Programmes launch and then quietly stall. Training is delivered, the system goes live, and within two quarters people have returned to the previous way of working. This happens the same way in general operations, in people and HR teams, and anywhere the work itself changes. The investment is recorded as complete while the benefit never arrives, because adoption was assumed rather than measured and support ended at go-live.
What we usually hear
We trained everyone and they still use the old process.
The system went live, but the benefit never showed up in the numbers.
Our sponsor approved the business case and then went quiet.
Middle managers are quietly working around it.
We cannot actually say how many people changed how they work.
Two quarters later, people drifted back to the old way.
If two or three of these sound familiar, the work below is where we would start.
Key solutions
Identify who is affected, how much their work changes, and where resistance is most likely, before the programme commits to a plan.
Work alongside the operations, people and HR teams who have to change how they operate day to day, through go-live and well past it. This covers AI adoption as much as system or process adoption.
Equip the executive sponsor and line managers to lead the change themselves, which is the strongest predictor of whether it holds.
Framework
AI adoption
Most AI rollouts are treated as technology projects. They fail for the same reasons every other change fails, only faster, because the tool arrives before anyone has agreed how the work should change.
People still need to know why the tool is there, want to use it, know how, be able to apply it to real work, and be reinforced until it is habit. Giving someone a licence covers the third stage only.
Trust is the extra hurdle. People will quietly avoid a tool they do not trust, and you will not see it in the licence numbers. Usage data tells you who opened it, not who changed how they work.
We treat AI adoption as a change programme. Impact by role, clear rules on approved use, managers equipped first, and adoption measured against the work rather than against logins.
If you are rolling out AI tools and seeing licences go unused, that is an adoption problem, not a product problem.
Expected business outcomes
Proof points
Prosci has benchmarked change management effectiveness against project outcomes across a large sample of programmes.
The strongest predictor is how well the people side is managed.
Prosci benchmarked change management effectiveness against project results. Programs with excellent change management were seven times more likely to meet objectives than those with poor change management. The gap is larger than any technology variable.
Source: Prosci benchmarkingMost transformation programs do not deliver what was approved.
Seven in ten large-scale programs miss the goals set out in the business case. The money is spent and the program is closed, but the operating result the board signed off never arrives.
Source: McKinsey & Company7 of 10 programs fall short of their stated goals
Alan Mulally joined Ford as CEO in 2006 and introduced the One Ford plan, together with a weekly Business Plan Review in which every executive reported status against the same format. The mechanism that made it work was behavioural rather than structural. Leaders were required to report problems honestly, and the first executive to show a red status was publicly supported rather than punished, which changed what the rest of the team was willing to surface.
Ford was the only one of the three large United States automakers that did not take federal bailout funding during the 2008 and 2009 crisis. The case is widely taught because the change was carried by sponsor behaviour and a fixed governance cadence, which is what the last two ADKAR stages require.
Send a short brief describing the situation. We respond with an approach, the team who would deliver it, and the measures we report against.
Request a proposal →