We Built Diageo’s First Marketing AI Agent
Updated: Sep 2
Learn how Voxly helped Diageo tackle one of the biggest compliance challenges in beverage-alcohol marketing - and what happened when the Diageo Marketing Code (DMC) met AI.
By replacing manual checks with an AI agent, Diageo crushed individual asset review times to 60 seconds and collapsed its entire multi-department sign-off loop from 3+ days down to just 1 hour.
Diageo reviews tens of thousands of advertising assets every year. Johnnie Walker. Baileys. Don Julio. Guinness. Across 150 countries, with different rules in each one.
Every piece of copy, image, and video gets checked against the Diageo Marketing Code (DMC) - an industry-leading marketing standard built by combining global and national standards for alcohol advertising with the brands' ethical marketing goals.
The Marketing Code sets Diageo's minimum standards for responsible marketing and also represent a cornerstone of their corporate culture and the way they do business.
It applies across all markets, and guides everything from research and development to marketing, promotion, and packaging. At the heart of the DMC is ensuring that all Diageo marketing is only ever directed at adults over the legal purchase age, and to encourage drinking in moderation.
Until 2024, the compliance process was entirely manual. Each asset took five to ten minutes to review by a brand manager. Getting full sign-off could take three days or more.
Then the volume of AI-generated content started to climb.
It wasn't hard to see where this was heading. The process that had always been stretched was about to break.
Diageo didn't come to us looking to remove humans from compliance. They came to us looking to give their compliance teams a better starting point.
The manual review process had real problems:
Speed: Ad review was consuming up to 50% of the creative process timeline. Slow reviews and approvals could mean missed market windows.
Consistency: Rules are nuanced. A Johnnie Walker campaign in the US carries different contextual considerations than a Baileys activation in the UK. Consistent application of the DMC across brands, markets, and agencies was genuinely hard.
Scale: The explosion of AI-generated content was only making the volume problem worse. Manual checks couldn't keep up. Our solution was engineering to handle up to 100, 000 assets per year.
Risk: Non-compliant assets don't just cost money in fines. The reputational damage in a regulated sector like beverage-alcohol can be severe.
The ask was to explore whether AI could make a meaningful dent in all of this - without removing human accountability from the process.
What Was Built: An AI Agent for Responsible Marketing
Diageo partnered with Voxly, and we designed and built an AI agent called Dominic - a compliance AI agent trained specifically against the Diageo Marketing Code (DMC).
Dominic scans image, text, and video assets and returns structured compliance recommendations.
Not decisions - recommendations.
It flags potential issues, references the relevant sections of the DMC, and gives reviewers a faster, more consistent starting point. The human still makes the call. Dominic just does the legwork.
We worked iteratively with Diageo over several months, testing accuracy across different content types, brands, and markets. Rules were refined to reflect market-specific nuances - the US was a priority - and calibrated to the particular sensitivities of individual brands.
The service was designed to meet users where they already worked, minimizing training and disruption.
Dominic was connected via API into Diageo's Sitecore DAM platform, so compliance review became part of the existing asset approval workflow rather than a separate step. A Microsoft Teams extension and a conversational chatbot were also developed, giving users access to compliance recommendations without leaving their day-to-day tools.
Dominic was rolled out as a Proof of Concept for testing and validation. The focus was on building trust in the service getting users used to a new marketing workflow.
Results: Faster Better More Responsible Advertising
The project delivered measurable outcomes across every key performance indicator.
Individual asset review: 5-10 minutes → ~60 seconds (around 90% faster).
Compliance sign-off: 3+ days → approximately 1 hour.
AI accuracy: 90%+ vs. Diageo DMC authors. By the way, individual brand managers only agreed with the DMC around 50% of the time, showing how subjective brand campliance can be.
Capacity: engineered to handle up to 100,000 assets per year.
Estimated annual savings at full rollout: around £500,000+ GBP.
That last number matters.
But if you ask us, the more significant outcome was this:
"The compliance results spoke for themselves, but we also greatly valued the learning - understanding how to scope, test, and govern an AI system within a regulated marketing workflow. That knowledge is now shaping how we think about AI adoption and agents across brand and marketing within Diageo." Dana Radu, Diageo. Marketing Risk and Governance Product Lead.
Diageo now has direct, hands-on experience of what it actually takes to deploy an agentic AI system inside a regulated marketing environment. That institutional knowledge — about data preparation, model evaluation, human-in-the-loop design, and change management — is worth more than any single efficiency metric.
What Marketing Leaders in Regulated Industries Should Take Away
This case study is about more than just about using AI to save time on compliance checks. Here are some factors to consider when using AI in a process that carries genuine risk.
Start with a standard. The Diageo Marketing Code was the foundation on which the AI agent was built. Without a clear, written compliance framework, there is nothing to train against and no way to evaluate accuracy. Organisations that want to automate compliance review need to invest first in codifying what good compliance looks like — not just as a legal document, but as a set of evaluable criteria.
Accuracy must be tested not assumed. Voxly and Diageo ran months of iterative testing. For any AI application in a regulated context, a structured evaluation programme - including market-specific and brand-specific use cases - is a must.
Put AI in the tools your team already uses. Users don't want to start using new tools if they can help it. That often means a new username and password, a new bookmark, etc.
So we embedded the Compliance AI agent into Diageo's SiteCore DAM and Microsoft Teams - making it easier and more familiar to get started and drive take-up.
Human oversight is a feature not a bug. The AI agent generated compliance recommendations, not compliance decisions.
The system was designed to augment the judgement of trained reviewers - giving them a faster, more consistent starting point - not to replace it. Organisations should be cautious about any AI compliance tool that removes human accountability from the loop.
A Proof of Concept for genuine learning and validation. Diageo used the PoC not just to validate a product, but to develop institutional understanding of agentic AI. The experience of scoping, building, testing, and deploying - even at limited scale - produces knowledge about what these systems can and cannot do.
If you're dealing with high creative volume, complex compliance requirements, or multi-market brand governance - and you want to understand what AI could actually do for your team - talk to us.
Summary of KPIs & Milestones
The first purpose built Marketing AI Agent at Diageo (PoC was delivered in Feb 2024).
Individual asset reviews cut from 5-10 min to 60 sec per asset.
Compliance sign-off from 3+ days to 1 hour.
Over 90% tool accuracy vs. Diageo DMC experts.
£500K+ in estimated annual savings, had the service gone fully live across the company.
