No AI can clean up your shit.

Over the past few weeks, I’ve noticed a recurring theme dominating conversations in the marketing operations space: the hype surrounding artificial intelligence (AI) and the fear of missing out (FOMO) that seems to be driving many enterprises to jump on the bandwagon.

It’s an exciting time, no doubt, with AI promising transformative changes in how we operate, analyse, and strategise. However, beneath the surface of this enthusiasm lies a concern that I believe we cannot afford to overlook.

For the past 2 decades, enterprises across industries have been diligently collecting and hoarding massive amounts of data, often under the banner of digital transformation.

This data accumulation was supposed to be the foundation for future innovation, a treasure trove to unlock insights and drive growth.

Yet, in many cases, this process has been carried out without a robust, well-thought-out foundational data strategy to ensure the data's quality, usability, and reliability.

Now, as every enterprise rushes to build and integrate AI capabilities into their operations, I see a significant problem emerging—and my concern centers squarely on the quality of the data being fed into these systems.

In my previous newsletter, I explored the concept of marketing operations as a system that often feels like a mysterious ‘Black Box’ to many professionals in the field.

I was referring to the intricate blend of art and science that transforms various inputs — such as budgets, creative strategies, and media plans — into measurable outputs and outcomes, like user engagement, reach, and ultimately, business success.

While the previous newsletter focused on the functional benefits of simplifying and upgrading marketing workflows, along with the inherent challenges that come with such transformations, we cannot ignore another equally critical aspect of this system: the quality of the data itself.

Data is the lifeblood of any marketing operation, and its quality directly impacts the effectiveness of every process, tool, and technology we deploy—including AI.

My perspective on this issue is straightforward, yet it carries implications for enterprises looking to harness AI effectively.

If the input data is unsanitized, unreliable, or, to put it bluntly, absolute ‘shit’ — think messy budgets, poorly structured plans, inconsistent creative strategies, fragmented media inputs, region-specific efforts that lack standardisation, agency back-and-forth that introduces errors, inadequate verification efforts, or haphazard reporting practices—then the outputs generated by your systems will inevitably reflect that poor quality.

You will end up with ‘shit’ data dashboard outputs, offering skewed or misleading insights into key metrics like share of eyeballs, user engagement, reach, or market penetration.

And, as a natural consequence, this can lead to ‘shit’ business outcomes —think diminished mindshare, ineffective purchase intent, missed sales targets, stagnant market share, and lacklustre quarter-over-quarter (QoQ) or year-over-year (YoY) growth.

No amount of AI, no matter how advanced, can magically clean up the mess.

Its business value depends on how good the quality of the data is.

While the hype surrounding AI often paints it as a game-changer—and in many ways, it has the potential to be—I find myself unable to shake off some deeper, more fundamental questions that enterprises need to grapple with.

• Are we at risk of missing out on the basics in our rush to adopt the latest technologies?
• Are we succumbing to the allure of shiny new objects—AI, generative AI, agentic systems—while neglecting to strengthen the very foundation on which our growth and success depend?

These are not just rhetorical questions; they are critical considerations that could determine whether our AI investments yield meaningful results or simply become expensive distractions.

In another the above newsletter, I drew a comparison that I think is worth revisiting here: I likened AI to a new pandemic, with data serving as the "virus food" that fuels its spread and impact. If the grassroots data—the raw, foundational data that enterprises collect and store—is not clean, consistent, and reliable, then the entire system is compromised from the start.

For enterprises today, ensuring that this foundational data is of the highest quality is not just a best practice—it’s an absolute necessity. Without a clean and solid foundation, any investment in advanced systems, including AI, risks being rendered utterly worthless. To me, the conclusion is crystal clear: there’s simply no point in building sophisticated systems on top of a shaky, unreliable base.

That brings me to the practical side of this discussion.

How can enterprises address this challenge and set themselves up for success in the AI era?

I’d like to share a decision framework that I believe can help guide organisations toward better data practices and, ultimately, more effective AI implementations.

1 ) First and foremost, focus on automating every process workflow in a way that ensures the data captured requires minimal or no cleaning.

This means investing in vertical-focused solutions that are tailored to the specific needs of your department or function, rather than relying on generic, one-size-fits-all ERP systems that often fall short in ensuring clean data capture. For example, use a marketing-specific tool for your marketing operations, a sales-specific tool for your sales team, and so on. By aligning your tools with the unique requirements of each domain, you can significantly reduce the risk of capturing messy or inconsistent data, setting a strong foundation for your AI initiatives.

2) Adopt a mindset of skepticism and diligence when it comes to your data—believe only what you see, and make it a habit to frequently validate your data.

Too often, enterprises take for granted that all the data they need is readily available and accurate, only to discover later that they’ve been operating under false assumptions. This is a common pitfall for companies that have been hoarding data for years without investing sufficiently in a foundational data strategy to ensure its quality and accessibility.

When these organizations finally start asking for specific data to fuel their AI systems, they often realize that critical pieces are missing, inaccurate, or unusable. Regular validation helps mitigate this risk and ensures that your data is truly ready to support your AI efforts.

3) Let’s talk about a step where many companies either overspend or completely neglect: building your own data warehouse.

If your organisation has already gone down this path, it’s imperative to ensure that all your data can be seamlessly stitched to common masters—standardised frameworks that allow for consistency and interoperability across your datasets.

However, if you haven’t yet invested in a data warehouse, there’s good news: in today’s AI-agentic era, you may not need to stitch everything together in a traditional sense.

Instead, you can explore building domain-specific agents on top of domain-specific data, creating a layered structure—like an agent pyramid—where higher-level agents can coordinate and leverage the outputs of lower-level ones. This approach can offer greater flexibility and scalability, allowing you to harness AI more effectively without getting negatively impacted by overly complex data integration efforts.

4) Finally, once you’ve laid this groundwork, it’s time to let AI work its magic.

You can set up AI-enabled workflows with Marketing Cloud Platform (MCP) integrations across your organization to streamline processes and unlock new insights. I’ll be honest—this step might require some significant up-skilling for your team, as AI technologies often demand new competencies and ways of working. But trust me when I say that this is an investment every individual and every organization needs to make in today’s rapidly evolving landscape.

The benefits of AI—when built on a foundation of clean, reliable data—are well worth the effort.

My hope is that this framework helps to temper the FOMO surrounding the AI hype that’s currently sweeping through the marketing operations space. More importantly, I hope it underscores the critical importance of clean, well-organised data as the bedrock of any successful AI strategy. Taking steady, deliberate steps to ensure the quality of data infuses confidence in AI initiatives and boosts the likelihood of success.

I’ll be back next week with more insights and ideas to share. Until then, happy marketing, and let’s keep building systems that truly deliver value!

Robin

__

P.S. I would like to give credit to Eduardo Ordax for the inspiration. He is the Lead Gen AI at Amazon Web Services (AWS) and his perspectives have had a strong influence in writing this article.