As economic headwinds strengthen, Publicis Sapient global boss Nigel Vaz thinks AI can come into its own – but not via large language models. Small language models, he suggests, can deliver both cost out and growth in, cutting development time and compute overheads and bringing revenue rapidly through the door.

Large language models have grabbed headlines over the last year or two. “For me, that's at the level of a browser. The application layer that sits above those two is where actually most of our clients will derive value,” he told Mi3. Within that application layer, per Vaz, “you need a small language model … Can you build an architecture that has inherent understanding of a specific business problem?”

Flower power

He gives a client example – the UK’s Royal Horticultural Society. “It’s a couple of hundred years old now, and this work was actually presented to the late Queen,” per Vaz.

“They had a tonne of knowledge. But the problem they had was ‘how do we get people to understand how to take care of their plants?’” The solution was “a Shazam for horticulture”, pairing the society’s knowledge with a visual interface.

“You take a picture of the flower. It tells you what that flower is. But … very quickly it starts to tell you 'this flower has this deficiency or issue and here’s a link to our commerce platform that allows you to buy the things to address it',” says Vaz.

“That was incredible, because it harnessed all of the wealth that they had, and we built a small language model on top of a very focused set of applications in order to solve that problem.” But without heavy fees, cloud and compute costs. “There's huge choices that you can make which influence these outcomes pretty dramatically,” per Vaz.

Get out of jail

He gives another, intended to show that while the likes of Salesforce and Adobe are now pushing a similar line in fast-to-market low customisation agentic AI models, nothing is truly out of the box – leaving space for nimble swat teams, or “Navy SEALs”, as Vaz describes Sapient.

“We’ve worked with the Los Angeles County Public Defender’s Office, where the LA public defender had attempted five transformations in the past and not really succeeded,” says Vaz.

“This one did, because it was essentially [AI] leveraging all of the data and casework that a public defender needed to support a client – in fact, an Oscar-winning documentary maker made a film about this. It's called Forgiving Johnny.

“It was built on Salesforce tech in order to provide them the best way in which to defend this client, by connecting a lot of data points. Forgiving Johnny talks about this guy, Noah Cox, who was the attorney defending Johnny Reyes, who would have gone to jail for a long period of time had this platform not enabled Cox to connect the dots about the best way to defend Johnny. The fact was that Johnny actually had significant mental health challenges and developmental disabilities that meant that he needed treatment rather than prison time,” says Vaz.

Reyes was facing the prospect of up to 20 years in jail. Instead, he was moved into appropriate treatment – which would essentially never have happened if his lawyer had had to manually wade through a mountain of paper casework.

AI not oven ready

The LA County Public Defender’s Office example was built on Salesforce’s platform – with Salesforce now pushing hard to own agentic AI and low cost, fast to market products, and rival Adobe likewise touting new agentic wares.

“But even at Dreamforce, Salesforce is not saying ‘you are CMO X and you buy this licence and suddenly your marketing problems can be solved by your own people’. They are not saying that, because software was never built like that,” says Vaz.

“Think about consumer software. When's the last time you downloaded a new operating system on an iPhone, and then suddenly you realise, ‘it's going to take me three hours of configuration to actually make sure that my music is back on here and my data is transferred across every single app – and that's us as people and our personal tolerance,” says Vaz.

“The minute you put it in the context of an enterprise, that utopia has never been the case. The systems integration industry and the IT services industry have been implementing software sold by software companies for the best part of 50 years. Whether it's SAP, which is – quote, unquote – out of the box, or Oracle, or CIBO, or Clarify for CRM, you can go through that entire lifecycle and there's always been people who needed to implement the thing.

“Even the most automated of stacks today – something like a ServiceNow – requires a tonne of configuration and implementation expertise from companies.”

Hence he thinks the likes of Publicis Sapient and its ilk are facing less disruption than larger consulting competitors.

“Our model has historically positioned us as the Navy SEALs, if you will, in the context of the IT services landscape. We were never going to be the regular army with hundreds of thousands of people and large outsourcing deals. That was not how Sapient was built. We were built to solve some of the toughest problems our clients were facing, particularly in a digital context, so we would get in, do those things and get out,” per Vaz.

“Now what we're starting to see is actually not having as many people, and leveraging AI to take cost out from clients and drive growth, is creating real opportunity for us.”

Pressure rising for CPG, telco, retail

“Pretty much every one of the companies that we look at on a global basis is pressed significantly from a growth perspective,” says Vaz. Hence focusing on projects that can scale while delivering both cost out and growth. He cites consumer packaged goods retail and telco as examples.

“If we look at the CPG companies, we've seen massive increases in prices that consumers are facing, but they haven't actually made much progress in trying to get better penetration of their products across segments who aren't buying or where they're disintermediated, particularly in the context of younger buyers,” says Vaz.

“At the same time, they're seeing significant inefficiencies in supply chain, significant inefficiencies in the context of marketing, where companies are basically saying, ‘we need to fundamentally rethink our product portfolio, given macro pressures and disintermediation’, and so you're starting to see big transformation in the context of how they're spending marketing dollars, how they're allocating supply chain dollars to build a different idea,” he adds.

“The same is true of retail, where we've seen huge pressures, particularly in the in the western world, because that inflationary pressure has been passed down to consumers. Retailers are under even more pressure from a growth perspective. So we're seeing things like retail media networks suddenly take on a life of their own. They were nice to have a few years ago, but now they are source of cash that these companies didn't actually have before.”

Which changes the economics of retail – and the difference between retail media haves and have-nots will become more marked.

“Most of the cash that these businesses generate, with retail margins being where they are, most of that is allocated to be reinvested. In this case you either have something that you can actually invest to fund new ventures or you can actually deliver back to shareholders as profit,” says Vaz. “It gives you optionality.”

He says a similar dynamic applies to telco.

“We're starting to see the same things – huge pressures competitively, content becoming more expensive to produce and distribute, that content not yielding as much return, and we're seeing significant optimisation,” says Vaz.

“If you look at all of these businesses, you're seeing clear trends of trying to find new sources of growth, trying to take cost out of our business, and both of these are creating the impetus for transformation.

“The challenge you've got is the short-term pressures. So the biggest areas of growth from our perspective, is where we are able to do both of those things in a one-two punch.”

Australia: transformation lag?

Vaz thinks Australia offers plenty of opportunity in that regard, and sees an immediate prospect within the immigration department.

“My view is, I think Australia is further behind in many industries from where the rest of the world has got to already,” he says. Vaz's guess is a lag of perhaps five years – and in some cases, more.

“When I flew in, I was scrambling for a pen for the last 15 minutes of the flight. I’ve been to 17 countries in the three months since I was last in Australia, and have literally not had to fill out a paper form. It’s double-sided: They have a digital solution at the airport, but you still need paper. What’s going on?”