We’ve all seen the headlines promising that AI will revolutionise our organisations overnight. But having spent the last 18 months running adoption programmes across multinational businesses, I have found that the distance between installing a tool and actually changing how people work is significant. There are dozens of factors to consider, but here are the top eight lessons I’ve learned that truly move the needle on AI adoption.
1. Data-driven targeting is non-negotiable
Don’t guess where AI adoption is happening; look at the data. By using actual usage metrics, you can identify specific departments or cohorts that need the most support. This clarity allows you to create targeted campaigns rather than broad, generic communications. Crucially, you must secure buy-in from leadership within these specific areas to ensure these interventions land effectively.
2. Real AI adoption comes at the use-case level
Having a foundational knowledge of AI principles and being able to prompt effectively are important skills for all employees, but it’ll only get them so far. The true value of AI comes from applying it to specific time-consuming processes that would benefit from the optimisation and automation capabilities of AI tools. A change manager, despite their many virtues, does not know these processes - that’s where your local subject matter experts (SMEs) are needed. Give them the AI skills and the mandate to apply this technology in the processes and pain points they know inside and out. I saw this with one champion and their sales team, and their performance hugely outstripped other sales teams’ output, because they worked hand in glove with AI.
3. Leadership must be active, not just performative
We often see leaders championing AI externally while their own usage remains low. This is often because the use cases for senior executives, who may not be doing as much hands-on content creation or data analysis, aren't as obvious as they are for junior staff. One effective way to bridge this gap is through mentoring. By pairing advanced AI users with members of the executive team, you create authentic storytelling opportunities. When leaders can give real-world examples of how the tool helped them, it resonates far more than a corporate mandate.
4. Build accountability into your champion network
A network of energetic AI users is vital, but their enthusiasm can wane if they lack direction. To keep your champions engaged, each group should be accountable for a specific adoption plan, reporting regularly on their impact. Since this work is often 'side-of-desk', we recommend implementing 'monthly themes', focusing on a single, manageable topic like prompt engineering or data analysis. If you have the budget, consider ring-fencing capacity for dedicated staff whose sole responsibility is to drive adoption within their specific division, geography, or function.
5. Balance progress with platform stability
Your partnership with technical and product teams is the heartbeat of your adoption strategy. The AI landscape moves fast, and it is tempting for product teams to focus purely on the next big update. However, constant shifting of priorities can leave the core platform unstable or confusing. The maxim should be: March confidently forward with one eye on the horizon. You must deliver on your near-term roadmap to provide certainty, but you must also balance new features with the maintenance of existing tools. If you neglect the platform’s performance, you lose the trust of your users, particularly those who were resistant to AI in the first place. Once they are gone, winning them back is incredibly difficult.
6. Know when to move from rollout to business as usual
How do you know when a project-oriented drive should shift to Business As Usual (BAU)? The answer lies in the data plateau. Monitor your metrics; once your usage data stabilises, you have likely reached the ceiling of what is possible for that specific adoption phase. You may find that 50% of your users are highly active, while the remaining 50% are occasional or inactive. Rather than spending excessive resources trying to convert the final 20% who may have found alternative workflows and tooling, focus on a 'continual drumbeat' of support. This includes regular learning sessions, global campaigns, and ongoing champion support to ensure adoption remains at the level you have achieved.
7. AI adoption is an operating-model problem, not a tooling problem
Organisations often have access to powerful AI models yet still struggle to share data, redesign workflows, or scale successful pilots beyond individual teams. For change managers, this reinforces that a tool rollout without workflow, governance, and cross-team knowledge-sharing redesign will stall. The change plan needs to target significant areas of the operating model, not just tool training.
8. Rising stakeholder expectations need to be managed explicitly
Stakeholders are now expecting more scope and depth from the same teams, faster analysis, clearer ROI evidence, and greater transparency about where productivity gains actually come from. This is a direct change management cue: expectation-setting and transparent benefits communication need to be built into any AI adoption programme, especially in environments where there is significant siloed data and uneven infrastructure, which will see much slower, uneven progress.
Sustained AI adoption needs people, process and proof
Most organisations are currently grappling with the same question: we have invested heavily in AI tooling, but how do we know if people are using it to its full potential? A comprehensive, data-backed change programme is the only way to ensure you get the outcomes you were promised, working with leadership, champions and SMEs to make real change in specific use cases at the local level. Define the effects of AI on your operating model, and ensure the business is clear on the value to expect and when.
If you would like to discuss any of these areas or share your own experiences with AI adoption, please get in touch.
