Banking

AI costs and ROI for banks

The cost of AI is not the licence fee. The cost of AI is the data work, the integration, the governance, the training, the maintenance, and the cost of getting it wrong. Most banks underestimate the total by a factor of three.

By Zakir Hoosen, Director, Fuzzelogic Solutions. Board-level guidance in plain English.

Every AI project has two costs. The cost you see on the proposal, and the cost that shows up six months later when the project is running, the data is messy, the integration is hard, and the business has not changed the way it works. A banking board that only looks at the first number is making a decision on incomplete information.

This guide covers what AI actually costs a bank, what the research says about returns, and the questions a board should ask before approving the budget.

The real cost of AI in banking

The vendor will tell you the licence cost. That is the number on the slide. It is also the smallest number in the total cost of ownership. The real cost includes data preparation, system integration, staff training, governance setup, ongoing maintenance, and the cost of disruption while the project runs.

In banking, there are additional costs. Regulatory reporting. Model validation. Audit trails. Customer communication when the system makes a decision that affects them. These are not optional. They are the price of operating in a regulated environment.

Two thirds of AI transformations do not deliver. The budget was spent. The return was not.

The 10-20-70 rule

Fuzzelogic uses BCG's 10-20-70 rule to frame the cost conversation with boards. It is a useful discipline.

1. 10% of the effort goes to algorithms. The model itself. The clever bit.
2. 20% of the effort goes to technology and data. The infrastructure, the data preparation, the integration.
3. 70% of the effort goes to process change. Changing how people work, how decisions are made, how the business operates.

Most AI budgets are built around the 10% and the 20%. Almost none budget for the 70%. That is where projects fail. The technology works. The data is ready. But the business does not change. The tool sits there, unused, while the old process continues.

In banking, the 70% is harder than in most sectors. Banking processes are regulated. Changing a decision process means changing the audit trail, the customer communication, the regulatory reporting. It is not a matter of sending an email and moving on.

What the research says about returns

The evidence on AI returns is mixed. Some organisations see real gains. Most do not.

That number should give every banking board pause. Eight in ten. Not a few stragglers. The majority. The problem is not the technology. The problem is that most organisations do not change the way they work to match the technology they bought.

Boards are rushing. The returns are not there. The gap between the promise and the delivery is the 70% that nobody budgeted for.

Questions a banking board should ask

Before approving an AI budget, ask five questions.

First, what is the total cost over three years, not one? Include data preparation, integration, governance, training, maintenance, and regulatory compliance. If the vendor cannot answer that, the proposal is not ready.

Second, what is the return, and how is it measured? Ask for the baseline. If the proposal says 30% improvement, improvement over what? If the baseline is not defined, the return is not measurable.

Third, who owns the return? Name the person who answers when the project does not deliver. Not a committee. A person.

Fourth, what is the cost of doing nothing? Sometimes the right answer is not to build. Sometimes the existing process is good enough, and the money is better spent elsewhere. A good consultant will tell you that.

Fifth, what happens if we stop? What is the exit cost? Can the data be extracted? Can the process revert? If the answer is no, you are locked in before you have started.

The honest version

Fuzzelogic is an Isle of Man firm that has spent nineteen years modernising banking, insurance, healthcare, retail, manufacturing, and government platforms. We have worked with nine regulated financial institutions. We tell boards what most consultants will not: the honest answer is sometimes that the cost of AI outweighs the benefit, and when that is the case, we put it in writing.

Your systems were built for a world before AI. Most can get there. We tell you which ones cannot.

Start with the assessment. Two to four weeks, fixed price, and you own the verdict and the roadmap whether or not we build any of it. When you are ready to talk AI, call Fuzzelogic Solutions and ask for Zak. www.FuzzelogicSolutions.com | info@FuzzelogicSolutions.com | +44 (0)1624 618950

Read next: AI readiness assessment for banks and AI governance for banking boards.