Bad AI outputs often start with a data problem

By Marianna Imprialou and Zach Bricker

Everyone wants to know why AI is falling short. 

Leaders expected a revolution, but most are seeing incremental or unreliable change.

The real issue is data connectivity. 

We surveyed over 400 marketers for our 2026 Marketing Data Report, and 36% named connecting data across channels as their absolute biggest hurdle. 

Teams are juggling three times the data they had in 2020, yet they still struggle to answer basic strategic questions. The information exists. It’s just locked in silos, poorly managed, and set up to sit in storage rather than driving decisions. 

When you plug that kind of messy data into an AI tool, you just get the same broken results at a much faster speed.

Marketing data governance: the unseen problem holding AI back

Marketing data governance is boring. Everyone knows it. The line from governance to profit isn’t always obvious: it's tedious to get right, and there's always something more urgent demanding our attention. So it gets pushed back.  

Can you answer the following questions on your company data?

• Who owns our data?
• Where does it go?
• Who should have access to it?

AI just made a lack of marketing data governance a very expensive bad habit.

It doesn't know your data is a mess. It just uses it, at scale and with total confidence, even when it’s missing critical data points due to limited access. 

Which means you either have to limit your use of AI to tinkering or fix your governance right now. 

For marketers, this is a particular challenge. Not because they don’t care about data, but because they’re being left out of the conversation: 52% of marketers say data strategy is owned outside marketing. 

Who decides what data gets collected? Probably IT. 

Who decides how it's stored? Usually data engineering. 

By the time it reaches a marketer, it's been shaped by people optimizing for technical efficiency, not business decisions. Meaning you get a whole lot of data, but not the data you really need. 

Stop patching your data stack — start fixing the foundation 

Only 6% of marketing teams have fully embedded AI in their workflows. But 80% are under C-suite pressure to make it happen.

That gap is an opportunity. Leadership has never been more motivated to invest in data infrastructure. Use that leverage before it closes. 

Here's what to do: 

Start with decisions, not data 

87% of marketers believe better analytics would improve their effectiveness — but 56% can't find time to analyze what they already have.  

Before adding anything to your stack, define your most important business questions and work backwards. What do you need to answer them? What data does that require? 

If your data can answer real business questions, you have enough. Don’t keep throwing more numbers at the problem until you’re successfully using what you already have.

Get marketing into the data conversation early

Marketing needs to be involved in what gets collected, how it's structured, and who can access it — or you'll keep inheriting data that looks comprehensive but is useless in practice.

Do your data governance homework 

Map your data. Define ownership. Document who has access and why. Make sure your key systems are connected. And check that your data is structured for decisions, not just storage. 

It’s not fun, but it’s the difference between AI that compounds your problems and AI that actually solves them.

The companies getting real value from AI in 2026 are the ones that fixed the right things first.

Want to know exactly what you need to do to get your marketing data AI-ready? Download our 2026 Marketing Data Report to find out what the best teams are already doing — and get a checklist to see where your gaps are.

About the author

Marianna Imprialou

Marianna is the Head of Data Science at Supermetrics. She's leading the development of Supermetrics' AI capabilities that turn complex marketing data into reliable, actionable insights. With over 15 years of experience in data science across academia and consulting, Marianna began her career as an Assistant Professor at Loughborough University, where she taught machine learning and statistics, published peer-reviewed research, and earned recognition for her academic work. Before joining Supermetrics, she worked as Director of Data & AI at EY, where she led data science teams and developed MarTech analytics solutions for global telecom and utility providers.

Zach Bricker

At Supermetrics' US operations, Zach serves as the Lead Solutions Engineer. With a genuine enthusiasm for data and its applications, Zach guides a dedicated team of solution engineers in the US region, which represents an impressive ARR of over $10M.