue becomes much less attractive once rebates, returns, delivery costs and service requirements are considered.
Humans are not particularly good at spotting these patterns across millions of transactions.
AI and modern analytical tools are.
Imagine being able to ask which customers have experienced the greatest margin deterioration over the past six months and why. Which salespeople consistently discount particular product groups more heavily than their peers? Which customers generate strong revenue but weak contribution after cost-to-serve? Which products are frequently sold together at unnecessarily low combined margins?
These are commercial questions, not technology questions.
And relatively small improvements can matter.
A business generating £50 million of revenue at a 30% gross margin produces £15 million of gross profit. Improving gross margin by just one percentage point increases gross profit by £500,000, assuming revenue and other costs remain unchanged.
That is why I would encourage businesses looking at AI to spend at least as much time thinking about margin intelligence as they do productivity.
Saving ten minutes writing an email is useful.
Finding £500,000 of margin leakage is considerably more interesting.
Pricing Is Particularly Well Suited to AI
Pricing is one area where I think there is considerable untapped potential.
Many businesses still manage pricing using a combination of price lists, spreadsheets, salesperson judgement and historical precedent. The result is often inconsistent discounting and very limited understanding of customers' actual willingness to pay.
AI doesn't need to take control of pricing to add value.
It can identify anomalies.
Why is this customer receiving a 14% discount when similar customers receive 8%? Why has the realised price of this product fallen while input costs have increased? Which customers are approaching renewal with margins materially below the target? Where are salespeople repeatedly overriding recommended pricing?
A commercial manager might eventually find those things manually.
The advantage of AI is its ability to keep looking.
That creates an important distinction. Some of the most valuable uses of AI won't involve asking it to do something. They will involve asking it to notice something.
Cost Reduction Needs More Honest Measurement
This is where I think businesses need to be particularly disciplined.
Imagine a finance team produces a monthly report that requires 40 hours of work. AI reduces that to ten hours.
It is tempting to claim a 30-hour cost saving.
But what cost actually disappeared?
If the same people remain employed on the same salaries, the answer is none.
The business has created 30 hours of capacity.
That capacity may still be extremely valuable, but it needs to be described accurately. Perhaps the finance team can now complete analysis it previously didn't have time to do. Perhaps growth can be absorbed without hiring another person. Perhaps external analytical support can be reduced. Perhaps roles can eventually be redesigned.
Each of those could create a genuine financial benefit.
But the benefit comes from what management does after AI creates the capacity.
I think this distinction will become increasingly important as companies start reporting the returns generated by their AI programmes. There is a danger of businesses adding together thousands of theoretical hours saved, multiplying them by employee hourly rates and announcing millions of pounds of benefits that never actually appear in the accounts.
That isn't an EBITDA improvement.
It is a capacity calculation.
The two shouldn't be confused.
Cost Avoidance Is Often More Realistic Than Cost Removal
There is another financial benefit that deserves more attention: cost avoidance.
Imagine a business is growing at 15% per year. Historically, every additional £5 million of revenue requires another two people in customer service, another finance administrator and additional sales support.
If AI allows the existing team to support that growth without those additional hires, the financial benefit is very real.
Nobody has been made redundant and no existing salary has disappeared from the P&L, but the future cost base is lower than it otherwise would have been.
That can be particularly powerful in growing businesses.
Rather than asking:
"How many people can AI replace?"
I think a much better question is:
"How much bigger can this business become without its cost base growing at the same rate?"
That is a fundamentally different way of thinking about AI productivity.
It is also much more constructive.
AI Can Improve the Quality of Decisions
There is another route to EBITDA that is harder to quantify but potentially significant: better management decisions.
Businesses generate enormous amounts of information, but managers rarely have the time to absorb all of it. Sales reports, customer feedback, market intelligence, operational data, financial performance, competitor activity and employee information all compete for attention.
AI can dramatically reduce the cost of analysing that information.
That doesn't mean asking AI to make the decision.
It means using it to improve the information surrounding the person making it.
A commercial director considering a price increase could use AI to analyse customer profitability, historic pricing movements, competitor information, customer correspondence and potential exposure before making a decision. A managing director considering an investment could ask AI to challenge the assumptions in the business case, identify missing risks and model alternative scenarios.
The financial impact of one better decision can dwarf thousands of small productivity improvements.
This is also why human judgement becomes more important as AI capability increases. AI can generate options, identify patterns and challenge assumptions, but somebody still needs to understand the business well enough to decide what matters.
Build an AI-to-EBITDA Bridge
One practical discipline I would introduce into any significant AI project is requiring the sponsor to explain how the initiative reaches EBITDA.
Not with a complicated business case.
Just a simple bridge.
What problem are we solving?
Be specific. "Improve productivity" isn't a problem. "Our salespeople spend six hours per week manually researching customers and preparing for meetings" is.
What will AI change?
Explain the operational intervention. Perhaps account research becomes automated and salespeople receive a structured customer brief before each meeting.
What operational measure should move?
Hours spent researching, meetings completed, quotations generated, response times or another observable measure.
What financial measure should move?
Revenue, gross margin, headcount requirement, external expenditure, cost-to-serve or another P&L measure.
How does that reach EBITDA?
This is the final test.
Consider two projects.
The first automates monthly reporting and releases 30 hours of finance capacity. Unless the business can explain what happens to those 30 hours, the EBITDA benefit is currently unproven.
The second identifies customers whose spending is declining and gives salespeople an intervention list every Monday. If that activity prevents £500,000 of annual revenue from disappearing at a 30% contribution margin with no meaningful increase in fixed cost, the route towards approximately £150,000 of contribution before any incremental costs is much easier to understand.
The second initiative might actually be less technologically impressive.
It could also be considerably more valuable.
Don't Automate a Bad Process
There is another trap worth mentioning.
AI can make processes faster without making them better.
If a business has an unnecessary report, automating its production doesn't suddenly make the report valuable. If the sales process is poorly designed, automating parts of it can simply help the business execute a poor process more efficiently.
This is why I would always ask whether the activity should exist before asking how AI can improve it.
Eliminate → Simplify → AI-enable → Automate
Not everything needs automating.
Sometimes the best productivity improvement is simply stopping doing something.
Measure AI Like Any Other Investment
Eventually, I think businesses will stop treating AI as something special.
It will simply become another form of investment competing for capital and management attention.
That means it should increasingly be subjected to the same questions we would ask of any other initiative.
What problem does this solve? What does success look like? What will we measure? What will it cost? What is the expected financial return? How quickly will we know whether it is working?
The technology may be new.
Those questions aren't.
One of the advantages businesses have today is that many AI experiments are relatively inexpensive. You don't necessarily need a six-figure investment to discover whether an idea has value. You can test something manually, prove the principle and then decide whether integration or automation is justified.
That makes the sequence important:
Problem → Experiment → Operational Value → Financial Value → Scale
Prove the value before building the infrastructure around it.
The AI Opportunities I Would Look for First
If I were reviewing a business specifically looking for AI opportunities that could influence EBITDA, I wouldn't begin with a catalogue of AI tools.
I would ask the management team to identify where money is currently leaking out of the organisation.
Where are customers leaving? Where are quotations being lost? Where are prices inconsistent? Where is margin declining? Where are rebates or discounts being applied unnecessarily? Which customers cost significantly more to serve than others? Where are expensive employees spending time on low-value administration? Which external costs exist because internal teams don't have enough capacity? Where will additional headcount be required as the business grows?
Then I would rank those opportunities by financial value, feasibility and risk.
That produces a very different AI roadmap from asking each department what software it would like.
It also makes the conversation much easier at board level because AI stops being an abstract technology investment and starts becoming a series of commercial hypotheses that can be tested.
From Productivity to Profit
There is nothing wrong with using AI to save time. In fact, it is probably where most organisations should begin because the experiments are relatively simple and the risks are low.
But it shouldn't be where the conversation ends.
The next stage is understanding what happens to the time you save.
Then looking beyond time altogether.
Can AI help you retain customers you would otherwise lose? Can it identify sales opportunities people cannot see? Can it improve pricing discipline? Can it expose margin leakage? Can it allow revenue to grow without overhead growing at the same rate? Can it give managers better information before they make expensive decisions?
Those are EBITDA questions.
And I think they are ultimately where the most interesting business applications of AI will be found.
Businesses don't need an AI strategy that demonstrates how much AI they are using. They need a business strategy that identifies where AI can create measurable value.
The question therefore isn't:
"How much time has AI saved us?"
It is:
"What did we do with the time, insight and capability AI created, and where did that appear in the P&L?"
When businesses can answer that question, AI has moved from productivity to profit.
Read More:
FAQs:
How can AI improve EBITDA?
AI can improve EBITDA by contributing to revenue growth, margin improvement or operating cost reduction. This might come from increasing sales conversion, improving pricing decisions, reducing cost-to-serve, automating work, improving productivity or avoiding future costs. The important step is demonstrating how the operational improvement created by AI converts into a measurable financial outcome.
Does AI productivity automatically improve EBITDA?
No. AI may save employees significant amounts of time without changing EBITDA. If the released capacity is simply absorbed elsewhere, the organisation has achieved a productivity improvement but not necessarily a financial return. Leaders need to determine how that capacity will be converted into additional output, increased revenue, lower cost or avoided expenditure.
How can businesses turn AI productivity gains into profit?
Businesses need to decide what will happen to the capacity AI releases. It might enable employees to handle more customers, increase sales activity, improve service without additional headcount or reduce the resources required for an activity. The financial benefit comes from converting the productivity gain into a measurable business outcome, not from the time saving alone.
What AI use cases can improve EBITDA?
AI use cases with the clearest potential to affect EBITDA are usually connected to important commercial or operational drivers. Examples include improving sales conversion, reducing lost quotations, increasing pricing consistency, improving customer retention, reducing administrative workload, lowering cost-to-serve and avoiding additional headcount as the business grows.
How should CFOs measure the financial impact of AI?
CFOs should look beyond AI adoption and productivity metrics and identify the financial mechanism through which value will be created. For each initiative, this means establishing the operational impact, identifying the affected business driver and measuring how that change influences revenue, margin or operating cost.
What is the difference between AI productivity and AI ROI?
AI productivity measures whether AI enables work to be completed faster or more effectively. AI ROI considers whether the value created justifies the investment required. An organisation can therefore achieve substantial productivity gains while generating relatively little financial return if those gains are not converted into measurable business value.
How can AI increase revenue rather than just reduce costs?
AI can support revenue growth by helping businesses identify opportunities, improve sales prioritisation, respond to enquiries faster, improve conversion rates, strengthen customer retention and make better commercial decisions. The financial impact should be measured through changes in relevant revenue drivers rather than simply measuring AI usage.
Can AI improve EBITDA without reducing headcount?
Yes. Improving EBITDA through AI does not necessarily require reducing existing headcount. AI can allow an organisation to grow without adding employees at the same rate, increase output using existing resources, improve sales performance, strengthen margins or avoid future costs. Headcount avoidance can therefore be an important source of financial value.
What is the AI-to-EBITDA Framework?
The AI-to-EBITDA Framework, developed by Paul Bensley, provides a structured way of connecting AI activity to financial performance. It traces value through five stages: AI Use Case → Operational Impact → Business Driver → Financial Conversion → EBITDA Impact.
Why do businesses struggle to prove the financial value of AI?
One reason is that organisations often stop measuring too early. They demonstrate that AI has been adopted, tasks have been automated or hours have been saved, but do not establish what happens next. The challenge is connecting those operational improvements to the business drivers that ultimately affect financial performance.
Comments