Ravi Rabheru: AI Adoption Starts With the Customer's Job to Be Done

Ravi Rabheru has a blunt read on AI adoption: most companies stall because they have not defined what value means for their customers. Rabheru, an independent AI advisor and member of the Forbes Technology Council who previously worked at Sainsbury's, argues that AI only takes hold when a team starts from the job its end customer is trying to get done and works backwards to the tool.

He made the case on Conveo's podcast, recorded live at Cannes Lions 2026. His examples run from cucumbers to supercars, and they land on a point insights and marketing leaders will recognize. The companies that get value from AI are the ones that understand why their customers, internal and external, behave the way they do.

Adoption follows value, and every team defines value differently

Rabheru starts with a warning about keynote stages. The demos people see at conferences are the north star, he says, and they typically run in demo environments. Real businesses exist to make money, and in his words even nonprofits need revenue so they can give back to the communities they serve. Any technology has to drive toward an outcome, and typically that outcome is revenue.

The trap is equating value with return on investment. ROI is money in, money out, with a multiplier. Value depends on who you ask.

"But value for a HR person is very different to a marketing person and value is different to a person that's working in supply chain."

Ravi Rabheru, Independent AI Advisor, Forbes Technology Council

If an organization can build a weighted average of value across its lines of business, he says, it has a strategy and can work backwards from outcomes. He has a specific way of getting there. Senior leaders join, plus people two or three levels below them, while skipping their direct reports, "because the minus ones will please the bosses." Customers come into the room too. Each group describes what value creation looks like for its part of the business, and the use cases land on a value matrix: how long it will take, what it depends on, and how much effort it needs, plotted on a two by two.

That approach also answers a question he gets about AI centers of excellence. Not every team generates revenue directly. Sales and marketing win most ROI-based budget fights because they physically bring in the money, yet they depend on product teams shipping features fast enough to sell. Value, in his description, ripples through the whole organization.

Value is defined by what reaches the end customer

Rabheru's clearest example comes from grocery retail. After Sainsbury's, he advised other retailers and pushed them back to basics: a supermarket's job is to get food onto shelves so people can take it home and eat it.

"People don't go and buy more cucumbers on a weekly basis because you've delivered a tech strategy where we've solved the data problems."

Ravi Rabheru, Independent AI Advisor, Forbes Technology Council

Cucumber demand stays fairly flat across a lifetime. You might buy more for a party or once you have kids. So the technology has to serve the cost and experience of selling cucumbers. "That's where you get adoption because you're thinking about the end customer and what they're doing," he says, and that means asking why people go into the supermarket in the first place.

He says he lives and breathes the jobs to be done framework, combined with Simon Sinek's Start With Why. A job to be done is the thing a person is actually trying to achieve, plus the reason behind it. Someone creating social media content has a job. The useful question is why they are creating it. Product strategies built on that question, he argues, are more successful. He sees the same failure in banking, where teams set out to transform the organization with data projects and never ask how the end customer will adopt the service.

Generative AI is a chance to redesign processes, and most teams only squeeze them

Rabheru sees two uses for AI. The first is acceleration, which he calls the basics. The second matters more. Generative AI understands natural language, so in his words, "It's not a data scientist workflow anymore." Anyone can prompt it. That opens a door to rethinking how work gets done, because, as he puts it, "a lot of processes for legacy organizations were designed 5 10 15 20 years ago."

Neobanks such as Revolut, he says, redesigned how people interact with their bank because they started from the consumer's job to be done. Most legacy teams take an existing process, trim the admin and call it transformation. That works in many cases. He would still start from the end point and ask what the new process should look like.

He is skeptical of the slogans that travel around this shift. Asked about mantras like "SaaS is dead," he answers: "My view on this is some of it's a lot of marketing fluff." An agent is still software. If a company replaces a CRM subscription with a set of connected agents and MCP servers, the money moves from the CRM vendor to the model provider. The vendor relationship changes, the spend may not.

Experimentation is welcome. Rabheru supports putting AI tools in everyone's hands so they understand the capability. He calls the current moment a honeymoon period, and says it ends when budgets meet reality.

"So you can only do certain number of use cases in a year. Even if you've got 200, you can only do maybe 10."

Ravi Rabheru, Independent AI Advisor, Forbes Technology Council

Companies also have to keep the lights on. Every shiny new tool forces a choice: stop a use case, ask for more budget, or fit the tool into the product strategy first. His advice is to ignore much of the market noise and ask what value a new tool will generate before deploying it. How to organize the effort, through a central center of excellence, communities of practice or experts seconded into product teams, depends on the appetite of the board and the culture of the organization. He has seen companies succeed with very different models.

Match the model to the job: a family car handles the school run

The mistake Rabheru sees most often right now is token spend. Teams default to the most capable model for every task. Once an agent is built properly, with guardrails, processing and skills, he says it can often run on a cheaper model such as Haiku or DeepSeek and reach the same outcome. Most of what he builds for himself runs on lower-cost models. His summary of the instruction he wants to give a model: "I don't need the model to think about things. I just need the model to just do the bloody thing."

He compares it to right-sizing cloud infrastructure, which companies are still doing years later.

"You don't need the best LLM for everything."

Ravi Rabheru, Independent AI Advisor, Forbes Technology Council

His advice to "treat your AI consumption as it's your own money" connects back to value. Cheaper running costs widen the gap between what a use case costs and what it returns, which makes the value case more credible to the people funding it.

For insights and marketing leaders, the why lives with the customers you do not have

Asked for the fastest way he has come to understand a customer, Rabheru tells the story of a car manufacturer facing a European mandate to build an electric vehicle. A luxury supercar brand cannot launch a four-door sedan that drifts from its brand. To build something that fits, the team had to immerse itself with customers, which in practice meant a lot of supercar shows.

The key move was talking to people who bought something else. Cars are functional or emotional purchases, he says, and a Ferrari buyer will not switch to a Lamborghini, though they might buy a Range Rover. Loyal customers will not reveal that.

"If you're only speaking to people that buy your thing, they're already loyal to you. You've got to go outside and think outside the box."

Ravi Rabheru, Independent AI Advisor, Forbes Technology Council

"Unless you go and speak to the customers, particularly competitor customers, you're not going to figure that out," he adds. He also recommends asking customers who left why they stopped buying, and bringing in third parties when a brand cannot run those conversations itself. That is the basis of good competitive intelligence, and it applies whether the customer is a car buyer or an employee using an internal tool. "Don't forget your internal people. They are incredibly important," he says.

He also points to where measurement is heading. Likes and view counts have existed for decades. What brands usually miss is the emotion behind them: the conversation that created the connection and drove the reshare or the purchase. In his view, "that real sentiment analysis and that emotional analysis is really really important," and he expects it to drive more personalized content and advertising.

For insights teams, three practical points follow from the conversation:

• Define the value your research creates for each internal customer, from marketing to supply chain, before choosing tools.

• Keep evidence from customers, lapsed customers and competitors' customers in one place so teams can query it later. Our guide to customer intelligence management covers how to turn research into reusable infrastructure.

• Pair the numbers with the emotional reasons behind them, which is where voice of customer research earns its budget.

Noah Plaue made a related case about getting AI used at enterprise scale in his episode on enterprise AI adoption.

Adoption follows understanding

Rabheru's argument comes back to one discipline. Start with the value and the job to be done of the end customer, internal or external, and the technology choices get simpler: which use cases make the cut, which model runs them, and which tools to ignore.

That discipline depends on knowing why people do what they do, including the people who chose someone else. Conveo's AI research assistant runs video conversations with real consumers, so insights teams can get to the why behind behavior faster and bring it into the decisions where AI budgets get spent.