Part of our guide: The real cost of AI
You're not looking for another article about how AI will "transform everything." You're looking at a budget, a board that wants efficiency gains and a growing pile of vendor proposals that all promise the world, none of them arriving with a credible business case.
Here is what the numbers actually look like.
The cost of doing nothing
Before we talk about AI ROI, let's talk about the cost of the current state.
In a typical mid-market business (100-2,000 employees), manual process costs are hidden across every function, with finance teams re-keying data between systems. Operations teams run reports by exporting to spreadsheets, customer service teams copy information from one screen to another and compliance teams manually review documents that could be parsed automatically.
None of these costs appear as a line item, and they're embedded in headcount, in overtime and in error rates. They also sit in the opportunity cost of people doing work that doesn't need a human brain.
A conservative estimate: 15-25% of total operational effort in a mid-market business is spent on tasks that are candidates for AI-assisted automation, not full automation. The human stays in the loop and the machine does the heavy lifting.
For a business with a £10m annual people cost, that's £1.5-2.5m in efficiency opportunity, and not all of it will be captured. Even capturing a third of it represents a material margin improvement.
What AI consulting actually costs
There's a wide range in the market, and the lack of transparency is part of what makes this hard to budget for, so here is a realistic picture.
The market breaks into three tiers, and at the top end Big Four firms charge six figures for a strategy engagement alone and take months to deliver. At the other extreme, tool vendors will give you "free" assessments designed to sell their platform, and the mid-market sweet spot is focused, independent consulting. That means strategy engagements measured in days not months, pilots that build real capabilities with governance and ongoing AI leadership that flexes with your needs. Our Breathe, Flow and Grow services sit in this space. The right investment depends entirely on the size and complexity of your business, so we scope every engagement individually rather than publishing a rate card.
Forget the vendor calculators, because the ROI framework that actually works starts with how you build a credible AI business case.
Start with cost avoidance, not cost reduction, because the most defensible AI business cases are built on avoiding future costs rather than cutting current ones. "We won't need to hire 5 additional people as we scale" is easier to approve than "we'll make 5 people redundant." Both statements have the same financial impact, and one of them is politically deliverable where the other often is not.
Then map effort reduction by process, and rather than calculating ROI for "AI" as a whole, break it down by specific processes: invoice processing, claims triage, contract review, customer onboarding and regulatory reporting. For each process, estimate the current cost (people × time × frequency), the expected effort reduction (typically 30-60% for well-chosen processes) and the implementation cost.
Finally, factor in the cost of delay, because if your competitors capture these efficiency gains 12 months before you do, they'll operate at a structurally lower cost base. That difference matters most in competitive markets and especially in PE-backed businesses with Value Creation Plans.
The research backs this up, and Microsoft's 2024 SME study found that businesses integrating AI report productivity gains between 27% and 133%. The wide range reflects the difference between "gave everyone Copilot and hoped for the best" (27%) and "strategically identified the right processes and built proper capabilities" (133%). The UK government's AI Action Plan projects a £47 billion boost to the UK economy from AI adoption. For mid-market businesses specifically, BDO's 2026 research shows that AI and productivity are now the primary growth lever for 42% of firms. The number that matters most is this: 61% of SME leaders say their company lacks a vision and a plan to implement AI. That means the opportunity is still wide open, and the businesses that build a coherent plan now will capture disproportionate advantage.
The business case structure
If you need to take something to your board, here's the framework.
Cost of current state: the calculation is total operational effort on automatable tasks × average cost per FTE, expressed as an annual figure.
Realistic capture rate: the figure to use is 20-40% of the theoretical opportunity, phased over 18 months, and that is conservative and credible.
Investment required: the shape is discovery + first pilot + ongoing support, scoped to your business. Get in touch for a straight conversation about what the right level of investment looks like.
Payback period: for well-chosen pilots the payback runs to 6-12 months and for portfolio-wide programmes it runs to 12-24 months.
Cost of inaction: the measure is competitor efficiency gap × time, and in PE-backed businesses that also means slower progress against the Value Creation Plan.
We provide this framework as a downloadable template, pre-structured with the right categories so you can fill in your own numbers.
A practical first step
The Breathe engagement at Oxygen Bubbles gives your leadership team a capability-scored AI roadmap with a clear business case for each priority initiative. You'll know exactly where the ROI is, what it costs to capture it and what to do first.
The work takes 5-10 working days, involves no tools to buy and delivers a plan your board can stand behind.
If you want a framework for structuring the business case, get in touch and we will share our template.
If your CEO is driving this conversation, share this with them: The Blockbuster Question: Is Your Business Model Ready for the AI Era?
If your CRO needs the governance view, share this: Introducing AI in a Regulated Business: A Risk Officer's Practical Checklist