AI ROI Stalls When Marketers Don’t Trust Their Data
Nearly eight in ten organizations report no significant bottom-line gains from AI.
McKinsey
In marketing, that gap is hard to ignore. Budgets are already allocated, tools are deployed, and leadership is watching. Yet many marketing teams still lack the data conditions required for AI to influence decisions with confidence.
Campaign data lives in Google Ads, audience records in a CRM, pipeline metrics in a separate dashboard, and engagement signals somewhere else entirely. The data is abundant, but none of it speaks to the rest. When AI draws recommendations from sources that don’t connect, the outputs can’t be trusted, and teams know it. Every recommendation gets second-guessed, every forecast reconciled manually, every insight validated before anyone acts on it.
That hesitation is where AI ROI dies.
Where Trust Breaks Down
Our 2025 Annual Data & AI Utilization Index, which surveyed practitioners, managers, directors, and executives across industries, found that 56% of respondents operate at a data maturity level of 3 or below on a 5-point scale. More than 42% said improving data quality and accuracy is their top forecasting priority. Half said they rely on a combination of data and intuition to make decisions because their data doesn’t provide a complete picture.
Accuracy of AI results is a big barrier. Right now we spend more time proofreading the results than we would have spent doing it ourselves.
Respondent
The pattern shows up across marketing organizations: attribution models conflict across platforms, forecasts shift depending on which dashboard you pull, and metrics can’t be defended in a room with finance or sales. Marketers fall back on manual validation, which ironically cancels out the efficiency AI is supposed to deliver.
The Real Obstacle to AI Adoption
McKinsey’s research shows that companies achieving real returns from AI are moving beyond single-tool deployments toward workflow redesign. That means rebuilding how marketing decisions actually get made, not adding a predictive layer on top of a broken reporting structure.
Agentic AI acts and decides autonomously across marketing and sales processes, further raising the stakes. A system that can execute campaigns, adjust bids, and route leads without human approvals at every step is only as reliable as the data it’s working from. If that data is inconsistent, the automation only scales the problem.
Over 65% of organizations in our survey reported still using Excel or Google Sheets as their primary data management tool, as they are familiar and accessible, but insights drawn from spreadsheets quickly become outdated. Manual data pulls produce errors, and when AI learns from fragmented, manually maintained sources, the recommendations reflect that.
Adding AI capability on top of disconnected systems results in more noise, not the progress or efficiency marketing teams were promised.
The Impact in Practice
Consider this client example. A high-tech drone software company growing at 30% to 45% year over year. Their proprietary SaaS platform generated user-behavior data, but it had no connection to their CRM or marketing systems. Campaign results were inconsistent and difficult to diagnose. Forecasting required manual reconciliation across sources.
A three-stage implementation connected their SaaS platform to Salesforce and Google Analytics, creating a unified customer record for the first time. Case management was automated. Marketing was rebuilt on top of live engagement data. Sales and service teams aligned on shared definitions and shared dashboards.
The result was a complete 360-degree view of the customer base, increased productivity across every team, and forecasting that no longer required manual reconciliation. With a sound data foundation, AI tools layered on top became genuinely usable because the inputs reflected reality.
Bridging the Gap Before the Agentic Era
More than 40% of early-stage agentic AI projects will be abandoned by 2027.
Gartner
All because the systems feeding AI models were never set up to produce consistent, trustworthy inputs. Organizations that skip the foundational work of connecting systems, standardizing definitions and establishing governance don’t get less value from agentic AI. They get faster, more automated versions of the same problems they already have.
What Progress Actually Looks Like
Financial returns from AI often lag operational improvements. Early progress is quieter than the headlines suggest. It looks like a marketer who pulls a report and doesn’t immediately open a second dashboard to check it. A forecast that goes to the CFO without three rounds of internal revisions first. A campaign decision made on Monday that doesn’t require a Friday postmortem to figure out which number was right. When data is consistent, governed, and connected across systems, AI stops being a tool that generates work and starts being one that eliminates it. Once people have no reason to distrust the data, the AI adoption will accelerate.
