Almost every senior leader we work with has been tasked with embedding AI more deeply within their working practices. But while there has been undoubted progress around the use of specific LLM tools (Claude/ChatGPT) the promise of more systematic progress is not - yet - as visible. Enterprises have poured an estimated $30–40 billion into generative AI, and yet 95% of pilots stall before they ever reach production (MIT NANDA, State of AI in Business 2025).
The result is that many capable leaders in well resourced companies are finding that there is a gap between what they set out to achieve, and what they see happening in practice. We call this the intention trap: the gap between what leaders intend to do and the results that they see.
This deeply human phenomenon is of course not unique to AI. If you’ve ever set a new year’s resolution, you will know that the intention you set on 1 January does not automatically translate into follow-through over the rest of the year. But just as it is possible to put in place a set of structures to help you follow through on your new year’s resolutions, there are steps that leaders in large organisations can take to ensure that their intentions get turned into concrete actions.
So for those who are all too familiar with the Intention Trap, we have put together three actions all senior leaders can do to help you move from intention to action.
1. Set your strategic vision for AI
Use the CogCo AI Integration Framework (below) to help you write down the purpose you would like AI to fulfil in your organisation
The companies that see real value in AI tend to share one trait: a clear, multi-year vision that someone at the top genuinely owns. And yet, according to one recent report, some 86% of large enterprises do not yet have a documented AI strategy (Human at the Helm of AI, 2026). In these organisations, only 22% of employees say that AI is meeting or exceeding their expectations - a figure which jumps to 66% when a company does have an AI strategy (Thomson Reuters, Future of Professionals, 2026). We recognise that setting out on the journey of creating a years-long strategy can seem daunting, especially in a world in which the technology might be changing by the day. So to help this process, we have found that it is useful to start with a simple question: how much decision-making and control do you want your AI systems to have?
And to help answer this question, we have developed the CogCo ‘Levels of AI Integration’ Framework, which helps to clarify and visualise what different integration levels might look like in practice. Do you want to maintain human decision-making at all levels (Level 1); or do you want to move to a world in which no manual (human) input is required to execute tasks, which are delivered fully by AI agents (Level 5)?
Answering this question will of course depend on lots of different considerations (from how willing you are to trust and rely on agentic outputs to your team’s current or future capabilities). But we have found that undertaking this exercise is a very useful way of supporting the work required to set your longer term strategy.
2. Model the behaviour you want to see
Demonstrate the (AI Adoption) behaviours you want to see your team embracing.
AI is fertile ground for overconfidence bias, a phenomenon where our perceived capabilities run ahead of our actual competence. Across six experiments, people who repeatedly watched someone perform a skill grew steadily more confident they could do it themselves, while their actual performance didn't improve at all (Kardas & O'Brien, 2018).
We have found that these findings are especially pertinent in the context of AI tools. Because the people pushing for adoption are rarely the people who then have to make the tools work in their daily job. The judgement sits with the sponsor; the effort sits with the team. So a leader who has seen an AI demo work, rather than worked with it, will likely underestimate what adoption actually asks of the people doing it.
Closing that distance is what marks out the organisations getting real value from AI. In one recent business review, organisations were three times more likely to be judged to be AI high performers when respondents strongly agreed that their senior leaders demonstrate ownership of AI initiatives and are actively engaged in using AI (McKinsey, 2025).
To help you move from the sidelines into the driver’s seat, focus on three behaviours. Which of these are you currently doing, and where could you be doing more?
- Model for your team how you would like AI to be used, rather than endorsing it from a distance. Where this is beyond your skillset, engage with your team about their needs, and about where the tools genuinely help. Use this to inform decisions on where you invest.
- Take the time to learn how to use the tools on real work rather than in a demonstration: among the CEOs BCG identifies as furthest ahead, when it comes to effectively leading AI adoption, two-thirds spend more than six hours a week upskilling themselves.
3. Understand your team and tailor the message
Understand where your team is at on the AI adoption curve
Employees are not a single audience. One of the most widely used models for understanding technology adoption in workplaces found people fall anywhere between Pioneers on one end, who strongly engage with AI to Avoiders who will resist the implementation of new technology completely.
The latest research suggests that the largest group of all in relation to AI adoption are the Sceptics, who are worried about how trust-worthy or reliable AI outputs will be. One recent study, for example, the proportion of people in the US who said that they were more concerned than excited about AI use grew from 37% in 2021 to 50% in 2025 (Pew Research).
The key learning for senior leaders here is to start with the assumption that you will have people across your organisation who are in very different places when it comes to AI adoption. And that it is very likely that a significant minority of your employees might be sceptical about what the future vision might imply for them and the way that they work.
This is something that CogCo has done a lot of work on recently, in particular through our BengoAI platform, which enables organisations to create sophisticated segmentations of their employee base. Together with AI personas of each of these segments that you can have interactive conversations with. Building up an understanding of the different groups within your organisation enables you to think about how you might take forward initiatives that might work with different types of employees. Pioneers are likely to be energised by newness and by the chance to lead the way. Hesitators may need proof it already works for people like them. Sceptics may want to see how this adds value to their work, and be told why they should care. Your communication is what decides whether each group sees AI as a threat or a genuine benefit.
The order matters as much as the content. Pioneers are already with you and they can become the proof the Hesitators need. Sceptics are your largest group and your biggest opportunity. Avoiders are the slowest to shift, and starting there drains the effort you need elsewhere. Deciding where to spend your attention first is what turns a tailored message into a change in how the organisation actually works.
A closing thought
These three actions - setting a strategic vision; modelling the behaviour you want to see; and understanding your employee base - together aim to tackle the most frequent sticking points organisations face when trying to step out of the Intention Trap. The questions may not be easy to answer right away, but we hope they serve as a starting point and provide you with a human-focused approach to tackling the AI challenge.
