For most of the post-ChatGPT period, AI investment in UK mid-market businesses has been treated as a discretionary experiment. A line item on the innovation budget. A project the CTO and the CFO discuss for fifteen minutes a quarter and otherwise leave alone. That framing made sense in 2023. It is now actively dangerous.
In 2026, AI investment has crossed the line from experimental to compounding. The maths underneath it have changed, and most CFOs have not yet noticed. I founded Cybix specifically to help UK businesses navigate this shift, and I want to lay out — in capital-allocation terms — what is actually happening to the businesses that have moved early, and what is happening to the businesses that have not.
The asymmetric maths nobody is showing you
Most CFOs evaluating an AI initiative use the same framework they use for any technology investment: cost up front, savings over time, payback period, risk-adjusted return. This produces a number. The number is usually unspectacular. The case sits in committee for another quarter.
What this framework misses is something Cybix has watched play out across multiple client engagements: AI investment in 2026 produces compounding returns within a single business, not just linear ones. Once a workflow is automated, the cost of automating the next workflow nearby drops sharply. The data layer is reusable. The integration patterns are reusable. The operating model is reusable. The team is faster. The Cybix engagements that have been live longest now ship new automation in a fraction of the time the first one took, with a fraction of the marginal cost.
The implication for capital allocation is profound. The first AI workflow has a payback profile any CFO would recognise. The fifth workflow has a payback profile no CFO has built a model for, because traditional capex maths does not capture the compounding effect of a maturing AI operating capability. This is the single most important number CFOs are not currently tracking, and Cybix’s internal benchmarks suggest the gap between what a standard business case predicts and what a mature programme actually returns is now substantial.
The cost of staying still
The reverse of compounding is decay. A business that has not invested in AI delivery capability does not stand still — it loses ground every quarter relative to competitors who have. The Cybix view, formed across engagements in six sectors, is that this gap has now opened to the point where it is visible in operating cost ratios, customer response times, and pricing flexibility.
Concretely: a UK financial services firm with 20% of customer onboarding automated has a different cost-to-serve than a competitor at 0%, and the gap widens every month. A retail business with automated returns processing has different working capital dynamics than one without. A healthcare provider with automated triage has different staffing requirements than one without. None of these gaps appear in a single quarter’s P&L, but they appear in the trend line, and by the time they show up in benchmarking reports, catching up is much more expensive than keeping up.
CFOs who have not yet built the cost-of-inaction line into their AI business case are working from a 2024 model in a 2026 market. Cybix’s experience across regulated and unregulated sectors is that the gap can be quantified, and finance teams that quantify it tend to fund AI delivery materially differently from those that don’t.
Why this is now a capital allocation question
The natural objection is that AI investment is operating expenditure, not capital expenditure, and therefore not a capital allocation decision in the technical sense. This is becoming a less useful distinction. The Cybix model treats every meaningful AI engagement as having a balance-sheet effect, because the resulting capability — automated workflows, reusable data layers, in-house engineering competence — is durable and produces compounding returns.
Whether or not it appears under capex on the books, the decision to invest (or not invest) in AI delivery capability is the same shape as a capital allocation decision: significant cost now, durable competitive position later, opportunity cost of waiting.
CFOs who are still treating AI as innovation-budget OpEx are pricing it wrong. Cybix’s experience is that the businesses that grasp this earlier are the ones that fund AI engagements at the scale the maths actually justifies, and those are the businesses that build the compounding advantage. The Cybix engagements that have produced the most economic value to date are the ones where the CFO ran the case as a capital decision, not a project budget.
What CFOs should be tracking
If you are the CFO of a UK mid-market business and want to bring AI investment into the same rigour as your other capital allocation decisions, four KPIs are worth adding to the next finance pack.
Operations automated, by workflow. Not pilots. Not deliverables. The percentage of a defined operational workflow that now runs without human touch. The Cybix benchmark, in our most mature engagements, has reached up to 84% on a single workflow.
Marginal cost of the next workflow. Track what it cost to automate the first workflow versus the second, third and fourth. If the cost is not declining, the compounding effect is not happening, and the AI capability is being built as a series of disconnected projects rather than a cumulative platform. Cybix tracks this curve on every engagement and treats a flat curve as a programme failure indicator.
Operating cost per unit, pre- and post-automation. Customer service cost per ticket. Compliance review cost per filing. Reconciliation cost per invoice. Whatever the unit of operation is in your business, that number should be moving in one direction after automation.
Cycle time on new automation. The single best leading indicator of compounding is whether the time-to-deploy a new workflow is shortening. If it is, the operating capability is real. If it is not, what looks like AI investment is actually a series of one-off integrations. The Cybix benchmark is that engagement four onwards typically ships in half the calendar time of engagement one.
These four KPIs together produce a capital allocation picture that the existing finance machinery does not generate by default. Building them into the quarterly cycle is, at this point, a leadership decision more than a tooling one.
What this means for your next budget cycle
The practical recommendation is straightforward. If your business is in the AI-experimenting cohort — pilots in flight, no automation in production — the next budget cycle is the one that decides whether you spend the next three years compounding or the next three years catching up. The maths now favours decisive investment, not incremental experimentation. The Cybix experience is that the businesses that commit at this stage are the ones that look materially different by 2028.
If your business is already in the compounding cohort — automation in production, marginal cost declining, second and third workflows underway — the recommendation is to defend that position. The competitors who have not yet started are not standing still by default; they are looking at the same maths and may be about to commit. The window where AI-ready operations is a defensible advantage is open now and will narrow over the next 24 months.
Either way, the conversation has moved from “should we?” to “how aggressively?” — and that is a conversation the CFO needs to lead, not delegate. The capital allocation lens is the right one. The Cybix view is that boards which adopt it now will, by the end of the decade, be operating at a structurally different cost base than boards which did not.
—
Dan Spence is the CEO of Cybix, a London-based AI consultancy that helps UK and international businesses convert AI investment into compounding operational capability. The Cybix engineering team — drawn from former Apple and Google language-model engineers alongside compliance, automation, recruitment and software specialists — has automated up to 84% of operational workloads for clients across banking, healthcare, telecoms, retail, oil and gas, and fashion. Cybix measures every engagement in the currency CFOs care about: operations automated, marginal cost reducing, cycle time shortening. More at cybix.ai.
