AI’s Tornado Effect

Vodafone, Ocado, Deliveroo and the NHS on what AI really means for their industries, and for your sales pitch

AI has been described as a tornado, tearing through industries at speed, leaving both transformation and turbulence in its wake. At Boardwave Live, leaders from Vodafone, Ocado, Deliveroo and the NHS shared what the storm looks like from the inside, and what vendors hoping to sell into their organisations should know.

From groceries to global robotics

James Matthews, Deputy CEO at Ocado Group, was quick to reframe perceptions.“Most people in the room probably think of Ocado as vans and groceries,” he said. “But the majority of the group is selling software (primarily) but also robotics, to other retailers around the world. Even when we’re doing robotics, we are doing software and using AI; it sits at the top of the robotics stack and enables it to do what it can do.”

The examples are far from theoretical. In one facility alone, Ocado’s AI-powered robots perform “a million picks a week,” James explained, half of all the food handled there. The challenge, he said, is packing a bag of groceries that has never been arranged in that particular way before. That task would be impossible with traditional programmed robotics. But generative AI and vision systems are making it viable. “We are enabling robotics to do things that simply were not possible before,” he said.

The NHS: scale and stakes

For Richard Meddings, former Chair at NHS England, the stakes are even higher. “I’ve been in an adjacent room to a surgeon removing a tumour,” he said. “The tumour is scanned, cast into pattern recognition, and before the patient is closed up, AI helps assess whether all the cancerous material has been removed.”

The examples tumbled out: AI in brain scans, breast screening, dermatology, ophthalmology, ambulance routing, and home-care monitoring. Each is designed to improve precision, reduce unnecessary admissions, or free clinicians’ time. But Richard was clear-eyed about the challenge. “It is an ecosystem, not a single organisation: 220 trusts, 6,400 GP practices, 13 ambulance services. The legacy complexity is enormous.”

That scale makes architecture and standards as important as individual breakthroughs.

Take ambient documentation: voice-to-text that links into appointment systems, prescriptions and patient records. “The danger is lots of CTOs just buy dictation tools,” Richard said. “What we need is voice-to-text that integrates with underlying systems.”

Vodafone: 600 models and counting

At Vodafone, AI is woven into the fabric of operations. “We have got over 600 AI models already working on top of our data ocean - the fifth largest in the world, at 27 petabytes,” said Amanda Jobbins, Global CMO & Commercial Director at Vodafone Business.

The priorities fall into three buckets: productivity inside the company, better products and services for customers, and support for employees themselves. Customer service is one focus. “We have deployed customer care super agents: co-pilots that help agents resolve complex network issues faster. It has massively improved first-time resolution.”

The ROI is beginning to show in marketing too. “Translation used to take two weeks. Now it is two hours. We can translate into 60 languages, reduce production times, and improve open rates. That is a huge return.”

Deliveroo: experimenting at speed

Carlo Mocci, Chief Business Officer at Deliveroo emphasised that AI had been core from the start. “We process well over a million orders daily and our logistics are fully automated. Whether it is designing the best routes or matching orders to riders, AI powers the system.”

Personalisation is another frontier. “Two customers sitting 100 metres apart may see entirely different app experiences. It is as if we are running hundreds of thousands of hyperlocal stores.”

Deliveroo was also an early adopter of generative AI, but Carlo warned against inflated expectations. “We saw an opportunity in coding, account management, and customer service. But not every experiment moves the needle. In fact, I read that 95 per cent of corporate generative AI pilots fail to generate measurable ROI. That rings true. The difference is building frameworks that let employees experiment safely, at scale. Many of our best ideas came from frontline staff.”

Flexibility, he added, is key. Procurement cycles are shortening. “We can now onboard a vendor in a week. We want optionality, because nobody knows what AI will look like in 12 to 18 months.”

Lessons for the pitch

For vendors, the panel’s message was blunt. If you are selling AI into enterprises, do not pitch hype. Pitch solutions.

James at Ocado: “We have got engineers who are hungry to experiment. We are running 15 or 16 trials right now. But you need to show how your tool solves a specific problem.”

Carlo at Deliveroo agreed: “Talk to business leaders, not just tech departments. Our CMO led our adoption of AI in marketing tools. It was not procurement-led. And we expect contracts to be flexible.”

Amanda at Vodafone underscored the need for co-creation: “With our scale, you are not going to sell us an off-the-shelf AI product. It has to be bespoke, working closely with our teams. Sometimes that means co-developing.”

Richard was perhaps the most cautious. “The NHS has 44,000 IT platforms. Overpromising is dangerous. What we need is connectivity of underlying systems. That is where entrepreneurs can help.”

A tornado, but with direction

AI is reshaping industries from grocery fulfilment to cancer care. The mood from the buyers was pragmatic: excitement tempered by the hard realities of scale, legacy, and ROI. Carlo at Deliveroo reminded the room that “95 per cent of corporate generative AI pilots fail to generate measurable return on investment” - a figure that resonates across industries. Vodafone described running steering committees and endless pilots to test what sticks. The NHS, with its 44,000 IT systems, warned against hype and urged vendors to focus on integration.

Yet for all the caution, the panelists were unanimous. This is not a passing storm. The tornado is real, and vendors who want to ride it need to prove that their AI solves problems buyers are facing today.

The uncertainty is also the opportunity. As James at Ocado put it: “We have got 15 or 16 different trials going on right now. For now, let 1000 flowers bloom, and then we will narrow it down to what really works.”

For AI companies pitching into buyers like these, that sentiment is both a warning and an invitation. The field is wide open, but only those who can turn experiments into tangible results will survive the storm.

You can watch the full session here.