The meter is running and nobody is reading it.
Control the cost. Improve the forecast. Show the return.
AI changes both the cost base and the assumptions behind your business plan. We help you assess spend, forecast scenarios and measure commercial impact, with finance keeping ownership of the decisions.
growth in tokens used against 497% growth in spend, January 2025 to April 2026 (vendor data from its customers).
Ramp AI Index, April 2026of IT leaders say they have been hit by unexpected AI charges (vendor survey).
Zylo SaaS Management Index, 2026of companies say they fully track the cost of AI; 22% see little until the invoice arrives.
KPMG Global AI Pulse, 2026Six problems in finance's own words, and what each one costs.
rise in tokens used between January 2025 and April 2026, while spend rose 497%. Tokens are the units AI is billed in.
Ramp AI Index, April 2026 (vendor data)Unit prices fall, yet bills rise because use grows faster. What it costs you: a budget line that is a guess, and variances found after the month closes.
Spend sits in four places: per-seat licences, usage inside the cloud bill, AI add-ons that arrive as renewal uplifts, and tools on expense cards. What it costs you: a number nobody can stand behind.
Ramp's April 2026 data show premium models took 56% of AI cost for 46% of tokens (vendor data). What it costs you: paying top price for tasks a smaller model could do.
Zylo's 2026 research found 36% of AI licences go unused (vendor survey). What it costs you: recurring spend on seats that deliver nothing.
Business units control most software spend, so the same job is bought several times. What it costs you: duplicate tools, and no volume leverage with suppliers.
KPMG UK's 2026 survey found organisations with full cost visibility were four times as likely to report an established return (25% against 6%, small sample). What it costs you: investment that cannot be defended at the next review.
Tokens are growing much faster than spend, and spend is growing faster than most budgets.
Plain usage, not headcount, now drives the bill. The chart uses only the sourced figures above.
Bar chart. Tokens used grew by 1,001 per cent and AI spend by 497 per cent, so usage grew about twice as fast as spend.
Tokens used (the units AI is billed in)+1,001%
AI spend+497%
Usage has grown about twice as fast as spend.
Stop asking whether AI is worth it. Ask which spend lines you can defend in one sentence.
A monthly total tells you the bill rose. It does not tell you why. Cost per task does: it shows whether the answer is a cheaper model, a shorter instruction or a process that over-runs. For each line, write one sentence: what it does, for whom, and how you would know it is working. A line with no sentence is not cancelled. It is flagged, and someone is asked to own it. That turns an argument about whether AI is worth it into a list you can manage.
Method drawn from GenWA's published guidance. It is a way of working, not a promised saving.
Six ways finance gets a number it can stand behind.
Cost Meter, Savings Finder, Value Register, Model Choice and Board Report are in early access. We confirm in writing what is live for you.
Cost Meter sets budgets and alerts per team and task, forecasts where spend is heading and names unused seats. One person can own the total.
About Cost MeterWhere the saving isSavings Finder finds premium models on routine tasks, repeated documents and duplicate seats, and estimates the saving in pounds with the drivers named. It is an estimate from your data, not a promise.
About Savings FinderThe right model for each jobModel Choice sends routine work to a suitable approved model and keeps premium models for the tasks that need them, under your rules.
About Model ChoiceCost against valueValue Register records the benefit and the cost of each use of AI, using figures your own teams record.
About Value RegisterA report for the boardBoard Report puts usage, spend, issues and value on one page on a regular cycle.
About Board ReportAn independent first viewThe Independent AI Audit analyses your usage data and invoices and reports where spend sits. Its Fees and any credit are agreed in writing.
See advisoryThree moves for the finance lead.
Three routes, matched to where finance sits.
Analysis of your usage and invoices, cost savings worked out from your data, model fit and a forecast. Analysis is delivered through our advisory service.
See advisoryProductCost Meter, Savings Finder and Value Register keep the number current after the audit. Set up with your team in an agreed environment.
See Cost MeterIncubatorIf finance wants a job automated, such as monthly board-pack preparation, we define, build and run it with a cost per outcome you can see.
See the IncubatorThe three colleagues who make the number real.
Holds the list of tools and the usage data behind the invoices.
For the CTO and CIOChief operating officerKnows which workflows depend on AI and what rework they create.
For the COOSees licences and renewals, and can renegotiate once the real use is known.
A company set no limit, and the annual AI tools budget was gone in four months.
Uber's chief operating officer was reported in 2026 to have said the company spent its annual AI tooling budget in about four months after staff were encouraged to use the tools as much as possible. It then introduced monthly caps per employee per tool, and the chief operating officer said it was very hard to draw a line from usage to features shipped. The lesson for finance is that a cap, a live meter and a measure of outcome are needed before the budget is exhausted, not after.
Reported by Bloomberg and TechCrunch, April to June 2026. Public incident. GenWA had no involvement.
Want a first view of what you could change?
Share your usage data for a figure based on your own spend, or book a demo of the cost view.
Cost falls when the work is matched to the model.
Two bars. Before: all work runs on premium models. After: routine work moves to smaller models, so the same work costs less.
- Premium models, kept for complex work
- Smaller models, for routine work
Keep reading
Analyse
Model maxing: why your AI bill grows faster than its value
Using the biggest AI model for every task feels safe. Why it happens, what it costs and where to start.
By Steven Myers · 5 min read
Grow
Time to value needs an operating model.
Good AI architecture connects the technology to the people, decisions and work required for the first useful outcome.
By Steven Myers · 2 min read
Grow
AI is a maze. Keep score on value, not just tokens.
Chasing the next breakthrough is absorbing. A useful AI operating model keeps the route, the cost and the result visible.
By Steven Myers · 2 min read
- AI costs spread across licences, consumption and teams
- Forecasts built on assumptions AI has changed
- Pressure to cut costs while funding growth
- Benefits that count saved minutes without showing cash impact
- Funding, fundraising or exit preparation without clear AI evidence
- Bring AI spend and ownership into one financial view
- Build scenarios and stress-test the assumptions behind the forecast
- Prioritise work by measurable return and time to value
- Prepare monthly board packs and decision-ready financial models
- Support funding and diligence preparation with traceable evidence
AI changes both the cost base and the assumptions behind your business plan. We help you assess spend, forecast scenarios and measure commercial impact, with finance keeping ownership of the decisions.
Find out more
Starting figures are examples. Edit the assumptions for your own work. This estimates released capacity, not guaranteed cash savings.
Challenges
Method: tasks × (current minutes × assumed reduction − review minutes) ÷ 60. Capacity value = net hours × loaded hourly cost. Implementation cost and revenue effects are excluded.
Straight answers.
How can a CFO see total AI spend in one number?
Start from usage as well as invoices. Cost Meter shows spend by team, tool and model with budgets and alerts, across connected AI. Before that, the audit builds the first total from expenses, cards, subscriptions and cloud bills, with a team against each line.
Why can't procurement control this on its own?
Procurement sees licences, but much AI spend arrives through expense cards, usage fees and features added to software you already pay for. Starting from usage finds spend that never passed through a purchase order, and shows where the same job is bought twice.
How do I forecast a cost that follows use?
Use cost per task rather than cost per month. Once you know tasks, models and volumes, you can forecast on the drivers. Our analysis reports where spend is likely to go and what changes it, such as model choice and shorter instructions. It is a forecast, not a guarantee.
What would a saving come from?
Typically three places: premium models used on routine work, licences nobody uses or buys twice, and context resent with every request. We size each from your data. Third-party research shows routing can keep most quality at much lower cost on some benchmarks, so we test against your own tasks.
How do I show a return before committing?
The audit gives a first view, and its Fees and any credit are agreed in writing. Value Register records task, outcome, cost and time saved for each use of AI. We do not promise a fixed return. You build the case from your own figures.
What does the audit cost?
Indicative audit fees, excluding VAT, are £7,500 for 50 to 250 people, £12,500 for 250 to 1,000 and £19,500 for enterprise. They are confirmed in writing before you commit and Fees and any credit are agreed in writing. if you go on to a platform plan.
What do you need from me to start?
Three months of expenses, card statements, software subscriptions and cloud bills, plus the team that owns each. If you have usage exports from AI tools, share them through the usage check, which reads them locally in your browser with no calls to an AI service.
How do we choose the first AI investment?
Bring AI spend and ownership into one financial view. We start with the problem, the accountable owner and a baseline. Agree a small scope, the cost of delivery and a measurable test before scaling.
How should we measure time to value?
Measure the elapsed time from an agreed start to the first accepted business outcome. Include setup, integration, human review and rework. A faster prototype is useful, but it is not the same as a result your business can use.
How does GenWA work with our team?
Choose the support you need: advice, modules, a scoped build, incubation or acceleration. We can build and run alongside your team. Your named people retain supervision and consequential decisions. Scope, fees, hosting and any outcome-based arrangements are agreed in writing.
See where the tokens go, and what to change.
Prices are confirmed in writing before you commit. We do not promise a fixed return.
Research behind these priorities
The five priorities are GenWA’s editorial selection, informed by the sources below and your role. The 2027 opportunities are planning scenarios, not a survey ranking or forecast.