From ‘Human In The Loop’ To ‘Human In The Lead’: Three Shifts That Change AI Adoption

​There’s a cartoon that circulates in enterprise circles. Someone asks: “What do we want?” AI. “When do we want it?” Now. “When will it happen?” We don’t know.

Every leadership team meeting I’ve walked into across Southeast Asia has some version of this problem. The CEO is energized, the board is aligned and the vendor is signed.

Six months later, the pilot is dead. Nobody announces it. It just stops while we reach for the usual explanations: Bad data. Wrong model. Poor vendor fit.

After eight years deploying AI across some of the largest consumer enterprises in the region, I’ve come to a different conclusion: The failure often lives inside a phrase that sounds perfectly reasonable: “human in the loop.”

The Trouble With ‘Human In The Loop’​

Think about who actually handles the transaction at a business, like the procurement manager dealing with thousands of supplier negotiations, and the cashier at a retail counter who is reading a customer in real time.

These people compose the operating machinery of any enterprise, and their daily decisions determine whether an AI investment produces returns. When you tell them they’re “in the loop,” here’s what they hear: You are a checkpoint. You may be temporary.

The doom narrative around AI doesn’t help. Every headline about job displacement lands on the shop floor as a threat. The person at the counter might go slower than necessary, not because they’re obstructionist, but because it’s rational self-preservation.​

There is a proverb in Malayalam that captures this well, Vadi Koduthu adi medichu, loosely translated as: Give someone a stick, and they will beat you with it. Enterprises that push AI onto front-line workers without reframing the relationship shouldn’t be surprised when those workers subtly ensure the pilot never quite delivers.

In my experience, “human in the lead” is a better framing. In this model, the AI exists to extend human judgment, not replace it. The person closest to the decision determines what gets acted on.

Here is what that takes in practice:

1. Change the conversation before the rollout starts.

Most AI implementations fail on day one, not at deployment. The framing used in the first team communication sets the adoption trajectory for everything that follows.

If the message is “we’re introducing AI to improve efficiency,” front-line workers will likely hear “we’re measuring you.” If the message is “this system is here to help you perform better and earn more,” the dynamic shifts.

At a retail entertainment company in Southeast Asia, cashiers were given access to a performance monitoring system that tracked their own quality metrics daily. No manager pushed them. No targets were handed down. They logged in every morning, reviewed their own numbers and began self-correcting.

Within months, team performance improved, overall store sales moved up, and individual incentive payouts increased. Some of the top performers saw significantly more. Nobody beat them with a stick. The data was theirs, the improvement was theirs, and the reward was theirs.​

2. Connect AI output directly to individual incentives.

This is where the framework becomes tangible.

Consider a cashier at the point-of-sale terminal of a retail entertainment company. A loyalty card is tapped. In that moment, an AI system surfaces two things: the customer’s value tier and their most frequently purchased category. The cashier reads the room, mood, pace, and tone and decides whether to suggest a relevant offer or simply complete the transaction. That decision stays with the person standing at the counter.

When both signals align the AI’s data and the cashier’s read of the customer, upsell conversion improves. Sales go up. Incentives go up. The cashier comes back the next day wanting to use the system again because they saw the outcome in their paycheck.

That is what self-driven adoption looks like, and it starts with making the benefit personal and immediate.​

3. Give the human the time to actually lead.

The third shift is about removing the work that crowds out judgment.

At a large commodity business in Southeast Asia, a procurement team of 12 was handling 2,500 purchase requests a month. Each request typically contained six line items, and each line item required a minimum of three vendor quotes.​

The workflow was: Pull the purchase request from the ERP, identify vendors, draft emails manually, wait for quotes, copy figures into a comparison template, analyze and decide. Roughly 80% of the team’s time was consumed by that coordination sequence, work that required accuracy but no commercial thinking.

When AI took over the sourcing, drafting and data-consolidation steps, the procurement team’s time shifted to what they were actually hired to do: negotiate. With more time in conversations with suppliers, they saw a reduction in procurement costs.

The AI didn’t negotiate. The humans did. The AI just cleared the path.

This is the operating principle behind the human in the lead model. The AI handles volume, repetition and data assembly. The human handles judgment, relationships and decisions. Neither replaces the other because neither can do the other’s job.

​Conclusion

Think of fire. Uncontrolled, it destroys. Once directed, it powered the engines that took us to the moon. AI sits at exactly that inflection point, where leaders must show their teams that they are the ones holding the torch, not standing in AI’s path.

Leaders who keep framing AI as something deployed on their workforce will keep getting the same result: compliant pilots, quiet resistance and rollouts that look good in board decks and produce nothing on the P&L.

The shift isn’t complicated. When the cashier, the buyer, the person at the last mile believes the AI is there to help them win, adoption becomes in their self-interest. And self-interest, in any organization, is the most reliable engine there is.