Episodes

Stephan Pretorius: The Agentic Agency

Stephan Pretorius, CTO of WPP, on how AI is rewiring the agency, from agentic marketing platforms to output-based pay that only works when one partner controls every lever.

Apple Podcasts Spotify
 ·  The Business of Marketing  · S5 E104  · 33 min
Read the interview

"ad trafficking is a soulless, you know, kind of like life-destroying exercise"

The Conversation

From media and telecoms law at Columbia in the mid-1990s, to launching an ad network in South Africa, to selling his own company to WPP and now sitting as its chief technology officer, Stephan Pretorius has spent a career at the join between technology and marketing. He argues that AI is not simply making agencies cheaper, it is changing the nature of the work itself, and rewriting how agencies get paid.

Pretorius trained as a lawyer, studying media and telecoms law at Columbia in the mid-1990s just as the internet was being commercialised, and used it as his ticket out of law and into digital media. He launched, by his account, the first ad network in South Africa, then founded Acceleration, an ad tech and martech business he sold to WPP some 12 years later. At WPP he ran Acceleration, helped then-CEO Mark Read turn around Wunderman, and in 2018 became chief technology officer, a remit that stretched from systems integration to the whole of marketing services, from strategy to creative to production to media. He built the group's open platform, and then AI arrived.

In this conversation with host Justin, Pretorius argues that AI is reshaping the agency on three fronts at once: how consumers search, discover and buy, how work is produced, and how teams collaborate. He describes WPP Open, an agentic platform where humans orchestrate specialist agents, super agents that compress two weeks of strategy into five minutes, and a Veo 3 film that would once have needed a feature-film budget. Most striking is the commercial change, with recent pitch wins that are partly or wholly output based, where the agency gets paid when the client sells a product. It only works, he says, when one partner controls every lever.

Key Takeaways

  • Pretorius has taken a long and windy road from law to technology leadership: media and telecoms law at Columbia, digital media in South Africa, founding and selling Acceleration to WPP, a turnaround stint at Wunderman, and the CTO chair since 2018. He argues that this generalist path is an advantage, not a compromise, and that network theory from his legal training still underpins how he thinks about platforms today. Versatility, in his view, is underrated against the cult of the specialist.
  • He frames the durability of agencies around fragmentation. As media fractured through mobile and the social economy, knowing how to execute in one channel became easy, but knowing how to allocate across channels and how they combine became harder. That complexity, rather than any single medium, is what keeps clients coming to marketing services companies. Agencies should follow the eyeballs and stay unsentimental about any particular channel.
  • On AI, Pretorius separates the boring efficiency story from the real prize. Tasks like ad trafficking can and should be automated. The excitement is doing work that was once impossible for reasons of time or budget: super agents that compress two weeks of strategy into five minutes, and AI film production that replaces feature-film budgets. He rejects the AI slop meme, arguing that in skilled hands the quality is now so high that spotting the AI means it was used badly.
  • The commercial change may be the sharpest argument. Clients moved from wanting AI to make work cheaper, to wanting more content for the same money, to wanting better content. Because WPP can now control the whole supply chain, its newest models are partly or entirely output based, getting paid when the client sells a product. Pretorius is clear this only works when one partner controls strategy, creative, production and media together.
  • On leadership, he credits bottom-up, distributed innovation over central control. WPP trained its workforce on AI since 2019, built the Hex apprenticeship for AI-native talent aiming at a thousand by 2028, bought and spun out research labs, and put a no-code agent builder in everyone's hands, yielding thirty thousand agents in a month. He governs it with first-principles guardrails on copyright and client data, and warns that the acceleration of AI-driven software raises open questions about productivity, differentiation and trust.

Topics Covered

  1. 01 AI in agency work
  2. 02 Agentic marketing platforms
  3. 03 Output-based commercial models
  4. 04 Distributed innovation and adoption
  5. 05 Governance and first principles

Key Exchanges

05
01 Beyond efficiency and automation, where do you think AI unlocks meaningful strategic impact?

composite complex strategy agents that can do about two weeks of work in about five minutes

A lot of people focus on AI for efficiency, and that's the boring part. There are many tasks humans should never have done. Ad trafficking is a soulless, life-destroying exercise, mechanical and high precision, and it can be largely automated. What's exciting is using AI to do things you either couldn't do before or didn't have the time or budget for. In strategy, we've built composite complex strategy agents that do about two weeks of work in about five minutes, giving not just the obvious answer but the non-obvious one. When Veo 3 launched we made a football-through-the-ages film for Google, about 15 AI-generated scenes that would once have needed 15 locations and a feature-film budget, in less than a week with about three thousand prompts. The quality is now so good you can't tell it's AI, and if you know that it's AI, it's not being used well.

02 This revolution changes how brands buy and how service providers sell. How are you evolving your commercial models?

So we get paid when the client sells a product.

In year one, clients asked how much cheaper AI would make things. Very quickly they said they didn't want it cheaper, they wanted more content, more personalised. Then they wanted better content. We've adjusted our unit economics and keep making our commercial models more aggressive. When you control all the levers, strategy, creative, production, media, social, influencers, you can be far more aligned with clients on how you get paid. Many of our largest new models and recent pitch wins are partly or entirely output based. So we get paid when the client sells a product. But you can only do it when you control all the levers.

03 From your vantage point, how do you see AI transforming the way agencies and brands operate?

instead of having five different agencies do, do the project, you now have five specialists working in a team with a bunch of agents on Open, and you can get the work done.

For the first time I've had to think not just about how technology lets us deliver differently, but how it changes the nature of work itself. AI is changing how consumers search, discover and buy, how we produce, and how we work. We built WPP Open, which integrates the end-to-end supply chain on one data layer, then agentify processes: sometimes fully automated, sometimes humans working with agents. We call it agents in the loop, not humans in the loop, where humans orchestrate workflows and bring in specialist agents. So instead of having five different agencies do the project, you now have five specialists working in a team with a bunch of agents on Open, and you can get the work done.

04 How are you ensuring WPP and the team adapt to this transformation?

Within a month, we had thirty thousand agents built in the business

Adoption and change management is the hardest thing. You need to drive distributed innovation. No one can sit centrally and understand all the use cases and then come down from the mountain with the tablets to tell people how it will be. You empower people with general purpose technologies, explain how they work, put guardrails around them, and let them free. We launched a no-code agent builder to the entire organisation at the end of 2024, and within a month we had thirty thousand agents built in the business. Our best ideas for super agents came from the community. That bottom-up energy is what's driven the transformation more than anything.

05 If we sit down in five years' time, what are we talking about?

we're in for a wild ride in the next couple of years

Marketing services won't look the same in five years. Being a trusted growth partner in the age of AI can mean commerce platforms, influencer programs, data analytics, CRM and digital products, so what we sell and are famous for will be very different. The existential question is software and human work. Teams message me saying their app is writing itself, and they're serious. People are shipping forty thousand lines of code a day. If you can articulate a specification well enough, anyone can spin up software of any complexity, and if all knowledge workers can do that, what does it mean for productivity, differentiation and trust? We're in for a wild ride.

S5 E104Season & Episode
33 minDuration
30,000 AI agents built in the first month
1,000 Creative technology apprentices targeted by 2028

"if you know that it's AI, it's not, it's not being used that well."

Hear Stephan on
The Business of Marketing
Season 5 Episode 104 33 min
Apple Podcasts→ Spotify→ Stephan's full profile→

Keep Reading