Fund the operating change
AI Transformation Business Case
Build an AI transformation business case that compares credible alternatives, distinguishes cash from capacity and makes uncertainty visible before investment.
An AI transformation business case should justify a change in how the business operates and the commitment required to achieve it. It must explain why the proposed option is preferable to continuing as today, repairing the process or using existing software differently. The presence of AI does not remove the need for those comparisons.
The decision maker needs to see the whole commitment: discovery, data preparation, delivery, integration, staff participation, transition and ongoing support. They also need to know who will convert an improvement into a business benefit. A convincing demonstration and a theoretical time-saving calculation are useful evidence, but neither is a complete investment case.
Define the decision and the counterfactual
State what management is being asked to approve now. Is it an investigation, a bounded trial, a production release or a multi-unit programme? Give the scope and the next decision point. A request for a small discovery budget should not imply approval of the entire transformation.
Describe what happens without the investment, including planned changes that will happen anyway. If a new enterprise system already includes the required workflow function, its benefits cannot all be credited to a separate AI project. If demand is expected to rise, explain the evidence and the staffing or service consequences under the current approach.
The UK Treasury’s business-case guidance uses a structured approach to project and programme decisions. It is a methodological reference, not a Saudi approval requirement. The example below uses original assumptions to make an investment comparison transparent.
Compare options that could genuinely solve the problem
Consider an illustrative industrial-inspection business preparing reports from authorised field notes, photographs and standard checks. Its objective is to reduce avoidable report-preparation work without weakening professional review. The options are process repair, workflow-software configuration and AI-assisted preparation. A qualified reviewer still accepts the report; the AI option does not acquire authority to certify an inspection.
Each option needs a fair description. Process repair might standardise the evidence pack and remove duplicate entry. Workflow configuration might enforce completeness and assemble known fields. AI assistance might help interpret variable notes and draft narrative. Compare the complete task, including corrections and exceptions, rather than timing only the part that makes one option look strongest. Use-case discovery should establish which problem deserves this comparison.
| Option | Initial investment | Annual running cost | Assumed annual avoided expenditure |
|---|---|---|---|
| A: process repair | SAR 20,000 | SAR 8,000 | SAR 25,000 |
| B: workflow configuration | SAR 40,000 | SAR 15,000 | SAR 40,000 |
| C: AI-assisted preparation | SAR 120,000 | SAR 40,000 | SAR 60,000 |
These amounts are hypothetical planning inputs, not supplier quotes, observed customer results or Saudi market averages. The assumed avoided expenditure represents temporary report-preparation support the business could actually cancel. It must be evidenced by the relevant budget and an operational plan before it becomes a defensible cash benefit. The options are alternatives; their benefits are not additive.
Show the arithmetic and its limitations
At steady annual operation, net cash contribution is avoided expenditure minus running cost. That gives SAR 17,000 for A, SAR 25,000 for B and SAR 20,000 for C. Over two years, subtracting the initial investment gives:
- A: 2 × (25,000 − 8,000) − 20,000 = SAR 14,000.
- B: 2 × (40,000 − 15,000) − 40,000 = SAR 10,000.
- C: 2 × (60,000 − 40,000) − 120,000 = −SAR 80,000.
This simplified comparison excludes discounting, tax, financing, inflation, timing differences and residual value. It assumes the full annual benefit and cost begin immediately and remain constant. A real cash-flow model must reflect the organisation’s finance policy, payment schedule and adoption ramp. If benefits arrive later, the early cash position will be worse than this illustration.
Under these assumptions, the AI option is not the strongest two-year cash choice. That is a useful finding. It may have additional quality or growth value, but those claims require their own evidence and should not be inserted simply to rescue a preferred solution. Review wider cost categories in AI transformation cost in Saudi Arabia.
Keep released capacity separate from cash
Suppose, separately, the same business processes 12,000 reports a year and a trial suggests a four-minute reduction in complete-task effort per report. That is 12,000 × 4 ÷ 60 = 800 hours of potential capacity. At an illustrative loaded labour value of SAR 120 per hour, the capacity is valued at SAR 96,000. It is not automatically a SAR 96,000 reduction in expenditure.
If employees remain on the same payroll, the business needs a redeployment plan: additional work they can complete, backlog they can remove or hiring they can avoid. If the SAR 60,000 temporary-support saving in option C uses those same released hours, do not add SAR 96,000 as another cash benefit. Show the capacity measure as explanatory evidence and count the realised financial effect once.
Assign a benefit owner who controls the operating action. A technology supplier cannot cancel a staffing budget or guarantee additional demand. Distinguish the transformation’s total benefits from the contribution of advisory work; the consulting ROI framework addresses that separate attribution question.
Test the assumptions that could reverse the decision
Focus sensitivity analysis on the uncertain variables with decision impact: eligible volume, effort after review, adoption, avoided expenditure, ongoing support and the initial commitment. Do not create dozens of scenarios that leave the underlying evidence unchanged. Show which assumption needs validation before the next investment.
For option C to reach a two-year undiscounted cash break-even under the same simplified timing, annual avoided expenditure must be SAR 100,000: SAR 40,000 annual running cost plus SAR 120,000 divided by two. That is a threshold to investigate, not a forecast. If the business cannot identify a credible route to that amount, a cash-only case does not justify this option over that horizon.
Also inspect the downside if the trial fails. Which setup costs are recoverable? Can prepared information be used for process repair? What commitments continue after cancellation? A feasibility study should target the assumption most likely to change the choice, rather than merely prove that a model can draft a report.
Include value that is important but not yet monetised
A more complete report, reduced frustration or improved traceability may matter even where a reliable cash estimate is unavailable. Define the outcome, baseline and evidence needed. Keep it alongside the financial case with its own weight in the decision. Do not label an unmeasured improvement as guaranteed revenue or treat every avoided error as a prevented major incident.
Record potential disbenefits too: review burden, weaker staff understanding, customer confusion or dependence on a difficult-to-replace supplier. Explain how the proposed scope controls those effects and which remain uncertain. Leadership can approve a non-financial objective, but it should do so consciously and understand the cost.
Turn the case into a funding instrument
The approval pack should identify the preferred option and alternatives, financial assumptions, non-financial outcomes, downside, benefit owners and conditions for the next stage. Ask finance to challenge the cash classification and operations to confirm the adoption and redeployment assumptions. Assign someone to update the case when evidence changes.
Separate advisory acceptance from software acceptance. An accepted options analysis does not prove production readiness; consulting versus implementation explains that boundary. When seeking senior approval, use the board presentation guide to make the required decision and uncertainty legible.
Open for AI’s agentic AI consulting can help frame use cases, readiness and a sequenced investment decision. To discuss an AI transformation business case, bring a redacted workload baseline, the credible alternatives and the budget decision you need to support. Confirm which financial analysis and evidence outputs are included in the engagement.
FAQ
Your questions, answered.
Should time saved be recorded as a financial saving?
Record it first as released capacity. It becomes a cash effect only through a credible action such as avoiding planned expenditure, reducing paid temporary support or absorbing work that would otherwise require additional spending. If the same hours support a cash estimate, do not monetise them again as a separate benefit. Keep the operational measure and the financial consequence linked but distinct.
How do we build an AI business case before reliable benefit data exists?
Fund the smallest useful investigation, with a budget boundary and a question that could change the decision. Use ranges for provisional assumptions and identify their evidence owners. Do not present an exploratory estimate as an approved return. The initial case can justify learning; production funding should depend on a stronger understanding of workload, adoption, review effort and actual financial consequences.
Can an AI project be approved with a negative cash case?
It can be considered for an explicit non-financial or strategic objective, provided the decision maker understands the cost, alternatives and evidence. Separate that rationale from claims of cash payback. For example, improved traceability may justify expenditure under a defined objective, but its value should not be invented as a revenue figure merely to make the financial model positive.
Should the business case include staff time spent on implementation?
Yes. Show the required internal effort even when it does not create an additional payroll payment. It consumes capacity that could support other work and can constrain delivery. Separate incremental cash cost from internal opportunity cost so finance can interpret the model correctly. Include process-owner time, review, training, data preparation and transition work rather than only supplier invoices.
What is the best sensitivity test for an AI investment?
Test variables that could change the preferred option or the funding decision. That often includes complete-task effort after review, eligible volume, adoption and ongoing operating cost. Calculate the threshold needed for the case to work, then ask whether the evidence supports it. A large scenario table is less useful than a clear explanation of the uncertain assumption on which approval depends.
Who should own benefits after the transformation is delivered?
The business leader who controls the operating change should own the relevant benefit, with finance agreeing its measurement. A programme manager can coordinate reporting, and suppliers can contribute evidence, but neither should be assumed to control staffing, sales conversion or customer adoption. Name the owner and the required action before approval, then review the benefit after delivery against the agreed baseline.