Only 25% of AI Projects Deliver Expected ROI — It's a Leadership Failure

Full text

For years, the central question around AI was whether it really works the way much of the market’s hype says it does.

That debate is settled, and we’ve seen what AI can do. The more important question now – and one most organizations still aren’t asking – is whether we’re working alongside it the right way.

That shift in framing matters. Budgets have been allocated, tools deployed, and pilot programs have graduated into full-scale production.

And yet, something is still missing.

IBM’s Institute for Business Value found that only 25% of AI initiatives have delivered their expected ROI, and just 16% have successfully scaled across the business despite years of investment and genuine enthusiasm for what the technology can do.

The problem isn’t AI itself; the bottleneck stems from what organizations have or have not built around it.

That’s a leadership problem, and fixing it will require more than buying better tools or scheduling more trainings.

Stop Investing in the Wrong Places

The instinct for most organizations has been to buy the latest platforms, stand up a few pilot programs, and bring in a vendor to train their workers. That approach addresses the surface-level challenge, but misses the greater underlying issue. The truth is the greatest barrier to AI maturity is the lack of investment in the human infrastructure needed to support it.

The companies seeing the strongest AI outcomes are rarely those with the most sophisticated or expensive models. They’re the ones that have fundamentally rethought how their people work. Among organizations that Boston Consulting Group designated as AI leaders, roughly 70% of resources went towards people and process changes, 20% to IT infrastructure , and only 10% to the AI models themselves. Most organizations have that ratio backwards.

When leaders become hyper-focused on deploying the right tools and launching the right uses cases, they neglect the organizational muscles that are essential to using AI responsibly and consistently. All the tools in the world won’t close that gap without the right training, guardrails, and policies to back them up. And building that support structure must be a leadership priority, not an afterthought.

The Productivity Gains Are Real, But Fragile

None of this is to say AI isn’t creating real value. It absolutely is – at least, for the companies using it well. But those gains are more fragile that many leaders realize. They evaporate when companies lack support for their employees across their interactions with the technology, or when they fail to clearly communicate where human judgement and critical thinking are still essential.

The data here is hard to ignore. Among employees who use AI on the job, less than 8% report receiving extensive training with their tools. And that number has barely budged despite a sharp increase in daily usage. Moreover, 60% say it often takes them longer to figure out how to accomplish a task with an AI tool than it does to simply do it themselves.

Companies are deploying AI faster than they are enabling people to use it, and in doing so, may be creating exactly the friction and confusion they were trying to eliminate.

What Leaders Owe Their People

This is where leadership has to show up differently. The gap between AI potential and AI reality isn’t going to close through procurement decisions or new rollout announcements. It closes with deliberate, ongoing investment in people. That means three things:

Focus trainings on people, not just tools: AI is evolving faster than training curriculums can keep up. Invest in training role-specific judgements, helping people understand where AI makes them faster and where it introduces challenges or risks.

Move beyond adoption rate metrics: If 80% of your organization is using AI and productivity is still flat or declining, adoption is the wrong metric. Measure time-to-completion on real work tasks and be honest about what you find. Some use cases that are slowing people down simply shouldn’t be using AI.

Stop treating AI policy as a compliance checkbox: Companies getting this right have built AI governance into how they plan and execute work daily, not appended into and acceptable use document. That means leaders who model where they use AI and where they don’t, and who are willing to explain why.

Making AI Potential a Reality

The technology is ready. But leaders need to be able to do more than allocate budget and monitor usage.

They need to decide when AI should and should not be used, what to rebuild rather than automate, and how to support their teams throughout all of it.

Those are the questions most organizations are still failing to ask, and until they do, the ROI gap isn’t going anywhere.

We've ranked the best HR software .

This article was produced as part of TechRadar Pro Perspectives , our channel to feature the best and brightest minds in the technology industry today.

The views expressed here are those of the author and are not necessarily those of TechRadarPro or Future plc. If you are interested in contributing find out more here: https://www.techradar.com/pro/perspectives-how-to-submit