AI ROI Archives | Thomas Anglero

A Pass-Fail Kpi Quietly Executes Every Idea It Was Not Built To See. Quote From Thomas Anglero, Strategic Ai Advisor. — Thomas Anglero on the AI value that pass-fail KPIs destroy.

Only about 12 per cent of CEOs say AI has delivered both cost savings and revenue gains. The usual response is to study that 12 per cent. The more useful question is why the other projects, the ones marked as failures, are quietly throwing away their most valuable output.

The 12 per cent is the easy part

Getting a return from an AI project is often not the hard thing it looks like. A company running its operations inefficiently will find that almost any competent AI project surfaces flaws worth fixing: process, supply chain, warehouse, pricing, contracts. That is where a good share of the 12 per cent comes from. It is real, and it is the floor, not the ceiling.

The bigger question is how you grow that number across every company, and the answer is not in the technology. It is in what leaders are not looking for. This is the same misdirection I wrote about in The Question Your Organisation is Asking About AI is Probably Wrong.

The spin-offs of innovation

On the way to a project’s stated goal, the people doing the work have epiphany moments. I call them spin-offs of innovation. A team sets out to cut licensing costs and, along the way, sees a new revenue model, a fix for a problem in another division, a better way to price, an idea nobody had scoped. Those moments are worth as much as the visible target, sometimes more.

That is where you grow the 12 per cent: by looking for the invisible spin-offs as deliberately as the headline goal, and by giving them permission and funding to become something. It is the opposite of the low-ambition approach I criticised in AI Leadership Failure: Set Low AI Goals, Automate the Obvious, Fail Quietly.

Why pass-fail measurement quietly destroys the value

Here is the honest version of the problem, and I want to be careful not to overclaim. I cannot tell you that the projects marked as failures were secretly full of breakthroughs, because by definition nobody counted. That is exactly the point. When you measure a project as pass or fail against a single goal, you guarantee that every discovery outside that goal goes uncounted, unfunded and unremembered. You will never know what was lost, which is its own kind of loss.

The KPI becomes the executioner. An idea gets surfaced, and because it is not the one thing being measured, it is quietly killed. So the fix is not a better dashboard for goal A. It is to make the question itself broader: what else did this project uncover, and what will we do with it? That is where the real return, and the real bonuses, should come from. It is the same reason value leaks long before the P&L moves, which I set out in The P&L Moves Only After the Leader Does the Work, and why the phase everyone wants to skip is the one that matters, in Into the Chasm and Back.

The 12 per cent is what got measured. The far larger number is everything a pass-fail scorecard threw away before anyone thought to look.

If you are leading your organisation through this, I work with a limited number of senior leaders each quarter. Get in touch at Anglero.com.

Thomas Anglero is a Strategic AI Advisor, keynote speaker and author of Intro to Artificial Intelligence. He has delivered over 450 keynotes across 30 countries for organisations including IBM, the WHO, the World Government Summit and the European Commission. He founded the IBM Watson AI Lab for Cancer at the Oslo Cancer Cluster and closed over $500 million in enterprise transformation deals as CTO and Chief Innovation Officer at Cognizant.

Frequently asked questions

Why do only about 12 per cent of AI projects deliver both cost and revenue gains?

Reaching a single stated goal is often the easy part. The low number reflects how projects are measured, not what they could produce. Most of the real value shows up as discoveries outside the goal that no one counts.

What are spin-offs of innovation?

The unplanned discoveries a team makes on the way to a project’s stated goal, such as a new revenue model, a fix for another division, or a better pricing approach. They are often worth as much as the headline target.

How does measurement destroy AI value?

Measuring a project as pass or fail against one goal guarantees that every discovery outside that goal goes uncounted, unfunded and unremembered. The KPI becomes the executioner of ideas it was never set up to see.

How do you increase the 12 per cent?

Look for the invisible spin-offs as deliberately as the headline goal, and give them permission and funding to become something. Make the question broader: what else did this project uncover, and what will we do with it?

The disappointing return on AI in most companies is not a failure of the technology. It is a failure of leadership that released a budget and never said where to dig. The companies now seeing a return are not the ones that spent the most. They are the ones whose leaders decided, before a single token was spent, which problem the AI was there to solve.

You approved the budget. You may have approved a large one. And now, sitting in front of a board that wants to see the return, you have a dashboard of usage and very little else to point to. If that is the quiet position you are in, this is for you, because the reason is not the one you have been told.

The gold rush nobody learned to mine

There is a picture I keep returning to: the gold rush. Everyone wanted in. So everyone bought a shovel. What almost no one did was learn to be a miner. A shovel does not find gold. You do not dig a hole in your back garden and strike a vein. You study the ground, you find where the gold actually runs, you buy the right piece of land, and then the shovel matters. Dig anywhere else and all you have is a hole and a sore back.

That is precisely what happened across the corporate world. Leadership bought the shovels. The budget went up, the tokens were purchased, the memo went out: everyone use AI, we are subsidising the tools, start digging. What was missing was the only thing that mattered. No one was told where to dig. No one was shown how to use the tool. No one was given the problem worth solving. An unlimited budget of tokens pointed at nothing in particular is not a strategy. It is a slush fund with a fashionable name.

You measured the tool. You never measured the leadership.

This is where the conversation about measurement turns on its head. Everyone wants to talk about putting the right measures around AI use. The harder question is the one no one asks: what were the measures on the leadership? Because a budget was spent, a great deal of money was spent, and no target was set for the people who spent it. The tool was measured. The leader was not.

Consider where the accountability actually sits. The finance chief is there to protect the company’s money. The chief executive is there to protect the company overall. The technology chief is meant to hold the strategy and the plan. The information chief is meant to guard how the organisation’s time and resources are used. When a year of spending produces a dashboard and no return, that is not the tool underperforming. That is four seats at the top that set no goal, protected no outcome, and held no one, including themselves, to a result.

The generation that never had to earn a return

There is a deeper reason this keeps happening, and it is uncomfortable. A whole generation of senior leaders built their careers in an era of nearly free money. For decades, capital cost almost nothing. You could finance a project on borrowed money at close to zero and never feel the weight of it. In that world, inefficiency hides easily, and you can spend your way past a problem rather than solve it. AI removed that cover. It asks a leader to know exactly where the value is and to go and get it, and a leader who has never had to earn a return in that way is suddenly exposed. This is the same failure I keep naming: set low goals and automate the obvious and you have bought maintenance, not transformation.

The leaders getting a return did the unglamorous work first

The leaders who are getting the return did something unglamorous first. They looked honestly at their own operation, found where it was actually broken, and pointed the AI there. And here is the part most miss entirely: every AI project is a culture project before it is a technology one. The tool does not transform a company. The decision about where to aim it does, and that decision is leadership work that cannot be delegated. The most common failure I see is a proud leader who handed the project to a likeable person and stepped back to admire it from a distance. That is not delegation. It is abdication.

This is the moment to move, not retreat

None of this means the money is lost or the moment has passed. AI budgets are being cut across the market right now, and most leaders are reading that as proof the technology failed. It is the opposite. It is proof it was deployed without direction, and it is the clearest signal yet that the leaders who move now, with a plan and a target, will pull away from the ones still digging holes at random. The uncomfortable truth is that the partners you have relied on to guide this were often the ones selling you shovels by the hour.

So when the board asks why the AI spending has not produced a return, the honest answer is not that the technology disappointed. The honest answer is harder, and it is the beginning of getting it right. You did not lead your AI project. You just bought more shovels.

Frequently asked questions

Why do most AI projects fail to deliver a return?

Most AI projects fail because leadership released a budget without deciding which specific problem the AI was meant to solve. The technology performs; the direction was missing. Companies that see a return identify where the value actually is before spending, rather than funding broad, unfocused use.

Is cutting the AI budget the right response to poor results?

Cutting the AI budget treats a leadership failure as a technology failure. Poor results usually mean the tool was deployed without a target, not that AI cannot deliver. The leaders who move now, with a clear problem and a measurable goal, tend to pull ahead of those who retreat.

Who is responsible when an AI investment produces no measurable value?

Accountability sits with leadership, not the tool. When spending produces a usage dashboard and no return, it reflects a leadership team that set no goal, protected no outcome, and measured the tool instead of measuring its own decisions.

What separates companies that get value from AI from those that do not?

The companies that get value did the unglamorous work first: an honest look at where their operation is genuinely broken, then pointing AI at that problem. It is a culture and leadership decision before it is a technology one, and it cannot be delegated.

If you are leading your organisation through this, I work with a limited number of senior leaders each quarter.

A Leader Reviewing Financial Figures Alongside An Ai Dashboard — Old data and old questions protect the losses. New questions move the P&L.

Once a leader has done the hard, honest work on themselves and the company, the P&L turns into simple arithmetic. The value leaks long before that, in companies that point AI at their old data and ask it the same comfortable questions that produced the losses in the first place. Change the questions and the numbers follow.

Where the financial value leaks

Most companies feed AI the old data, what the company has always said it is, and ask it to confirm the story they already believe. That old data is your current P&L, and usually a great deal of the L, which is why you are looking at AI to rescue you. Asking a powerful tool to defend a false picture of the business moves nothing. It just dresses up the same wrong answers. This is the same reason the honest reckoning has to come before any AI strategy.

The sequence that actually moves the P&L

The investment differs for every company and every group inside it, but the architecture is the same.

First, the data. Not only the old data, but new data, found by asking the right questions about what the company actually is, where it failed, and where it is really going. That core picture will usually be ugly, and the forecast uglier. That is the point.

Second, the strategy, built on top of that honest picture, with AI used as a tool and not as a saving light. A tool, nothing more.

Third, the right people alongside the AI, and then implementation, which might be a handful of people or thousands of AI agents, depending on what you are trying to achieve. It is the same architecture as the three phases of a real AI project.

The most important step is the first. The new data comes from new people asking questions the company has never asked itself, which is why who you put in charge of the work matters so much, and none of it happens unless a leader is willing to go down a painful path and see that they were not the leader they believed they were. That is why the inner work comes first. Skip it, and you are back to old data and old questions, which is exactly where the losses came from.

When the numbers move

There is no fixed timeline. Depending on the ambition, profitability can arrive in days, in months, or over a few years. The workflow is always the same; only the scale of the goal changes. What does not change is the order: the leader does the honest work, the real data surfaces, the strategy is built on it, and only then does the P&L start to move. The numbers are the easy part. The work that unlocks them is not.

Frequently asked questions

Why does AI fail to improve the P&L in most companies? Because they point AI at their old data and ask it the same comfortable questions that produced the losses in the first place. That old data is your current P&L, usually a great deal of the L. Asking a powerful tool to defend a false picture of the business moves nothing; it just dresses up the same wrong answers.

What is the sequence that actually moves the P&L? First the data, both old and new, found by asking honest questions about what the company really is and where it failed. Second the strategy, built on that honest picture with AI used only as a tool. Third the right people alongside the AI, then implementation. The order does not change; only the scale of the goal does.

Which step matters most? The first. The new data comes from new people asking questions the company has never asked itself, and none of it happens unless the leader is willing to go down a painful path and see they were not the leader they believed they were. That is why the inner work comes first.

How long does it take for the numbers to move? There is no fixed timeline. Depending on ambition, profitability can arrive in days, in months, or over a few years. The workflow is always the same; only the scale of the goal changes. What does not change is the order: honest work first, real data next, strategy on top, and only then the P&L.

Why is the inner work of the leader a financial issue and not just a personal one? Because skipping it sends you back to old data and old questions, which is exactly where the losses came from. The numbers are the easy part once the honest work is done; the work that unlocks them is the hard part, and it starts with the leader.

Advice to start AI with mundane tasks and lowered KPIs sets leaders up to fail quietly. Automating the known is maintenance with better tooling, not transformation. The purpose of AI at executive level is to surface what you have missed entirely: new revenue models, new market positions, and the structural weaknesses nobody is incentivised to find.

Why “start with mundane tasks” is dangerous advice

If the last article you read about AI told you to start with mundane tasks and redefine your KPIs downward, you are being set up to fail. Quietly. Politely. With dashboards that prove you achieved exactly what you aimed for, which was nothing worth achieving.

A recent piece in Fortune offered this guidance to leaders: focus on automating mundane tasks for immediate productivity gains, then redefine your success metrics as AI changes how value is created.

On the surface, it sounds reasonable. That is what makes it dangerous.

Focusing on mundane tasks means spending your AI budget on things you already understand. You are automating the known. You are optimising what exists. And you are calling it transformation.

It is not transformation. It is maintenance with better tooling.

What AI is actually for at executive level

The purpose of AI at the executive level is not to do what you already do, faster. It is to surface what you have missed entirely. New revenue models you have not considered. Market positions you did not know were available. Structural weaknesses in your governance, your processes, your assumptions, the ones nobody in your organisation is incentivised to find.

That is where the value sits. And mundane task automation will never take you there.

Why lowering your KPIs is the worse advice

The second piece of advice, redefine your KPIs to match what AI can deliver, is worse. It is asking you to lower the bar so you can clear it. To assign metrics that can be achieved and then declare success.

But if AI is doing what it should, your existing KPIs are the wrong KPIs. The new markets, the new models, the new processes, you have never measured them before because you have never seen them before. You cannot put a KPI around something you did not know existed until AI showed it to you.

Setting achievable KPIs around AI adoption is how leaders fail without realising they have failed. The numbers look fine. The dashboards are green. And the organisation falls further behind every quarter. It is the same reason AI is a culture project before it is a technology project.

Why you need minds that challenge your assumptions

I work with a team of AI advisors, the sharpest minds I can assemble, and we examined this article from every angle. Not one of them found a single piece of advice worth following.

That is not an attack on the publication. It is a warning to you. Be careful what you read. Be careful what you believe. Because I know you are in a position where you need answers, and the wrong answers delivered with confidence are more dangerous than no answers at all.

Surround yourself with minds that challenge your assumptions, the way my AI advisors and I together challenged this article. That is how you lead through this.

Frequently asked questions

Why is “start AI with mundane tasks” bad advice for leaders? Because it spends your AI budget on things you already understand. You automate the known, optimise what exists, and call it transformation, but it is maintenance with better tooling. It sounds reasonable, and that is exactly what makes it dangerous: it produces green dashboards while the organisation falls further behind.

What should AI actually do at the executive level? Surface what you have missed entirely: new revenue models you have not considered, market positions you did not know were available, and structural weaknesses in your governance, processes and assumptions that nobody is incentivised to find. That is where the value sits, and mundane task automation will never take you there.

Why is redefining KPIs downward a mistake? Because it asks you to lower the bar so you can clear it. If AI is doing what it should, your existing KPIs are the wrong ones, since the new markets and models have never been measured before. You cannot put a KPI around something you did not know existed until AI showed it to you.

How does a leader avoid failing quietly with AI? Surround yourself with minds that challenge your assumptions rather than confirm them. The wrong answers delivered with confidence are more dangerous than no answers at all, so pressure-test the advice you read before you act on it, and measure against what AI reveals rather than what you can comfortably achieve.

If you are leading your organisation through this, I work with a small number of senior leaders each quarter. Get in touch at Anglero.com.

AI adoption is rising across every industry, but return on investment stays flat. The bottleneck is not the technology. It is leadership: most executives treat AI as a tool to delegate to, rather than a capability to build around. Three changes turn AI use into measurable business value, and all three start with the leader, not the IT department.

The hard truth about AI adoption and business value

AI adoption is increasing across industries, yet the business value reflected on monthly spreadsheets remains stagnant. The hard truth for organisations facing this disconnect is that the bottleneck sits at the leadership level. Extracting real ROI from artificial intelligence requires executives to transition from passive observers to active AI leaders. You cannot expect a workforce to revolutionise their output if the leadership team still views generative AI as a simple text editor rather than a powerful strategic engine.

Stop typing, start talking: AI as your C-level advisor

To generate massive business value, leaders must fundamentally change how they interact with AI platforms like ChatGPT, Gemini, or Claude. You must treat your AI as a top-tier, $10,000-a-day strategic consultant available to you 24/7. Stop typing simple queries and start using voice to feed it complex, real-world business challenges. Whether analysing declining client trust or modelling market expansions, your AI can synthesise global data sets, process historical trends, and deliver predictive strategic options in 25 minutes, a task that would traditionally take a team of expensive consultants three months.

Flex your AI leadership to set the standard

Driving organisational adoption requires demonstrating the technology’s power in real-time. Leaders must “flex” their AI usage openly to set a new baseline for operational speed and efficiency. If you can command an AI to instantly cross-reference calendars and book 1-on-1 meetings during a live team call, you immediately raise the performance bar for everyone in the room. Combine this visible leadership with an internal AI champion who has the freedom to troubleshoot and train employees, and your company will rapidly transition from mere AI adoption to generating compounding business value. This is why AI is a culture project before it is a technology project.

Frequently asked questions

Why is my company’s AI implementation not generating ROI? The gap between AI adoption and business value stems from a lack of top-down AI leadership. Executives must stop treating AI as a typing tool for grammar and start utilising it as a 24/7 C-level strategic advisor to extract actionable, high-level business intelligence.

How do leaders drive AI adoption in their teams? Leaders must actively “flex” their AI usage in front of their teams. By demonstrating real-time AI capabilities, such as executing complex scheduling or data analysis live during a meeting, executives set a new operational baseline that forces the organisation to elevate its standards.

How should a CEO use generative AI? A CEO should use AI through voice interaction as a strategic sounding board. Instead of basic prompts, executives should feed the AI complex business scenarios, client data, and market variables to generate predictive charts, strategic options, and deep-dive analyses in minutes rather than months.

Generative AI is transforming workplaces faster than previous technologies. 23% of U.S. workers use GenAI tools weekly, leading to significant productivity gains. The result is time savings of at least four hours per week, primarily in tasks like writing and data analysis. Executives like the CIOs must balance innovation with workforce modernisation and ethical AI usage to maximise GenAI benefits.

The GenAI Revolution in the Workplace

The rapid adoption of generative AI tools represents a fundamental shift in how work gets done. Unlike previous technological advances that often required years of integration, GenAI is being embraced at unprecedented speeds due to its accessibility and immediate impact. Companies that foster an environment where these tools are properly utilised gain a significant competitive advantage.

Employees who leverage GenAI report not only completing routine tasks more efficiently but also having more time for creative and strategic thinking. This shift allows teams to focus on higher-value activities while automating the mundane. However, this transition requires thoughtful leadership to ensure the technology serves both organisational goals and employee development.

Five Executive Strategies for GenAI Adoption

Showcase Employee AI Innovation: Highlight AI success stories during every town hall or employee newsletter. This visibility normalises AI usage and inspires others to explore similar applications.
Stay Informed Through Regular Dialogue: Have employees explain their AI utilisation regularly. This keeps leadership current with evolving workplace practices and demonstrates genuine interest in employee innovation.
Recognise Individual and Team Achievements: Praise specific employees or teams by name for their AI implementations. Focus on improved outcomes rather than cost savings to maintain positive associations with the technology.
Enable Peer-to-Peer Learning: Encourage employees to conduct Masterclasses for colleagues interested in AI adoption. This creates an organic knowledge-sharing ecosystem that accelerates organisation-wide capabilities.
Champion Cultural Evolution: Embrace the inevitable changes in company culture rather than resisting them. Leaders who position themselves as AI champions set the tone for positive adoption throughout the organisation.

By implementing these strategies, executives create an environment where generative AI becomes a catalyst for growth rather than a source of uncertainty. The most successful organisations will be those where leadership actively guides this technological transition while empowering employees to innovate within appropriate frameworks.

Frequently asked questions

How much time does generative AI actually save employees?

Research indicates at least four hours per week for workers using GenAI tools, concentrated in tasks like writing and data analysis. With 23% of U.S. workers already using these tools weekly, the aggregate productivity gain across an organisation is substantial.

Why is generative AI being adopted faster than previous technologies?

Because of its accessibility and immediate impact. Unlike earlier advances that required years of integration, GenAI delivers value almost at once and needs no specialist training to begin, so employees embrace it at unprecedented speed.

What are the executive strategies for maximising GenAI benefits?

Five: showcase employee AI innovation publicly, stay informed through regular dialogue about how staff use it, recognise individuals and teams by name for outcomes rather than cost savings, enable peer-to-peer learning through masterclasses, and champion the cultural evolution rather than resisting it.

Why should recognition focus on outcomes rather than cost savings?

Because framing AI around cost savings invites fear of job cuts, while framing it around improved outcomes keeps the association positive. Employees adopt the technology more readily when leadership celebrates what it helps them achieve, not what it helps the company cut.