Written by Ash Amrite, Head of Solutions and AI Transformation, Merkle
AI has moved beyond the experimentation phase. The harder part starts now.
Dentsu's 2026 CMO Navigator found that nine in ten marketing leaders say emerging AI capabilities are already reshaping their strategies. And the direction of investment is equally clear: Merkle's 2025 Global CMO Navigator: CX Edition found that eight in ten CMOs said their organisations were making significant investments in generative AI.
What they want from that investment is telling. Increased marketing effectiveness was cited by 44% of CMOs, while 42% pointed to increased process efficiency and 42% to greater team productivity. By comparison, only 19% identified reducing marketing headcount as a key benefit.
AI, in other words, is increasingly being expected to make organisations work better, not simply smaller.
But that creates the next problem: what is “better” actually worth?
There is no shortage of enterprise technology investment sold on the promise of efficiency. What remains scarce is a credible, repeatable method for proving that efficiency actually materialised, and quantifying it in terms the CFO recognises.
And the scrutiny is increasing. Merkle's 2026 Transformation Gap research found that while 95% of enterprise leaders are allocating budget to generative AI, only 35% have achieved enterprise-wide adoption.
The question is therefore shifting from “can we use AI?” to “can we prove that using it is creating value?”
Productivity alone isn't a business case
Productivity is directional. It tells you things got faster or easier – but it doesn't tell you what that speed was worth, to whom, or whether the organisation captured it.
A team that resolves incidents 40% faster has undoubtedly improved a metric… but what has that improvement delivered? Has it recovered revenue? Freed people to do higher-value work? Reduced the need for additional hiring? Lowered customer churn or operational risk? Or has everybody simply gained another 20 minutes in which to answer email?
Without a structured value framework, productivity gains remain anecdotal, celebrated in quarterly business reviews and forgotten by budget season.
Realised business value answers a different question: not “are we more efficient?” but “what did that efficiency actually deliver?”
That is the thinking behind what we at Merkle call the AI Speed to Value Equation: a way of moving from activity metrics such as hours saved, tasks automated and tickets resolved to a defensible measure connecting operational change to financial outcome.
It starts by separating AI value into three distinct components.
1. Workflow efficiency: what is the time genuinely worth?
The most obvious AI productivity gain is time. Processes become shorter, manual steps disappear, and repetitive tasks are automated.
But converting time saved into pounds is not as simple as multiplying hours by salaries. A useful calculation looks like this:
Workflow Efficiency Value = (Annual Task Volume × Time Saved per Task × Hourly Rate) × Capture Rate
The important variable is the capture rate. Not every minute saved through AI becomes a minute productively reinvested. People switch context. Work expands. Processes contain friction. An organisation might theoretically release 1,000 hours without gaining productive use of all 1,000.
The capture rate is the honest variable many productivity calculations omit. By including it, you turn an aspirational saving into a more defensible one. This is particularly relevant in document processing, approval chains, data reconciliation, reporting cycles and other recurring operational tasks where volume and time can be reliably baselined.
But time is only the first part of the equation.
2. Issue resolution: what does being faster prevent?
Every unresolved or slowly resolved issue carries a cost. Sometimes that cost is visible: SLA penalties, customer churn, incident overtime or lost transactions. Often it is hidden: engineering distraction, compounding technical debt or reputational exposure.
As a result, the next part of the equation needs to capture both the operational benefit of resolving an issue faster and, where relevant, the wider business impact avoided as a result.
Issue Resolution Value = Labour Value of Faster Resolution + Business Impact Avoided
For an organisation where service disruption carries a direct revenue consequence, reducing mean-time-to-resolution isn't simply a service-desk metric; it's a P&L variable.
If reducing mean time to resolution (MTTR) by 30 minutes across 200 annual incidents prevents £10,000 of business impact for every hour of disruption, that represents £1 million in recovered value. Clearly, that number belongs in the board briefing – and it's the kind of number that turns a CIO's operational metric into a CFO's investment case.
First-contact resolution matters too. Every escalation creates another layer of cost, so improving resolution at the first point of contact compounds across the total issue volume.
This is where measurement starts to change the AI conversation. Instead of saying an AI tool “helps teams resolve incidents faster”, you can show what faster resolution protects or releases for the business.
3. FTE optimisation: what can people now do instead?
This is the most politically sensitive pillar, and potentially the most financially significant.
The obvious temptation is to treat AI productivity as a headcount equation: automate X hours and therefore remove Y roles. That both misrepresents the opportunity and creates unnecessary organisational resistance.
Our CMO research bears that out. Marketing leaders are far more likely to look to AI for greater effectiveness, efficiency and productivity than for reduced headcount. The more useful question is how much capacity has been liberated from low-complexity, high-volume work, and what happens to that capacity next.
FTE Optimisation Value = Cost Avoidance + Incremental Value from Strategic Redeployment
The first component covers genuine cost avoidance: for example, additional hiring that is no longer required because rising workload can be absorbed within the existing team.
The second looks at what happens when people's time moves to higher-value activity. When senior engineers spend less time on repetitive triage, for example, they can spend more time on architecture, innovation and revenue-generating work. If you establish a credible baseline for the value of those activities, that differential can be measured too.
Imagine automation absorbs the equivalent of 2.5 FTEs of low-complexity work. Those people don't disappear; instead, their capacity moves. The question becomes: what did the organisation enable those 2.5 FTEs to do that it could not do before? That's a much richer measure of transformation than jobs removed.
Putting the equation together
These three pillars provide the positive side of the equation. But no credible business case should stop there.
AI transformation costs money. It also creates friction – so realised value needs to account for both:
Realised Business Value = Workflow Efficiency Value + Issue Resolution Value + FTE Optimisation Value − Implementation Cost − Change Friction
The change-friction allowance is important to account for because almost every significant operational shift causes a temporary productivity dip. (People need training, processes need adjusting, and adoption takes time.)
Leaving that out might make an ROI projection look more attractive, but including it makes the model more credible to experienced finance and operations stakeholders – CFOs in particular – who have watched overstated technology business cases fail to materialise. Merkle's Transformation Gap research suggests that funding and ambition are not the fundamental constraints on enterprise AI; the much bigger challenge is moving from pilots and pockets of adoption towards transformation at scale.
And then there is one final variable: time. £2 million of value realised over 36 months is materially different from £2 million realised over 12. This is why speed to value determines how quickly an organisation validates its investment, how soon it can reinvest the return and, increasingly, whether the business case survives the next budget cycle.
Implementation cost is really Total Cost of Ownership
The “Implementation Cost” in the equation above is shorthand for something broader: the Total Cost of Ownership (TCO) of the AI capability across its full lifecycle, not just what it costs to launch. TCO breaks down into four categories.
Initial and acquisition costs cover software licensing or development, implementation and setup, and onboarding and training.
Operational and infrastructure costs are the recurring bills that keep the capability running – cloud hosting, and, for AI specifically, martech platform licences, API licences and token usage, which scale with adoption and vary by model and provider. This category also covers compliance, security and vendor management.
Maintenance and support costs include help desk services, SLAs, routine patching and the customisation needed as requirements evolve.
Internal and indirect costs are the least visible but often underestimated: staff time managing or working around the system, productivity lost to downtime, and eventual decommissioning and migration.
Modelling cost this way, rather than as a one-off figure, is what keeps the realised-value equation credible over time. A capability that looks cheap to deploy can look very different once licensing, token consumption and support costs compound at enterprise scale.
This takes cross-functional ownership: the CTO and CIO model infrastructure and licensing economics up front, the CFO translates that into realised-value terms, and the CMO flags which use cases will scale fastest – and where cost will therefore grow fastest too.
Three disciplines make the calculation work
None of this requires perfect data. But it does require greater discipline than simply declaring that AI has saved thousands of hours.
Three practices make the biggest difference:
Baseline before you optimise. You cannot calculate time saved without knowing how long something took before. Instrument processes before transformation, not after.
Separate volume from impact. High-volume, low-impact gains can dominate a headline number while masking underperformance on the issues that really matter. Report both.
Review quarterly, not annually. Realised value drifts. AI adoption changes, process volumes shift and cost structures evolve. A quarterly value review keeps the equation honest and identifies course corrections while there is still time to make them.
AI now has to prove its value
For the first phase of enterprise AI, experimentation itself had value. Organisations needed to learn what the technology could do, where it could operate safely and how people would use it.
Now we are moving beyond that phase. With nine in ten CMOs already saying AI is reshaping their strategy, the debate is no longer really about whether the technology will matter. The much more useful conversation is about where it creates measurable value – and how quickly organisations can capture it.
As AI becomes embedded in more workflows and attracts a larger share of transformation budgets, “we saved some time” is an increasingly inadequate answer.
The organisations that keep winning internal investment won't necessarily be those with the most impressive partner demos or the most compelling vendor pitch decks. They will be the ones that arrive at the budget conversation with a number, a methodology and the audit trail to defend it. That's as true for the CMO defending a martech roadmap as it is for the CTO defending an infrastructure spend.
That means translating the language of productivity (faster, simpler, better) into the language of business: pounds/dollars recovered, capacity freed, risk reduced and value created.
That translation is the difference between AI that gets funded as an experiment and AI that earns its place as infrastructure.
Want to understand what your AI productivity gains are really worth? Contact Merkle to explore how our AI Speed to Value Equation can help you establish a defensible baseline, quantify realised value and build the business case for what comes next.
