Only fund the AI projects with the best ROI.
Most organisations now have more AI use cases than budget, capacity or data to support them. We assess each on risk-adjusted return against total cost of ownership, prioritise the portfolio, and reshape promising candidates until the business case holds.
Discuss your investmentTreat AI like the growth investment it is.
Capital investment follows a disciplined process: a business case with explicit assumptions, a view of the downside, a hurdle the return must clear, and gates at which funding continues or stops. AI use cases rarely receive the same treatment. They arrive from vendors, business functions and innovation teams with incompatible assumptions and single-point forecasts. That makes them difficult to compare and easy to over-fund.
We bring investment discipline to the AI portfolio. Every candidate is assessed on the same basis: expected return, total cost of ownership, delivery feasibility and uncertainty. The result is a ranked portfolio with a clear decision for each use case: fund, reshape or park.
Unlike a passive investor, your organisation controls the asset. It can resolve the most material uncertainty before committing capital, redesign a use case that falls short, and measure realised value once the system is in operation.
Every business case starts from the cost of the current process.
For each candidate, we establish what the affected process costs to run today: the people, licences and systems behind it, taken from your finance records. Against that baseline we set the total cost of ownership of the proposed solution over its full life. That covers data preparation, integration, infrastructure and models, engineering, and ongoing maintenance and support. The difference defines the break-even: the improvement the use case must deliver before it returns anything at all.
We then test the assumptions behind the expected return. They typically come from three sources: the proposal itself, the people who run the process, and the KPIs already reported. We record each as a range. Where they diverge, the divergence is itself a finding. It identifies the assumption the investment decision depends on, and how much evidence supports it.
Know what realising the return will take before you fund it.
A business case can clear its hurdle on paper and still fall short in operation. For each candidate we assess three feasibility constraints separately, because each affects the case through a different mechanism.
- Data readiness. Whether the records the system depends on are complete, accurate and accessible, and what preparing them would cost. That cost belongs in the total cost of ownership.
- Process stability. Whether the process is settled enough for a system built on it to stay useful. A process that is still changing shortens the period over which the investment can pay back.
- Change effort. What the people who run the process must do differently, who has to approve the change and how long adoption will take. Slower adoption delays the benefits and lowers the return within the same horizon.
Where the current process is poorly documented, we apply process mining to the event logs of the systems it runs on. This shows how the process actually runs, whether it is worth automating and which steps to automate first.
A decision with its uncertainty attached.
We simulate the range of return against the full cost many times over, which gives a distribution of outcomes instead of a single forecast: the chance the use case fails to pay back, the assumption driving that risk, and the one worth testing before you commit. Each candidate leaves with a decision, fund, reshape or park, and the reason behind it. An estimate is not a realised return. Once a system is in operation, value is measured in use against the baseline and the forecast.
The same discipline we apply to cyber risk.
The team that evaluates AI proposals is the team that quantifies cyber risk for our customers using FAIR, the open standard for expressing risk in financial terms. Estimating uncertain loss in money with the assumptions visible is the discipline an investment case needs, with the upside added.
What could change the investment decision?
An order-entry process in a consumer goods business costs AED 2,466,000 a year in people, licences and systems. The proposed AI system costs AED 1,849,500 over three years to prepare data, integrate, run and maintain. Break-even is a sustained 25% reduction in the cost of the process before the investment returns anything. The proposal claims 40%. The process owners estimate 15 to 30%. Simulating those ranges against the full cost gives roughly a one-in-three chance of not paying back.
The assumption that matters most is how much saved time becomes lower cost rather than time spent elsewhere. Decision: reshape, test that assumption on one order stream for a quarter, then fund or park.
Illustrative example. Amounts use a fixed EUR 1 = AED 4.11, rounded from the Central Bank of the UAE rate for 22 October 2026, not live exchange rates.
Exhibit · Illustrative example
Roughly one in three simulated outcomes does not pay back the full cost
Order-entry process, AED 2,466k a year to run; proposed system AED 1,849.5k over three years