Why AI Adoption Stalls — and What Pivotal Does Differently
Most advice about getting a team to use AI follows a familiar pattern: give people access to a tool, run some training, hand out a prompt library, and appoint internal champions. Rishi Chande, Managing Director at Pivotal, argues that this playbook often misses the real problem.
In an interview with Peekwire, Chande set out how the consultancy approaches AI operating systems work — and why starting with the work, rather than the training, tends to produce better results.
Start With the Work, Not the Training
Chande said the standard approach works when the challenge is simply that people don't know how to use a tool. But in his experience, that tends not to be the problem.
"The challenge is that a lot of times, AI sits separate to where people are doing their work, so you are asking people to change their habits for a benefit they can't see yet."
Pivotal takes what Chande describes as the opposite angle. Rather than beginning with training, the team starts with the work itself. They identify one task the team does every week — something tedious that people are happy to hand over — and rebuild it with AI built in.
"Once people see the benefit it can bring, the barrier to working in a 'new' way is gone, people ask what else it can be used for, and the momentum starts to build."
For readers facing resistance to new tools, the practical takeaway is to find a single recurring, low-stakes task and demonstrate value there before attempting a broader rollout.
What Low Usage Numbers Really Hide
When Pivotal is brought in because a team isn't using the AI tools it already has, Chande says a few themes recur — ones that the pages ranking for this search tend to leave out.
The first is that people often are using AI, just not through the company account. "They're worried that it looks like they are cutting corners, or they're not sure what they can use it for. So the usage numbers look low, but it's just happening on personal accounts, which is not ideal all round."
Second, staff haven't been told what they can or can't put into these tools. Chande is blunt about vague guidance: "saying 'use it responsibly' isn't really a guardrail."
Third is trust in the output, which he says depends on how the AI setup is configured and on balancing where AI adds value against where people need to apply their own judgement and experience.
"The common denominator in these three boils down to ownership of AI internally, and whether or not that clearly sits in the leadership team, and how the leadership team communicates their AI vision."
That points to a practical lesson for leaders: clear data rules and visible ownership matter more than another round of tool training.
The Cost of Measuring Success Too Quickly
Asked what companies decide too fast, Chande pointed to the definition of success itself — typically a measure of usage, such as logins or number of agents, which he characterises as a measure of output or volume.
"What this misses is whether anything actually got better by using AI, and that's where people start to question, 'Is this actually working for us?'"
The damage, he says, is that trying to win back the mindset of the organisation six months later is much harder.
Pivotal's view is to start with a few teams and a few real pieces of work, then measure time saved or improvements in quality and consistency, and let that shape the wider rollout. Chande sums up the approach as "more evolution, than revolution."
Practical Takeaways
- Lead with a real task. Pick one recurring, tedious piece of work and rebuild it with AI embedded, rather than opening with training.
- Make the rules explicit. Vague guidance like "use it responsibly" doesn't tell people what they can put into a tool.
- Look beyond login numbers. Low usage may reflect personal-account use, not disinterest.
- Name an owner. Whether AI ownership clearly sits with leadership shapes how the rest of the organisation behaves.
- Measure what improved. Time saved, quality and consistency are more useful signals than volume of output.
Chande's answers suggest that the hardest part of AI adoption is rarely the technology. It is the habits, guardrails and measures of success that surround it — and those are decisions leaders make long before a tool is switched on.