Transformed from a traditional agency into an AI-first consultancy using Looker as a central intelligence layer
Overcame the enterprise trust barrier by using Looker’s semantic layer to ground AI and dramatically reduce hallucinations
Developed an agentic workflow where data actively triggers autonomous creative content generation
Implemented Conversational Analytics, allowing stakeholders to chat with data using natural language
Automated the creative feedback loop using BigQuery AI to detect creative fatigue and trigger instant optimizations
Overdose Digital built a central intelligence layer with Looker and BigQuery, then launched Project Bluesky to link data with creative production. They also used Gemini and Vertex AI to create an agentic workflow that generates hyper-localized assets with human oversight.
For years, Overdose Digital has managed nearly 300 clients, using custom reporting that relied on manual interpretation. This led to dashboard fatigue, where clients were presented with static graphs, which didn’t intuitively lead to action. With the industry evolving to adopt AI, Paul Pritchard, Overdose CEO, realized the single biggest barrier to enterprise adoption was trust, as clients feared hallucinations and untrusted results.
Overdose needed to transform from a reactive service agency into an "AI-first consultancy." They didn't just need a visualization tool, but a source of truth powerful enough to ground Generative AI. They aimed to build a system where data didn't just sit on a screen but actively directed the business, moving from telling clients what happened to automating what happens next.
We realize that our industry is changing, and we are now an entirely AI first business. Data underpins everything we do, and we believe Looker provides the new baseline of semantic data required to enable future focused, predictive work. It allows us to move beyond telling clients what happened, to telling them what to do next.
Paul Pritchard
CEO, Overdose Digital
Overdose Digital’s transformation rests on two key pillars: Project Titan (the foundation) and Project Bluesky (the innovation).
Project Titan centralized disparate ecommerce data into BigQuery, using Looker as the semantic brain to define metrics universally.
This created the governance required for Conversational Analytics, allowing stakeholders to chat with data and receive accurate and near immediate answers.
With this trusted foundation, they launched Project Bluesky, an agentic workflow that closes the loop between insight and action. Unlike standard GenAI tools, Bluesky utilizes Looker as an active "event bus." When Looker detects a specific data pattern, such as a dip in return on ad spend or a change in local weather, it triggers Vertex AI agents, including Gemini, Imagen, Nano Banana Pro, and VEO.
These agents, grounded in factual brand assets stored in Cloud Storage, autonomously generate hyper-localized creative briefs (such as "heavy coats for a blizzard in New York" versus "light jackets for London rain"). Because the data is grounded in Looker’s semantic layer, the system achieves near zero hallucinations, ensuring brand safety while automating production. Finally, Looker closes the loop by monitoring for creative fatigue, automatically triggering a "Synthesis Agent" to diagnose the performance drop, refine its strategy, and refresh assets in near real time.
The single biggest barrier to enterprise AI is trust. By using Looker’s semantic layer to ground our GenAI models, we’ve solved that. We are no longer just looking at dashboards; we have built a closed-loop system where Looker actively triggers agents to diagnose data and create actions. It’s not just reporting; it’s 'Agentic Commerce', and it virtually eliminated hallucinations.
Ryan Delaney
Chief Experience Officer and Co-founder, Overdose Digital
Overdose Digital is a pioneering global digital commerce agency delivering world-class execution across strategy, experience, technology, marketing, search, data, and AI.
Industry: Technology
Location: New Zealand
Products: Google Cloud, Looker, BigQuery, Vertex AI, Gemini