AI strategy needs more structure. The 3 circles might be the answer
I feel like we should create something of an AI swear jar where every time it gets mentioned the pot bills builds by yet another £1. Not least because I think I'd be very rich by now. But as much as there's a danger of saturating the market with discussion about AI, I still believe the truth is that many organisations still haven't successfully deployed a true AI strategy.
This was further cemented for me when the Boston Consulting Group shared their findings that 60% of businesses have not generated the financial return expected from AI investments and only 5% have scaled their use of AI to give them a competitive advantage.
I’m not sure this was really a huge surprise to me though. The exact numbers might have been a wake-up call, but who can we really point to as an example of someone who’s got their approach to AI 100% right?
Let’s take Duolingo. In 2024 they piloted a brave approach to AI automation with predicted workforce reductions of 10% as a result. The following year users criticised the app’s new approach to such a degree that Duolingo began rehiring. The approach was a complete 180 and allegedly led to staff questioning “are we using AI because it helps or just to say we’re using it?”
Another high-profile example is Klarna. They invested in AI agents to take on customer service queries, with 2024 seeing stats included the AI agent handling as much as two thirds of all customer service queries – reportedly the equivalent of 700 full-time humans whilst also reducing resolution time from 11 minutes to just 2 minutes. You can see why this may have been perceived as a highly successful approach. However, the AI agent may have excelled at basic tasks, but for a financial technology company it failed to deliver on the complex and emotionally charged disputes leading to reduced customer satisfaction and compliance risk hallucinations. Klarna began to hire more human agents again in 2026 with the company’s CEO Siemiatkowski stating that "Cost unfortunately seems to have been a too predominant evaluation factor when organizing this. What you end up having is lower quality.”
But Klarna and Duolingo are advanced and successful businesses. They will not have chosen their approaches to AI lightly. So, what can be done to be more robust in deciding the best course of action?
The three-circle approach provides the most impactful topline solution I’ve come across. It combines the business’ overarching goals, with a specific AI strategy before moving onto specific actions. Whilst not failsafe, it does include a useful framework for triangulating different areas before committing to actions.
Business goals: Duolingo’s organisational mission is to “develop the best education in the world and make it universally available”. Yet it strikes me that the AI strategy they deployed moved them away from delivering the best education. You could argue that with reduced costs maybe they could have made the app more universally available, but their chosen actions just didn’t correlate to their business goals. But they’re not alone. IBM, yet another highly successful company, used an AI tool as part of their drive to build a reputation for excellence in the healthcare field. However, their chosen tool was not trained on a wide group of real-life patient data files, but instead a limited number of hypothetical situations. The tools did not deliver the excellence intended and the approach was scrapped at an estimated cost of $62billion.
The hype and speed of AI must surely be a factor here, but the very basics of strategy being founded in the overarching goals of a business is a must. Before deciding on any future actions, I would suggest it is vital to group together and ensure the business goals are clear, fit for purpose and known and understood by everyone involved in a new AI strategy.
AI Strategy: Once the organisation’s direction is crystal clear, the AI strategy can emerge from it. This can avoid the use of AI for AI’s sake – and let’s be clear that a strategy can be as much about when not to use AI as it is when to use it. Having clarity on what you’re looking to achieve before making the decision about how AI can help is logical and can prevent being caught up in the hype of new AI tools.
But it still may not be that simple. Looking at the example above with IBM, perhaps this could have been a good strategy if the tool genuinely added value. An AI strategy often needs to consider what to develop or who to partner with to be able to get this right – and that is not an easy decision. In their 2026 “Global AI confessions report”, Dataiku explored the tensions and challenges leaders face when considering AI. One confession states: “I’m not afraid of investing in AI, I’m afraid of investing in the wrong vendors”. 65% of those surveyed worried about over-investing in a crowded AI provider marketplace.
So, what’s the answer? I might be a leadership trainer, but I don’t think it’s my bias talking when I say more training. More awareness in how to assess your needs and options before making a decision. Without a long history of AI success stats to review, leaders need to implement the most robust decision-making frameworks they can to make the best choices. Skills such as critical thinking, data storytelling and marketplace analysis need to be continually honed to make decision-making as strong as possible.
AI actions: This is the stage where your plans now hit the workforce at large – and getting your people on board can make your strategy a success. But equally, it can break it if it’s not handled well. It’s estimated that as many as 70% of change initiatives fail in business, and that’s before we add on the emotionally charged reaction to AI that could see resistance and non-compliance be even higher than usual.
Communication is more important than ever. It’s not just the messaging about strategy that people need to hear, but the human quality of why they are still valued and how they can work with AI and not be replaced by it. As Bernard Marr says, "Even the smartest AI strategy will fail without buy-in. Leaders need to build narrative intelligence and tell stories that connect innovation to mission, values, and human benefit”. We need to ensure that we link the roll out plans of our AI strategies to speak to our employees on a distinctly human level.
An article by Harvard Business Review had an angle I thought was worth exploring here. Built on the insights that some people can navigate the ambiguity of technological change better than others, they propose appointing these people to roles of “Digital Wayfinders”. At first, I felt this was similar to the roles of “Champions” that we so often see in change practices, but the Digital Wayfarers go beyond this role - investigating and even advising senior leadership on the right way forward. Their natural curiosity is harnessed, and the doors are opened for them to explore. People from within, and across, the business who understand it’s needs and are given the freedom to do what they do best. Then, much like the classic “Champion” role they can support the wider roll out by gaining more buy-in than a top-down solution.
And like all change projects, this needs to be well planned. I saw a video recently that compared AI tool rollouts to gardening. If you throw a bunch of seeds on the soil, but don’t take the time to cultivate, water and nurture their growth you won’t get the garden you imagined. Much like releasing tools to the workforce needs commitment to monitor, iterate and ongoing communication to see the business benefit you seek. The toughest seeds might sprout, but you won’t get the success you anticipated without real effort.
The three-circles is not an exhaustive approach, but it does align to 3 really important principles of clarity, decision-making and practical actions - and for that reason I like it. I think it can be easily adapted based on different goals and scaled depending on what needs to be achieved. It strikes me that when we’re trying something new, we can often neglect the basics, and this is a reminder of the importance of these 3 areas and a prompt to put in place appropriate planned steps.
References
https://www.aimbusinessschool.edu.au/why-abs/blog/why-70-of-change-management-initiatives-fail
https://aimultiple.com/ai-fail
https://hbr.org/2026/10/empower-your-most-curious-employees-to-drive-your-digital-transformation
https://media-publications.bcg.com/The-Widening-AI-Value-Gap-October-2025.pdf
