Marketers have no shortage of ideas for using AI. Turning those ideas into functional, revenue-generating marketing programs is another matter entirely.
That gap between ambition and execution took center stage at the September MarTech Conference panel titled “Built for yesterday: Why your data architecture can’t keep up with AI.”
The session focused on why legacy enterprise data stacks actively inhibit the speed AI promises—and why fixing the problem requires structural redesign rather than simply licensing another platform.
The panel featured three industry leaders:
- Koertni Adams, head of content and product marketing at MessageGears.
- Jacqueline Freedman, CEO and founder of Monarch Advisory Partners.
- Mike Maynard, chairman of Napier Partnership Limited.
- Moderator Kevin Haag, senior vice president of data strategy at Qualify Digital.
Real-time marketing demands real-time data
When an organization’s transition from reactive marketing to proactive execution fails, the breakdown almost always originates in the data pipeline.
“If your data is an hour or more old, by default your execution’s always going to be reactive.” — Koertni Adams
Marketing teams can hire exceptional talent and build efficient internal workflows, but brittle, slow-moving pipelines render those investments useless. Adams pointed out that this friction becomes obvious whenever a team attempts to activate a new customer attribute. If capturing a single data point requires a dedicated data science sprint, a new custom integration, and complex SQL queries, a straightforward request instantly devolves into a multi-month project.
Swapping out marketing automation vendors will not eliminate that bottleneck. If the underlying data architecture remains rigid, the exact same latency issues will follow you to the next platform.
Freedman expanded on this diagnosis, emphasizing that marketing leaders frequently confuse system design failures with software limitations.
“A new shiny tool doesn’t always fix broken issues that are outside of it.” — Jacqueline Freedman
Before evaluating new software, organizations must step back to map how information actually flows between systems — and how human beings interact with those tools on a day-to-day basis.
The problem isn’t a shortage of AI ideas
Legacy data architecture severely restricts activation. Most marketing teams have access to only a tiny fraction of the total customer data across their organization.
Adams noted that latency and restricted access limit every downstream effort, from real-time journey triggers to granular audience segmentation. A team might map out an exceptional multi-touch customer campaign, only to discover their technical stack lacks the accessibility to execute it. AI amplifies this disconnect: it vastly expands what marketers can imagine doing without expanding what their underlying infrastructure can support.
Maynard summarized this friction with a phrase often used at his agency:
“Ideas are easy, execution’s difficult.” — Mike Maynard
For B2B organizations in particular, the primary challenge is supplying AI models with enough rich contextual data to engage complex buying committees. Without deep context, AI simply generates higher volumes of generic messaging. Most organizations ultimately realize their issue isn’t AI capability—it’s access to data.
Freedman advised leadership teams to audit their motivations before investing heavily in AI tooling: Are you solving an actual business problem, streamlining an existing process, or merely trying to satisfy board pressure to deploy AI?
Maynard offered a similar warning: “Just throwing AI at it for AI’s sake is a waste of time.”
AI needs the full contextual picture
Even the most sophisticated machine learning models cannot infer missing enterprise context. Real-time execution relies on feeding models a holistic, continuously updated view of the customer.
“Your AI’s only as strong as the information that it has.” — Koertni Adams
If critical behavioral signals, purchase updates, or service interactions never reach the model, those blind spots directly ruin the output. Furthermore, organizations frequently neglect the return path: marketing teams pull data out of the central warehouse but fail to feed campaign interaction data back into it. This leaves marketing, BI, and data science teams operating from conflicting records instead of refining a single source of truth.
Composable vs. monolithic: No universal solution
One popular response to rigid infrastructure is abandoning monolithic marketing clouds in favor of composable, modular stacks. Freedman strongly advocates for this architectural pivot.
“Do you want a best-in-class stack or do you want a movable monolith?” — Jacqueline Freedman
She likened a monolithic platform to an aging house: every minor renovation risks exposing hidden structural problems. A modular architecture functions more like a custom build, allowing teams to swap out individual components without taking down the entire system—a critical advantage given how fast AI vendors evolve.
However, Freedman cautioned against expecting modularity to fix internal operational issues. “AI cannot fix your bad wiring. It will just make bad processes move really, really fast and go really, really wrong really quickly,” she said.
Maynard pushed back with an essential reality check for smaller organizations. While composable stacks suit large enterprises with deep engineering resources, smaller B2B teams often lack the bandwidth to manage and maintain dozens of point-solution integrations. For those teams, an all-in-one suite with “good enough” features is often far more practical than building a custom best-of-breed stack.
Audit before you act
For teams seeking immediate improvements without launching a massive, multi-year platform migration, the panelists agreed on three steps:
- Map your current data footprint: Freedman recommended starting with a comprehensive data audit to identify every repository housing customer information, who owns it, and whether true single-customer views exist.
- Expose broken feedback loops: Adams advised using the audit process to locate duplicated records and unlinked systems where campaign responses fail to flow back to the central warehouse.
- Focus on high-value signals: Maynard noted that capturing more data isn’t always the goal; capturing the right data is. A single critical data point — like how long a buyer plans to keep a vehicle — often provides far more actionable leverage than dozens of lower-value behavioral metrics.
Takeaways
To close the session, each panelist shared one practical piece of advice:
- Adams: Start small and avoid “boiling the ocean.” Test, refine, and optimize AI activation on a single targeted campaign before attempting enterprise-wide transformation.
- Freedman: Walk through your own customer journey end-to-end, from initial signup to post-purchase support, to experience the actual touchpoints firsthand.
- Maynard: Resist getting distracted by shiny technology capabilities. Focus relentlessly on customer needs and the data required to serve them.
View the agenda and watch the free, online MarTech Conference on demand today.