Most organisations are spending their AI budget in the wrong place. The data is clear on where the value comes from. The enterprises pulling ahead have figured this out.
Most enterprise AI programmes are well funded. Modern data platforms are in place. Pilots have run and shown promise.
And yet, when the board asks for a clear return, the honest answer is hard to give.
The technology is rarely the problem. The budget split is.
Boston Consulting Group estimates that around 70% of the value from generative AI comes from people, process and cultural change. Only 30% comes from the algorithms and the platforms that run them. Most enterprises are spending in the opposite ratio. That is the single biggest reason AI investment is not translating into AI returns.
This is the problem we see most often at Inovus. Capable teams. Real leadership commitment. Genuine funding. And AI programmes that stall before they deliver at scale.
Here is what is actually getting in the way, and what the highest performing organisations are doing differently.
BCG’s 10-20-70 framework says 70% of generative AI value comes from people, process and cultural change, not from the technology itself. Boston Consulting Group
When AI proofs of concept become science projects
There is a pattern that repeats across organisations of every size.
A proof of concept is built. It works. Leadership is encouraged. More investment follows. But then, somewhere between that early success and scaling across the business, the programme loses momentum.
The pilot keeps running. It gets refined and re-presented. But it never scales. It never delivers the return that was promised. What started as a proof of concept has quietly become a science project. Technically interesting. Carefully maintained. Not driving business outcomes.
This is not a failure of ambition. It is a structural problem. The proof of concept was built to prove the technology works. It was not built to prove the organisation can work differently. These are two very different challenges. Solving the first does not automatically solve the second.
The gap between individual AI wins and business-level returns is not a technology problem. It is a workflow design problem.
Why AI adoption is harder than cloud, ERP or mobile
Enterprises have absorbed big technology changes before. Cloud migrations. ERP rollouts. Mobile-first working. In each case, the core job stayed the same. People learned new systems and new interfaces. What they were fundamentally responsible for did not change much.
AI is different. It does not just add a new tool to an existing role. In many cases, it changes what the role is for.
This hits hardest for experienced professionals. People who have spent ten or fifteen years building deep expertise are now being asked to hand much of that work to AI, then redefine their contribution around supervising the output. That is a real professional and personal transition. It is not just a training exercise.
Employees whose managers actively support AI use are 8.7 times more likely to say AI has transformed how work gets done. Gallup
That number cannot be explained by which tools people have access to. It comes down to whether the people around them are actively helping the transition happen, or just standing back and hoping it does.
The organisations getting this right build it in deliberately. Internal AI ambassadors who help colleagues apply AI to their specific work. Change programmes running alongside the technical rollout rather than bolted on afterwards.
Workflow redesign is the single biggest driver of AI ROI
Of all the factors that determine how much value an organisation gets from AI, one stands out clearly.
Workflow redesign had the biggest impact on business results of the 25 factors studied in AI transformation programmes. McKinsey, 2024
Not the model. Not the platform. Not the quality of the data, though that matters. The biggest driver of financial return is whether organisations redesign their workflows, or simply put AI inside processes built for a different era.
Putting AI into a broken process does not fix it. It just makes it run faster. A workflow with weak steps in the middle becomes a faster version of the same flawed process. The enterprises getting real returns ask a different question. Not where can AI help inside this process, but what should this process look like if we built it from scratch today, with AI in from the start.
BBVA is a good example. Rather than rolling out tools and hoping people would adapt, the bank trained 250 senior leaders specifically on workflow redesign before scaling to the rest of the business. The result was 83% weekly active AI use across the bank. That came from leaders who understood what needed to change, not just which tools were available.
Putting AI into old workflows makes broken processes run faster. Rebuilding those workflows from scratch is what drives real returns.
Two AI governance risks most enterprises are not yet planning for
As AI takes on more work, and as individual tools become connected workflows, two governance risks appear that are not yet getting enough attention.
-
Over-reliance
As AI systems become more reliable, teams naturally trust them more. That is reasonable. The risk is when people start accepting AI outputs without applying their own judgement. The critical thinking that makes human contribution valuable can weaken over time if people are not actively engaged. Good governance needs to address this. Not just the risk of AI making mistakes, but the risk of humans not catching them.
-
Visibility
As workflows become more complex, AI systems will sometimes make decisions and hand off to other AI systems without a human reviewing each step. This creates blind spots. Outcomes reached through a chain of steps that no individual reviewed or approved. Enterprises need to design clear intervention points, decision transparency and human oversight into these workflows from the beginning.
For regulated industries such as financial services, insurance and healthcare, these are not future concerns. They are operational realities that need governance frameworks built around them now.
AI is a capacity investment, not a cost reduction tool
The most important reframe in AI transformation is this. AI is not primarily a cost reduction tool. It is a capacity investment.
There is a limit to how far you can reduce human involvement before the quality of output starts to drop. Cutting people and replacing them with AI, without redesigning what the remaining team does, produces short-term savings and long-term problems.
The enterprises getting the strongest returns are not the ones that cut the most. They are the ones that used AI to expand what their teams can do, then pointed that expanded capacity at growth.
The strongest examples we see follow the same pattern. Companies use AI to give their people more leverage rather than to cut headcount. They then point the freed capacity at growth. More client-facing time. Faster product cycles. Broader coverage. AI becomes a way to grow, not just a way to reduce costs.
Team structure also has to change. In AI-first organisations, the pattern shifts towards people who can oversee and direct more of a workflow. Routing. Deciding. Supervising. They bring in specialists for the moments where deep expertise genuinely matters. This is different from how most teams are organised today, where specialists own their end of a process and coordinators manage between them.
Helping people understand what their role looks like in this new model, and actively supporting them to make the shift, is one of the most important things an organisation can do to make AI work at scale.
How to measure AI transformation: what good actually looks like
Most enterprises measure AI adoption the wrong way. Licence usage. Login frequency. Time saved. These are easy to track. They are not what matters.
The organisations that get this right do something simple that most avoid. They rewrite job descriptions before they roll out AI, not after. A job description tells people what success looks like in their role. If it has not changed, the message is that the job has not changed. AI becomes a productivity tool rather than a fundamental shift in how work gets done.
Some of the largest technology companies have started factoring AI use into how they evaluate performance. But the more meaningful version of this is not measuring whether people are using AI. It is measuring whether the way people work is actually changing. Are workflows being rebuilt? Are business results improving? Are teams developing new ways of working that compound over time?
Projects with excellent change management are around seven times more likely to meet their objectives. Prosci
Seven times. That is what investing in the 70%, the people and the process, actually delivers. It is not a marginal gain. It is the difference between an AI programme that produces one good pilot and one that changes how the business operates.
Rebalancing your AI investment in practice
Rebalancing AI investment, spending more on people and process change and not just on technology, is a leadership decision with immediate consequences.
It means choosing delivery partners who understand both the technology and the operating model change required to make it work. Technical capability alone is not enough.
It means agreeing on business outcome metrics at the start, not usage metrics. What result is this programme meant to deliver? How will it be measured? What does success look like when the programme is finished?
And it means treating AI as a continuous journey. Building the internal capability to keep adapting as the technology develops, rather than treating each initiative as a one-off project.
The technology is the smallest part of the investment that needs to change. The operating model, the workflows and the people are where the return lives. The organisations that have understood this are already pulling ahead.
Talk to Inovus
At Inovus, we help enterprises unlearn pre-AI ways of working and deliver measurable outcomes through AI-first delivery. We build programmes around the ratio that actually drives value, investing in workflow redesign, people change and embedded governance alongside the technical work. We measure against business outcomes, not usage metrics.
If the gap between your AI investment and your AI return is where you are right now, we are happy to talk.
About the author
Filippo Di Pisa is CTO at Inovus, a data and AI consultancy backed by La Fosse Group, helping organisations unlearn pre-AI ways of working and deliver measurable outcomes through AI-first delivery.
Sources
Boston Consulting Group, AI Transformation Is a Workforce Transformation, 2026.
Boston Consulting Group, How Generative AI Is Transforming Business.
McKinsey & Company, The State of AI, 2025.
Gallup, Global Indicator: Artificial Intelligence.
Prosci, The Correlation Between Change Management and Project Success.
OpenAI, How BBVA is Scaling AI From Pilot to Practice Across the Org, 2025.