This article was based on the interview with Crunchbase CRO Ann Davis on the pressure to show ROI from AI by Greg Kihlström, AI and MarTech keynote speaker for The Agile Brand with Greg Kihlström podcast. Listen to the original episode here:
The pressure is on. If you’re a marketing leader in an enterprise organization, you feel it every day. The C-suite, the board, even your own team are all looking for the promised land of AI-driven Return on Investment (ROI). The market is saturated with a sense of FOMO, a frantic race to deploy the latest generative AI tool simply to be able to say, “us too.” This rush often leads to impressive press releases and internal announcements, but the subsequent results are frequently underwhelming, leaving leaders to question the technology, the investment, and sometimes, their own sanity.
The uncomfortable truth is that we’ve been looking in the wrong direction. The obsession with finding a “better AI” is a distraction from the real, far more challenging work that needs to be done. As leaders, we know that a brilliant strategy executed on a faulty foundation is destined to crumble. The same principle applies here. The most advanced large language model on the planet is rendered impotent by incomplete, siloed, and out-of-context data. In a recent conversation, Crunchbase CRO Ann Davis, a veteran of seven startups and a key leader during Looker’s acquisition by Google, articulated this challenge with precision. The path to meaningful ROI isn’t paved with more sophisticated algorithms; it’s built on the “unglamorous work” of data wrangling and infrastructure. It’s time we stopped chasing the AI ghost and started shoring up our foundations.
The Fallacy of the “Best” AI
In the scramble for AI supremacy, the most common question is often, “Which tool is best?” We compare the outputs of Gemini, Claude, and ChatGPT, looking for a definitive winner. According to Davis, this is fundamentally the wrong question to ask. The performance of any model is not an intrinsic quality of the AI itself, but a direct reflection of the data it can access. Without the right underlying dataset, you’re simply getting a slightly different flavor of a generic, and often incorrect, answer.
“I think they’re only as good as the dataset that they’re built upon, right? So if you’re asking, you know, them to do very general type, you know, Q&A, they’re gonna go out and search, you know, everything that’s available on the internet, and that’s probably okay. I think it’s more when you start to try to get into, you know, what we call, um, expert type of requests that you’re only gonna get responses based upon what data they have access to.”
For a marketing leader, this is a critical distinction. Asking a public-facing AI to “write an email campaign for enterprise CIOs in the financial services sector” will yield a plausible, yet generic, result. It lacks the essential context: Which CIOs have engaged with your content before? What are the specific pain points your product has solved for similar customers, as documented in your CRM? What signals from your marketing automation platform indicate they are in-market right now? Davis notes that even a prompt expert gets “very distinctly different answers” from different models, highlighting the absence of a single source of truth. The real work, she argues, happens before you even write the prompt. It’s about connecting the CRM data to the ERP system, ensuring your product usage data speaks to your customer support platform, and creating the connective tissue between your internal data silos. Only then can an AI, layered on top of this rich, proprietary dataset, begin to provide truly “expert” responses that drive business outcomes.
Finding Your Edge Beyond Public Data
If every organization is using AI tools trained on the same publicly available internet data, you haven’t created a competitive advantage; you’ve simply joined the pack. The playing field is level, which means you’re competing on the same generic insights as everyone else. True differentiation, and the resulting pipeline and revenue growth, comes from moving beyond this baseline and leveraging data sources that your competitors either don’t have or haven’t learned to use effectively.
“Everybody, you have to assume, is gonna be using that same public data. So you’ve got to find what’s your edge gonna be and how can you drive more pipeline and more conversion into revenue because you’re seeking out data sources that other people just don’t have or they haven’t learned to use effectively.”
This is where marketing leaders can make a strategic impact. Your first-party data—website behavior, email engagement, event attendance—is a unique asset. But the real power comes from enriching it. Davis points to Crunchbase’s non-public information about private markets as an example. An AI armed with that data can identify high-growth companies long before they appear on a public stock exchange. For marketers, the equivalent could be using specialized third-party data to identify companies that are rapidly hiring in a specific department, have recently adopted a complementary technology, or are showing other non-obvious buying signals. By feeding your AI a cocktail of your unique internal data and specialized external data, you move from generic targeting to predictive, contextual engagement. Your AI is no longer just a content generator; it becomes an insight engine that can surface opportunities no one else can see.
Shifting from Tools to Systems: Empowering Through Operations
The typical technology rollout in a large organization involves selecting a vendor, providing training, and handing the tools to the end-users with the expectation of self-service adoption. This approach is notoriously inefficient, creating a divide between early adopters who thrive and laggards who resist. With AI, this haphazard approach is even more dangerous. To truly realize efficiency gains, leaders must architect a new operational model where the “busy work” is systemically removed before it ever reaches the individual contributor.
“This is really about… where our go-to-market leaders, rev ops leaders, we’re really thinking about how we can take the busy work… out of the AE’s job with AI. Like, including, you know, pre-building dashboards and pre-mapping territories and pre-loading account intelligence. Um, rather than sort of handing the reps the tools and saying, ‘Okay,’ you know, like we usually do in January, ‘here’s your territory.’”
While Davis speaks from a sales perspective, the lesson for marketing is profound. This isn’t about giving your team access to ChatGPT and hoping they write better ad copy. It’s about your Marketing Ops team using AI to pre-build hyper-segmented audiences based on dozens of signals, automating the initial draft of a multi-channel campaign, or generating predictive lead scoring models that are continuously refined. Davis shared an anecdote about a sales rep whose annual territory planning, an exercise that used to take five weeks, can now be completed in 15 minutes. Imagine applying that level of operational leverage to your campaign planning, content strategy, or market analysis. It’s a fundamental shift from individual productivity to systemic empowerment, ensuring that the benefits of AI are felt consistently across the entire organization, not just in the pockets of a few power users. This is how you build a scalable, AI-powered marketing engine.
The pursuit of AI-driven ROI is a marathon, not a sprint, and many are starting the race on the wrong foot. The allure of sophisticated models and slick interfaces has distracted us from the foundational work that truly matters. As Ann Davis makes clear, the future of competitive advantage will not be determined by who has the “best” AI, but by who has the best, most connected, and most contextual data fueling their AI. The immediate task for every marketing leader is to look inward, not outward. It’s about auditing your data estate, breaking down internal silos, and strategically seeking out the unique data sources that will give your organization an inimitable edge.
When asked what we’ll be talking about a year from now, Davis’s answer was simple: quantifying the ROI. The leaders who will be having a positive conversation on that topic are the ones who are embracing the “unglamorous” but essential work of data strategy today. By shifting focus from the AI agent to the data it consumes, we can move beyond the hype cycle and begin building intelligent, efficient, and resilient marketing organizations. True agility isn’t about adopting technology the fastest; it’s about having the strategic foresight to build the right foundation first.
