Your AI Pilot Is Not Failing. It Is Dipping.

The San Antonio Spurs launched their AI rollout and six months in, only 14% of the organisation was actually using it.

If that was your business, you’d probably kill the project. Pull the plug. Tell the board it didn’t work. Go back to the way things were and write it off as a failed experiment.

The Spurs didn’t do that.

They pushed through. And now they’re saving 1,800 hours a month — every month — with 85% of their people actively using AI across the business.

Same tool. Same technology. Completely different outcome. The difference wasn’t the software.

Here’s what most businesses get wrong about AI adoption.

They treat a dip as a verdict.

Kamil Banc at AI Adopters Club puts it bluntly: “The absence of earnings lift is often a predicted phase, not proof the strategy is broken.”

There’s a concept called the productivity J-curve. It explains something that makes total sense once you see it, but kills AI initiatives every day because most leaders don’t know it exists.

Phase 1: Adoption. You get access. People experiment. There are early wins. Things feel promising.

Phase 2: The dip. Costs and friction go up as the organisation builds what AI actually needs to work. Workflows need redesigning. People need role-specific training. Data infrastructure needs sorting. Governance frameworks need building. Productivity drops. Leaders panic.

Phase 3: Harvest. The complements compound and scale. Productivity accelerates. You build what Banc calls a compounding advantage.

Most companies stop at Phase 2.

They experience the friction, see productivity stall, and conclude AI is overhyped. They were never going to see the harvest because they bailed before they got there.

And here’s the thing — the numbers are genuinely shocking.

78% of organisations have deployed AI tools. 6% report meaningful impact on earnings. 95% of AI pilots show no measurable P&L impact.

Not 40%. Not 60%. Ninety-five percent.

That’s not an AI problem. That’s a change management problem. And the data from Stanford, McKinsey, and MIT backs this up: the organisations in that 6% aren’t using different tools. They’re managing the transition differently.

McKinsey is direct about it: invest twice as much in change management as in building the solution. In practice, 80 to 90 cents in every rand of AI spend goes to technology and data. The people side — the training, the workflow redesign, the governance, the culture shift — gets what’s left over.

And then everyone wonders why their AI pilot isn’t moving the needle.

So back to the Spurs.

They’re the first NBA team to deploy ChatGPT Enterprise org-wide. The rollout was led by Charlie Kurian, their Director of Innovation and Strategy, and Jordan Kolosey, VP of Business Strategy, Innovation and Data Operations.

They didn’t start with a company-wide mandate. They started with 150 people.

Before anyone was required to use anything, they ran an in-person pilot, built custom onboarding guides, ran an internal AI hackathon, and let people find use cases that actually mattered to their jobs. No usage quotas. No forced adoption metrics. No “you will log in three times a week or explain yourself.”

Here’s the thing — earlier AI tool rollouts at the Spurs had failed. Because they required selling. ChatGPT was intuitive enough that people started requesting access before it was offered to them.

Honest question: when last did your team request more of a software tool before you’d finished rolling it out?

They also invested hard in the complement layer — the stuff that most organisations skip because it’s not as exciting as buying the technology. Workflow redesign. Role-specific training. Governance frameworks. They even built a “Culture GPT” trained on 50 episodes of an internal docuseries, so staff could embed the organisation’s values into everyday communications.

That’s not a product decision. That’s a people decision.

And it’s exactly why 14% became 85%.

So what does this mean for you?

The J-curve dip is not avoidable. But how long you sit in it — and whether you ever climb out — comes down to three things.

Fund the complements, not just the tool. Workflow redesign, role-specific training, governance, and data plumbing aren’t optional extras. They’re the investments that get you from the dip to the harvest. Skip them and you don’t avoid the dip — you just never exit it. Most AI budgets don’t account for this. They should.

Pilot before you mandate. A 150-person pilot isn’t hedging. It’s building the internal case before you ask the whole organisation to change how it works. It surfaces real use cases, creates peer advocates, and gives you the evidence to push through when the dip hits and someone in the boardroom suggests pulling the plug.

Measure fluency, not installs. Most organisations track deployment. Seats purchased. Licences activated. The Spurs tracked fluency — how deeply and consistently people were actually using the tools. That’s the number that tells you whether the change is sticking. Installs tell you nothing.

95% of organisations are sitting in the dip right now, calling it a failed strategy.

The Spurs cleared it in six months.

The technology was never the problem.

If your team is in the dip — or you’re about to start a rollout and want to avoid it — get in touch. This is exactly the work we do.