The AI Cost Savings That Didn’t Deliver As Promised.

Bain & Company surveyed 951 companies to understand exactly what AI actually delivered against what was promised in the boardroom. The results are stark: 40% of companies tracking their AI spending saw cost savings of less than 10%, despite targeting 11% to 20%. Only 4% hit savings above 30%.

Bain's own language for the finding was blunt — the technology worked, but the value didn't arrive.

Ninety per cent of the companies that missed their targets are increasing their AI budgets again next year regardless. Forty-four per cent are funding the next wave of spend directly from savings the last wave never actually produced. Bain called this what it is: a circular bet with a structural leak.

Read as a cost-accounting failure, this is an uncomfortable finding for any CFO. Read as a signal about where AI's value was always going to land, it's a far more useful one – and a far more interesting story for the companies willing to see it.

The Savings Were Never Going to Show Up as Savings

Every automation wave before this one followed the same shape. Robotic process automation, then machine learning, then generative AI — each promised a cost line that would shrink. Each time, the number moved less than projected, and each time, the explanation offered was execution: bad data, weak governance, and unclear ownership.

All of that is real. But it isn't the whole story. A cost-savings target assumes the value of AI shows up as headcount reduction or a smaller line item. For most knowledge work — and certainly for anything resembling communication, strategy, or judgment-based decision-making — that was never where the value was going to land. The value shows up as better decisions, faster responses, and work that previously didn't get done at all because there wasn't time or budget for it. None of that appears as a savings figure. It appears as growth, as reputation, as the capacity to do more with the same team — which is a different ledger than the one the original business case was built against.

Bain's own research on the companies actually pulling ahead makes this explicit: the winners aren't the ones who found better cost discipline. They're the ones who stopped measuring AI against cost and started measuring it against value produced at the enterprise level — better outcomes, not just fewer hours logged.

What This Means for a Romanian CFO

Romanian companies adopting AI over the past two years have largely followed the same playbook as their Western European counterparts: a business case built on efficiency, a target expressed as a percentage reduction, and a board expecting to see that number appear on the P&L within a year. Bain's data suggests that expectation was miscalibrated from the outset, and Romanian firms are not exempt from the pattern.

What is different locally is the opportunity this creates. A market still building its AI maturity has the chance to set the right expectation from day one, rather than unwind an overconfident one two years in. The Romanian companies that will look smart in three years won't be the ones that squeezed a cost line the hardest. They'll be the ones whose finance leadership asked a better question from the start: not "how much will this save us," but "what decisions, campaigns, or client outcomes will this let us produce that we couldn't produce before."

That reframing costs nothing and changes everything about how a board reads an AI investment's first-year results.

A Personal Observation

“I've sat in enough boardrooms across telecoms, technology, and financial services to recognise this pattern before Bain put a number on it. Every efficiency initiative I've been part of — long before AI — produced the same tension: the finance team wants a line item to shrink, and the operating teams know the real gain is somewhere the spreadsheet doesn't look. AI hasn't created this problem. It has simply made the gap between the two more visible and more expensive to get wrong, because the budgets involved are larger than any automation programme that came before it.” Steve Gardiner

The companies that will do well from here are the ones willing to tell their board a harder truth in year one: the savings may never arrive in the form promised, and that isn't the failure it looks like.

Where This Leaves the Communication Function

There's a version of this exact mistake happening inside communication and marketing teams right now, and it's worth naming directly. AI tools are being justified internally on the same basis — fewer hours on drafting, faster turnaround, smaller agency retainers — and boards will apply the same scrutiny Bain describes to those business cases too. A communications function that can only defend its AI spend in terms of hours saved will face the same uncomfortable review Bain is describing for automation generally.

The stronger case is the one Lighthouse PR builds with clients directly:

AI's role in strategic corporate communication and reputation management is not to cut the cost of producing content, but to widen what a communication function can credibly cover — more markets monitored, more scenarios prepared for, more precision in messaging under pressure — at a level a smaller unassisted team could never sustain.

That is a valuable case a board can actually evaluate honestly, rather than one that quietly fails the same test Bain just applied to the rest of the enterprise.

———

About the Author

Steve Gardiner (exec MBA) is a senior marketing and commercial leader at Lighthouse PR, bringing global experience from Accenture, Electronic Arts, Virgin Media, Telekom, and Etisalat. Latterly, as VP Business at Etisalat, he was responsible for $1.8B in revenue.

Today, Steve applies his strategic, marketing, and growth expertise to support Lighthouse PR clients as part of the agency’s service offering.

About Lighthouse PR

Lighthouse PR is a leading PR agency in Romania that works with a select number of organisations across Central and Southeastern Europe, delivering media relations, reputation management, crisis communications, social media and an extensive range of business growth and business continuity services — always led by senior practitioners.

We hold exclusive membership for Romania and the Republic of Moldova in both the Eurocom worldwide PR network and the CCNE, Europe's leading crisis communications network.

Lighthouse PR: Clear. Concise. Convincing.