Artificial intelligence is reshaping how brands understand and engage with their audiences — but with this transformation comes a question of trust.
For Elizabeth Dobbs, AVP of Marketing Technology, Data, and Growth at Databricks, the real power of AI lies not just in automation or analytics, but in how marketers use it responsibly to elevate strategy and creativity. Blending technical expertise with a deep understanding of human behaviour, Elizabeth is helping redefine what intelligent, ethical marketing looks like in the data-driven age.
Many marketers are exploring AI, but trust remains a sticking point—what does “trusted AI” actually mean in the context of marketing campaigns?
When we talk about trusted AI in marketing, it really comes down to three things: having a human-in-the-loop, ensuring contextual, intelligent output backed by enterprise data, and robust governance with tools like Unity Catalog – our centralized data catalog that provides fine-grained access control for structured and unstructured data in multiple formats on various platforms.
Marketers can’t just take AI outputs at face value; they need to know the data behind it is accurate, consistent, and aligned with how they ask questions.
That’s why iterative feedback was pertinent when we were building ‘MARGE’ (Marketing Genie), Databricks’ internal AI-powered business intelligence tool loaded with our marketing data. MARGE allows our own marketers to interact with data using natural language (e.g. which marketing campaign drove the most marketing qualified leads in Q2) and encourages users to validate the results with a feedback loop to ensure its accuracy and trustworthiness.
We also worked directly with marketers to understand specific Databricks marketing terms and how they would phrase questions and expect them to be answered. We then trained ‘MARGE’ to deliver validated outputs with a feedback loop from users. That consistency is what builds confidence and cultural alignment around the metrics that matter most.
However, evaluation and security are often the most overlooked and important piece of AI. Building AI without evaluation is like building a house out of wood with no understanding of construction. That’s why 95% of AI experiments never make it into production, because organizations can’t confidently guarantee accuracy, security, or quality from their AI models. At Databricks, we’ve spent a lot of time building the tooling, such as Agent Bricks, to change that. We strive to give marketers and their organizations a repeatable way to measure quality with transparent benchmarks they can show to legal and compliance teams, flexibility to use the best-fit model (whether that’s a partner’s LLM or one they’ve built themselves), and governance that ensures privacy and security rules carry through into production.
Trusted AI, then, is not just about flashy outputs. It’s about reliability, explainability, and governance, making sure you can trust that the “house” you’ve built will stand.
When you combine human oversight, fine-tuned models, and strong governance, you create a system marketers can rely on to make confident, data-driven decisions without second-guessing the insights.
How can data intelligence powered by AI uncover customer insights that traditional approaches often miss?
Businesses don’t just need general intelligence; they want domain intelligence and to build custom AI models using their proprietary data that will give them the ability to scale.
We call this data intelligence. In this ideal state, marketers are given the ability to do what traditional approaches never could, run exponentially more experiments, connect disparate signals, and uncover patterns across business silos. We have also seen that at this current stage, it is also not just about growth, but also being able to apply the unique marketing context across channels, audiences, and behaviors, while keeping data quality and security intact.
Read also: Government champions digital technology for inclusion and access
For marketers, this means intentionally shifting away from mundane and repetitive work and toward higher-value activities. A well-trained AI can handle the repetitive tasks like generating ad variations and repurposing copy so teams can focus on building strategy and high leverage campaigns.
The real promise is in applying expertise where it matters most: finding new ways to delight customers and driving growth with a tangible ROI. For example, PetSmart, one of the largest U.S. pet supply retailers with over 65 million loyalty members, built a customer 360 on Databricks to unify data, analytics, and AI. This enabled personalized engagement for a loyalty program driving nine out of 10 transactions. They discovered that customers who used their grooming service, PetGroup, spent far more, but only a fraction used it. Instead of testing a few A/B campaigns a week, PetSmart’s AI-based system ran thousands simultaneously, analyzing variables like buying behavior, geography, and timing. By week three, grooming sign-ups rose 22%. The system also revealed overlooked insights—for instance, emails sent on Mondays and Tuesdays outperformed others because customers plan grooming early in the week. That’s the power of pairing machine-scale experimentation with human expertise.
Such a result is not an outlier, but increasingly common as companies operationalize AI at scale. Those kinds of measurable outcomes make the business case clear to any CFO or CEO.
Of course, the foundation is critical. AI trained on bad data will only amplify the flaws. Getting your data house in order, setting clear quality benchmarks, and understanding cost at scale are all prerequisites. Do that, and AI becomes a multiplier for every marketing discipline.
The term “composable CDP” is gaining traction—can you explain what it is and how it transforms campaign activation across channels?
Think of a composable CDP like legos. Instead of buying a monolithic, one-size-fits-all customer data platform, you keep your data on a unified foundation — in our case, the Databricks Data Intelligence Platform — and then “compose” the right activation tools on top of it. Your BI and CDP sit side by side, all reading from the same governed data source.
That solves three problems marketers face every day:
- Silos and slow handoffs: The data doesn’t get copied around just to run a campaign. It stays in one place, accurate and governed.
- Bottlenecks: Marketers can build audiences and activate them across channels without waiting on heavy engineering support.
- Lock-in and cost: Because the tools are modular, you can swap them in and out as your needs evolve, instead of being stuck with an expensive all-in-one system.
The Databricks Intelligence for Marketing acts as the foundation and single source of truth for the composable CDP, unifying customer and campaign data with built-in AI. It lets every marketer — regardless of technical background — self-serve insights, personalize experiences, and activate campaigns in real time, all while staying on a trusted data platform.
How are organisations leveraging AI-powered workflows to make real-time decisions without relying heavily on technical expertise?
The data has always been both the lifeblood and the barrier to any experimentation program.
Using natural language, companies are reducing the traditional analyst bottlenecks that come with design and measuring outputs – freeing up marketers to bring experimentation to the center of their execution strategy.
B2C companies with short test-to-revenue cycles typically have a strong experimentation muscle, get signals quickly, and immediately translate them into output. That agility has allowed B2C leaders like Grammarly and Skechers to make real-time personalization part of their DNA.
Grammarly, for example, processes 5 billion user events every day. By moving to the Databricks Data Intelligence Platform, they’ve been able to make those events available for analytics in under 15 minutes instead of four hours, cutting costs by 90% while delivering 110% faster querying. That means marketers and product teams don’t need to wait on analysts or rely on complex SQL — they can tap dashboards, test ideas, and act on real-time insights almost instantly.
The bigger picture is that AI has ended the old debate around whether personalization is worth the effort. With the barrier to entry so much lower, the conversation has shifted to how can we personalize most efficiently and effectively? And that’s what these AI workflows deliver – the ability for business users to access trusted insights in natural language, run experiments at scale, and make authentic, customer-first decisions with speed and confidence.
What are some communication challenges marketers face when trying to integrate AI into existing strategies, and how can they overcome them?
The potential of AI in marketing is endless — and the reality is, AI is the worst it’s ever going to be right now.
The biggest communication challenge for marketers is trusting the AI when speaking to it. Trusted AI is what holds many people back, which is why the most powerful use cases today are human-in-the-loop: letting AI handle execution within scope while marketers provide oversight.
The first step in integrating AI effectively is looking at how they can augment human workflows; working alongside humans. AI is especially good at the things we dislike most – the repetitive, structured work like drafting, tagging, and running campaigns.
At Databricks, we’ve built AI agents to solve exactly these challenges. Two in particular stand out: Agent Tagatha, which automates content tagging across thousands of assets—reducing what once took months to just hours—and Agent Atlas, which redefines segmentation. Instead of relying on static rules, Atlas blends roles, titles, and real-time behaviors to ensure every person receives the most relevant content. Since deployment, Agent Atlas has driven a 30% lift in engagement across key audiences, powered by insights from millions of users and tens of millions of behavioral signals.
Marketers succeed with AI by pairing automation with human judgement, using oversight to set thresholds, validate outputs, and continuously refine the models.
Looking ahead, how do you see trusted AI reshaping the marketer’s role in building authentic, customer-first campaigns?
AI is going to fundamentally reshape the marketer’s role by making everyone a growth marketer.
Historically, running something like an account-based marketing (ABM) campaign came with a huge operational burden — custom creative, custom copy, manual segmentation, and the risk that choosing the wrong accounts could render all that effort ineffective. AI changes that equation by allowing marketers to make bigger bets, test more ideas, and do it all with much more agility. You can have several oars in the water at once and quickly see what’s working.
The real winners will be marketers who embrace experimentation — trying, testing, refining concepts and campaigns. And the differentiator will be creativity and the ability to ask the right questions. Prompt engineering and context setting are becoming core skills, because the marketers who can do so will be the ones who use AI most effectively to build authentic, customer-first campaigns.
For this to happen, leaders also need to encourage creativity within their organisation and help teams gain hands-on confidence with AI. That’s why we’re also investing in building more comfort and familiarity with AI within the Databricks Marketing teams. We’ve recently hosted a Shark Tank-style hackathon to get cross-functional marketing teams together to surface common challenges, prototype AI solutions, and even do a bit of “vibe coding” to bring ideas to life.
I don’t believe that AI is going to take your job, but a marketer who knows how to use AI better might.
Elizabeth Dobbs
As AVP of Marketing Technology, Data, and Growth, Elizabeth Dobbs operates at the intersection of strategy, advanced analytics, AI, and scalable technology. She leads cross-functional initiatives that convert customer insights into measurable growth, arming go-to-market teams with the data, tools, and infrastructure to move faster and make smarter decisions.
Elizabeth has built high-performing marketing operations, scaled experimentation across channels, and architected systems that turn complex tasks into automated, repeatable solutions. Her contributions span full-funnel optimization, the deployment of composable CDPs, and the integration of AI throughout the marketing stack to accelerate performance and drive efficiency.
