AI is revolutionising industries by enabling companies to rethink how they deliver value to customers. Yet, the challenge is often in moving beyond pilot projects to scalable, impactful use cases.
By following a structured process and guiding principles, companies can better harness AI to improve customer experience and drive innovation and growth. A number of organisations, including Diageo, a global leader in the beverage industry, have successfully adopted such an approach – offering useful lessons for others.
1. Start small (and smart), then scale
Focus on high-impact, low-complexity use cases to demonstrate value and create buy-in across the organisation. Use a “pilot and scale” approach to test in controlled environments before full deployment. Focus on areas where inefficiencies exist, customer needs are unmet or new possibilities remain untapped. Small, nimble teams can often lead this phase. The goal is to build early proof points and momentum without requiring major organisational change.
For instance, Diageo’s Breakthrough Innovation teams applied an internal rapid experimentation process called Ignite to assess and de-risk ideas. One challenge they tackled was helping customers navigate the intimidating process of selecting whisky. By partnering with sensory science machine learning company Vivanda, they developed Flavor Print, an AI tool that matches consumers’ flavour preferences with product recommendations. The platform’s success in 40 markets and 20 languages has helped guide marketing and product decisions.
2. Business value first, AI second
Scaling AI solutions requires a disciplined focus on outcomes. Define clear business objectives before selecting AI technologies. AI should be a means to solve problems or create opportunities, not the end goal, and only use cases that demonstrate clear business value and align with strategic goals should be expanded. Metered funding – where resources are released in stages based on proven results – ensures that AI investments remain outcome-driven.
Diageo’s HALO initiative, a generative-AI-powered platform that allows consumers to co-create personalised bottle labels and other merchandise, provides an excellent example. After successfully testing the concept, the company offered customers the opportunity to create unique bottle labels for Johnnie Walker whisky by training the AI engine on brand elements and the artwork of Scottish artist Scott Naismith. This resulted in a 110 percent sales uplift for Johnnie Walker Blue Label at a 20 percent premium.
3. Data as a strategic asset
Ensure your data is clean, accessible and well-governed. This includes breaking down silos across departments and ensuring a robust data infrastructure. A study from Harvard Business School highlighted that organisational silos are the top barrier to leveraging data effectively. The second biggest barrier is data standardisation and technical integration. Clear data-sharing protocols and a powerful and accessible “data lake” that aggregates all relevant data help address both challenges.
Starbucks tackled this through its Deep Brew platform, which enables hyper-personalised customer experiences based on transaction data, preferences and behavioural insights. Their enterprise data analytics platform and data lake unify data from various sources, which is processed through a compute layer within Deep Brew. This powers personalised recommendations, customised offers and supply chain optimisation, all delivered through touch points like the mobile app, digital drive-thru, website and social media.
4. Cross-functional collaboration
To support AI implementation at scale, it’s crucial to build an AI leadership ecosystem. Foster a network of AI champions across functions and invest in upskilling teams to work effectively with AI technologies. These partnerships can provide additional expertise and access to cutting-edge capabilities.
Diageo’s AI council exemplifies how leadership and collaboration can drive innovation and efficiency. This cross-functional forum includes people from IT, legal, procurement, digital, innovation and planning. There is a bottom-up element where every new AI trial needs to be flagged by markets or brands via a digital process or portal, and approved to ensure that knowledge sharing, risk mitigation and efficiencies are fully considered. The council will then own specific strategic workstreams such as training agenda, responsible AI framework or AI insight agenda.
5. Speed with discipline: Agility experimentation at scale
Once high-potential opportunities are identified and executed, the next step is to build more strategic agile experimentation capabilities. Rapid innovation cycles, grounded in learning plans and evidence scorecards, allow organisations to test and iterate AI solutions at scale. To sustain experimentation across use cases, brands and regions, companies need a top-down strategic vision, bottom-up empowerment and the right organisational infrastructure. This phase demands systems, governance and cultural readiness to move from isolated pilots to an enterprise-wide test-and-learn engine.
At Diageo, opportunities are only pursued if they can scale across at least two of three areas: brands, categories, regions or channels. They define the “size of the price” at the start and adjust as they learn. Projects with the highest expected return, based on potential impact and risk, are prioritised. One example is Seedlip, an AI brand ambassador launched within three months in 2023 with limited functionality. User feedback and usage helped improve it and add new features, tripling conversion rates. The learnings from this smaller challenger brand are now informing the rollout into global hero brands.
6. Technology stack scalability
Invest in flexible technology architectures. Headless setups where the frontend (apps, chatbots) is separated from the backend (data) allow companies to scale and adapt quickly as new AI models and providers emerge, while mitigating risks. Building internal capabilities alongside a network of external, solution-agnostic advisors is key. Corporations may not be able to attract top AI talent but can work with procurement, IT and legal to build and evolve their partner ecosystem. However, architectural flexibility also brings increased complexity and integration challenges. Strong technical oversight is needed to avoid creating fragmented systems, security vulnerabilities and mounting operational costs.
Unilever is a good example of achieving tech-stack scalability through a modular, headless architecture. By decoupling data, models and interfaces, the company has built a flexible ecosystem that allows it to integrate best-in-class AI tools from multiple providers, while maintaining control over core capabilities. A global data strategy – including proper data integration from various sources into one data lake – ensures consistency and interoperability across markets and functions. This enables real-time decision-making and rapid deployment of AI use cases, from demand forecasting to generative content creation. As a result, Unilever was able to stay agile, reduce vendor lock-in and continually adapt to a fast-evolving AI landscape.
Guiding principles
While the six steps outline what companies must do to leverage AI, a set of guiding principles defines how they must operate. These principles cut across all stages of AI maturity – from initial pilots to enterprise-wide deployment – and help organisations stay focused, agile and customer-centric.
- Base decisions on evidence: Use data to prioritise use cases that align with strategic goals.
- Move fast: Adopt a test-and-learn approach, prioritising experimentation and fast iteration. Quick wins build credibility and momentum.
- Sponsor and empower from the top: Champion AI at the leadership level and empower teams to drive use cases from the ground up.
- Focus on the customer: Design AI use cases that meet real customer needs, simplify their decision-making and offer more personalised experiences.
- Build for trust: Integrate responsible AI principles – such as transparency, fairness and privacy – into every stage of development and deployment.
AI is a continuous journey of discovery and reinvention – a way to reimagine how businesses engage with their customers. Now’s the time to start, experiment and see what it can really do.
