AI Delivered The Savings, But Can Your Business Sustain Them?

Direct Source Verification: This story is aggregated from Forbes (forbes.com). Full reporting rights and copyright belong to the primary publisher.
Kelvin Cheema is a Global CIO and transformation executive advising boards on technology, AI, transformation and enterprise value creation.

Kelvin Cheema is a Global CIO and transformation executive advising boards on technology, AI, transformation and enterprise value creation.

gettyImagine two businesses each reducing the cost of processing a transaction by 20%. Both report a legitimate improvement. In one, AI helps employees resolve difficult exceptions, captures specialist knowledge and strengthens the organization’s ability to improve. In the other, it performs work through which employees once developed judgement, leaving the business increasingly dependent on a supplier or a shrinking group of specialists.

The saving is identical. The future economics may not be. The figures are illustrative. The distinction is real.

I believe boards risk treating an immediate productivity gain as proof of sustainable enterprise value. Reducing the cost of today’s work matters. But it does not tell us whether the business is better equipped to deliver tomorrow’s performance. For that, we must examine what happens to the knowledge, judgement and operational capabilities behind the saving.

Research by Tania Babina, Alex He and Renhao Jiang associates AI investment with productivity growth through the accumulation of organizational capital: durable, firm-specific knowledge developed through learning by doing. Their findings suggest that AI can make an organization more capable rather than merely less labor-intensive.

That is the favorable outcome: lower costs accompanied by stronger capabilities that support future performance. But the same technology can change how those capabilities develop.

A controlled 12-week study involving more than 200 chess students offers a useful warning about the design of AI assistance. Students with additional help available on demand improved by 30%, compared with 64% among those receiving structured assistance. Chess is not corporate finance, and the experiment does not establish an effect on enterprise profitability. It illustrates a narrower point: Making assistance more convenient does not necessarily produce better independent learning.

Consider finance. Automating reconciliations can release capacity, strengthen controls and give teams more time to investigate anomalies. That is an attractive investment. But what if automation also removes the experience through which future finance leaders learn to recognize an unusual transaction, challenge an unexplained balance or understand why a control failed? Transactions may be processed faster while the organization becomes less able to investigate the exceptions that matter most.

I would not preserve inefficient manual work simply to train people. That would mistake a development method for a commercial objective. I would redesign how judgement is acquired, tested and retained while allowing automation to perform the work it does better.

The question is not whether AI should replace tasks. It is whether management has accounted for the capabilities those tasks once developed.

When AI changes how work is performed, it can also change where critical expertise resides and who controls it.

One business may embed knowledge in reliable workflows, accessible systems and employees who can improve them. Another may achieve the same cost reduction while becoming dependent on proprietary technology, external specialists or processes its people can no longer meaningfully challenge.

Neither arrangement is automatically superior. Buying external capability can be more economical and resilient than maintaining everything internally. But dependency has a price, even when it is absent from the original productivity calculation. What happens when a supplier changes its commercial terms? When an automated process fails under unfamiliar conditions? When the business must adapt a workflow that nobody internally fully understands?

These are not arguments against automation. They are questions about the cost of sustaining it. A serious investment case should account for switching costs, recovery capability and the consequences of losing control over critical knowledge.

This is where an operating model question becomes an investment question. Two businesses can report identical reductions in operating costs while warranting different assumptions about future margins and cash generation. One may have strengthened its ability to scale, adapt and resolve problems. The other may need recurring remediation, expensive supplier support or scarce specialist intervention to maintain its reported performance.

Those are different economic propositions.

A saving that creates reusable capability can support operating leverage. A saving that creates fragile dependencies may carry costs that emerge long after the initial benefit has been recognized.

The uncomfortable possibility is that both investments could pass the same financial approval process and subsequently be reported as equally successful. The business case captures the reduction in operating cost. The benefits dashboard confirms its delivery. Yet neither necessarily tests whether the organization has retained the ability to challenge, recover or adapt. That is a governance weakness, not merely a technology risk. Management may have delivered exactly what it promised while leaving the enterprise less capable of sustaining it.

Capability is not the only determinant of AI returns. Demand, implementation economics, reliability and competition also matter. But a board cannot judge the durability of a saving without understanding what the enterprise can still do, what it must buy and what it may struggle to recover.

A productivity dashboard measures what became cheaper. It does not necessarily reveal what became more valuable, more fragile or harder to replace. That is why I would challenge an AI business case that stops at hours released, headcount removed or cost per transaction. Can the enterprise still investigate exceptions, recover from failure, challenge its providers and develop the judgement required for decisions it remains accountable for?

AI can deliver exceptional operating leverage. It can also build organizational capabilities that compound beyond the original investment. But neither outcome follows automatically from a reported saving.

Before approving an AI investment, I would ask three questions: 1. What capability will this investment create, transfer or stop developing? 2. Where will that capability reside? 3. And what will it cost to sustain, recover or replace?

Productivity can be measured this quarter. The consequences of lost expertise, concentrated dependency or stronger institutional knowledge may emerge years later. The real test is not simply whether AI delivers the saving. It is whether the enterprise can sustain the value.​

Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?

Original Source
https://www.forbes.com/councils/forbestechcouncil/2026/11/18/ai-delivered-the-savings-but-can-your-business-sustain-them/
Visit Forbes ↗
SHARE STORY:
𝕏 f in

Related Coverage in Business