Bastien Parizot has spent the best part of two decades harnessing the power of technology to transform businesses. 

Today, as SVP Global Business Services at Reckitt – the consumer goods giant behind household names including Dettol, Lysol, Durex, Finish and Nurofen – he leads one of the company’s most consequential bets: the integration of AI into global operations.

It is a role that positions Bastien at the helm of Reckitt’s enterprise-wide AI transformation.

The goal, as Bastien frames it, is to “reinvent our end-to-end processes and enable a more effective, simplified Reckitt”.

Applying AI systematically

Reckitt’s brands are used in hundreds of millions of households every day. The company has built its reputation over more than 200 years on product superiority and consumer trust. 

Now, it is betting big on AI to sustain that edge in a fast-moving market.

Reckitt’s AI journey began in earnest in early 2024, when the leadership team asked a pointed question: given the rise of AI and the explosion of available data, what could the company do differently? The answer set in motion a programme to apply AI systematically across the business.

Marketing came to the fore as a natural starting point. Already one of Reckitt's most data-mature functions, with existing experience in machine learning, it offered fertile ground for experimentation – and the results exceeded expectations.

After a successful proof of value, Reckitt deployed a suite of AI-powered tools to more than 1,000 marketers, covering reporting, creativity and innovation. The outcomes were striking: a 60% boost in marketing efficiency and a significant reduction in time spent on routine undertakings. 

Bastien reflects: “Our pilot showed that time spent on everyday tasks – for example, a post-campaign media analysis – could be reduced by up to 50%, with better quality.”

From programmer to AI leader

Bastien’s path to the AI frontier was far from direct, but in hindsight feels inevitable. Technology has been a constant thread throughout his career, regardless of the specific function he was serving.

“Whatever role I’ve held over the years, technology has always been a core enabler of transformation, execution and impact,” he says.

“From programming, digital and now AI, technology has been the differentiator for businesses for decades. Whether it brings disruption or opportunity depends on how you choose to approach it.”

That philosophy has kept Bastien close to emerging technology through successive waves of innovation. It also explains his appetite for a role that, by his own admission, has no template.

Bastien describes his working day as anything but predictable. As the architect of a newly-created remit, he moves fluidly between strategic conversations and operational detail. 

He continues: “In one day I can move from reviewing how automation is helping us identify gaps in contracts, to working on how we accelerate AI across the enterprise, to a deep process and data review with our procurement team in India or a very exciting working session with the team in China on livestreaming.

“Reckitt’s purpose, global footprint and diversity – combined with the pace of transformation internally and externally – make my days some of the most intense of my career so far, but also the most exciting.”

The pilot trap

For all the excitement surrounding generative AI, most enterprise adoption efforts continue to fall short. The vast majority of pilots never make it to production – and Bastien has a clear view on why.

“The most common mistake firms make when adopting AI is trying to launch too many projects at once,” he observes. 

“We know business leaders are under increasing pressure to demonstrate how they're integrating AI. But too often, a lack of strategic focus makes it more difficult to move pilots out of the experimentation phase.”

The data is sobering. According to research from Boston Consulting Group (BCG), only 4% of companies take a structured approach to AI deployment – yet that small cohort represents the top tier of value creators. 

Bastien also challenges the assumption that deploying a large language model (LLM) off the shelf is sufficient: “The other hurdle is thinking we just need to deploy a GPT solution and ask it questions.

“While we use available LLMs, we have seen the value is unlocked from applying tailored solutions to our problems – ensuring the business context, tacit knowledge and access to structured and unstructured data is built into our applications.”

Perhaps most importantly, he argues that scale must be considered from day one. 

“It is less about the technology,” he goes on, “and more about change with people and how they work, process and the knowledge and context we need to capture and codify for leveraging AI.

“We follow the 70/20/10 principle: it's 70% about people, 20% about process and data and 10% about technology.”

Measuring what matters

Proving ROI remains one of the central challenges in enterprise AI.

At Reckitt, Bastien has developed a disciplined framework for assessment that resists the temptation to measure activity rather than impact.

“It starts with defining what success looks like,” he contends. “At Reckitt, we assess AI through three questions: is it doing what we expect? Does it provide us with a more effective way of doing the work? And is the output at minimum on par with our current ways?”

The company’s Insights Generator and concept development solution illustrates the approach in practice. The tool brings together more than 40 internal and external data sources, giving product teams faster access to consumer and market intelligence.

“It provides access to insights 70% faster and helps us create product concepts that are testing twice as well as before,” Bastien adds. “When deployed at scale, results like this deliver a clear and measurable return on investment.”

The SLM shift

Debate is growing around whether general-purpose LLMs will eventually give way to smaller, more specialised alternatives. 

Bastien outlines a nuanced picture shaped by industry context and use-case specificity, where certain sectors gravitate towards highly-specialised small language models (SLMs). 

“Some industries – like insurance, banking and pharma – will probably benefit from SLMs because their work relies on deep, domain-specific knowledge,” he says. “For many other sectors, the real value will come from taking strong general models and continuously refining them with context, data and use-case-specific tuning.”

The trajectory, in Bastien’s view, points towards AI that acts less like a chatbot and more like a genuine operational partner: “Purpose-designed solutions – often fine-tuned with a company's data, creative standards, regulatory needs and processes – allow AI to shift from being a chatbot to becoming a true agentic partner.”

Balancing innovation with privacy and security

As AI becomes embedded in core business processes, the question of data governance becomes increasingly urgent. For Bastien, the answer lies in applying the same rigour to AI that responsible companies already apply to data more broadly.

“Enterprise-grade AI requires and benefits from the same standards of privacy and security that already govern how companies handle data,” he says. “Our models and data operate in secure environments and we are very selective about any third-party APIs or external solutions we use.”

Reckitt has invested in guardrails, observability tools and a robust responsible AI assessment process for every application it deploys. The firm has also developed its own responsible AI principles alongside consumer and data governance frameworks.

One area demanding fresh thinking is unstructured data – the documents, conversations and institutional knowledge that Gen AI can unlock but that traditional governance frameworks were not designed to manage. 

“We need to approach data governance differently with Gen AI, in particular turning a whole new set of information into data,” Bastien asserts. 

Beyond technology and process, he holds firm to a foundational principle: “We believe strongly in keeping a human in the loop. AI can support better decisions, but people remain accountable for making them.”

Scaling with discipline

Having proven AI’s value in marketing, Reckitt is extending its approach into R&D and across other functions. Lessons learned in the early stages are shaping how the company thinks about scaling.

“Scaling Gen AI is a continuous learning journey, so ‘success’ is relative and a moving target,” Bastien admits. “To avoid getting lost in the sea of pilots, we have started to look at our business operations – what is our current process and what can AI do for us? – before even talking about technology.”

Reckitt’s approach is deliberately focused. Rather than pursuing a broad sweep of simultaneous deployments, the organisation has adopted what Bastien calls a “deep and narrow” strategy – going function by function and stress-testing each use case against criteria of feasibility, effectiveness and quality. 

Scaling, he stresses, is ultimately a human challenge: “It lands if it is useful and our teams are using it. So, we spend a lot of time ensuring those AI solutions are part of the day-to-day and not just a fancy technology that fades.

“We’re seeing strong traction and promising results and that gives us confidence as we continue our journey to transform how Reckitt operates with the help of AI.”