Blog: Business case
What does your AI cost on a difficult day?
Published by Forever Foundry
The demonstration handles a straightforward request in seconds.
Your business also handles incomplete orders, disputed invoices and customers whose circumstances do not fit the standard process.
Those difficult cases belong in the business case.
AI costs change with the work
On 19 October 2026, Gartner published findings from its survey of 2,002 CIOs and technology executives, conducted between May and July.
While 58% reported increased pressure to deliver cost improvements through AI, 51% expected AI to increase the total cost of owning technology solutions. Gartner highlighted that AI costs can depend on what happens when a system runs, rather than simply on the initial purchase.
These are surveyed expectations, not measured cost increases across every organisation. Read Gartner’s findings.
The FOFO™ view
An average cost per request can hide the cases that consume most of the effort.
Imagine a logistics business introducing AI to resolve delivery queries.
A routine enquiry might need one lookup. A disputed delivery could require several searches, checks against different records, a human review and a second conversation with the customer.
A cheap first response does not necessarily mean a cheap resolution. Usage that grows with each case is also why an AI agent needs a spending limit.
Customer service knows which enquiries return. Operations knows which records are incomplete. Finance sees credits and compensation. IT understands the systems and usage charges involved.
Listening across those functions gives leadership a more credible picture of the investment. It is one reason your CFO should help choose the AI problem.
Test the awkward cases before committing
Build the assessment around a representative mix of routine requests, exceptions and peak demand. For each, record:
- Whether the problem was resolved correctly.
- Total AI and system usage costs.
- Human checking and correction time.
- Repeat contacts, escalations and rework.
- Time to resolution.
Compare the total cost per correctly resolved case with the current process. Test how the economics change when volumes rise or more cases need intervention.
That may support automating routine work, assisting people with complex cases, or fixing missing information first. The evidence should determine the approach.
AI Compass™ listens across the organisation to uncover friction points and exceptions. Slid3rs™ helps teams rank opportunities by business value and delivery complexity. FlowRoom™ supports leadership agreement on priorities and ownership.
A credible AI investment needs to work economically on the days when the work is difficult.
Related reading
Next