When the CEO of HubSpot talks about what’s holding companies back with AI, marketing operations leaders should pay attention. In a recent LinkedIn post, Yamini Rangan shared that nearly every customer conversation she has about AI eventually circles back to one thing: data.
Her argument is simple and, in our experience, completely correct. Messy data has always been a problem, but AI has raised the stakes dramatically. When humans worked with imperfect data, a rep would catch the duplicate account, an SDR would notice the strange email address, a campaign manager would spot the broken segment before launch. When AI agents act on that same data, nothing gets caught. Bad records become bad emails sent to real customers. Ambiguous fields become bad deal intelligence. And because agents operate at machine speed across connected workflows, errors don’t just occur — they compound.
The line from her post that every marketing leader should sit with: when your data lives in silos with unclear definitions and poor quality, “AI doesn’t fix that. It inherits it.”
The ARR problem lives in your marketing database too
Rangan’s example involves ARR — finance defines it one way, sales another, both legitimately, and an AI agent handed that ambiguity will confidently return the wrong answer every time. A human who understands the business has to decide which definition applies before any agent touches the data.
Now apply that same test to a typical B2B marketing database, and the problem multiplies:
What counts as an MQL — and did sales ever actually agree to that definition? Is “region” populated by the SDR team’s convention, the events team’s convention, or whatever the last list upload contained? Is that account one company or five, because it exists under five spellings across your CRM and marketing automation platform? When “industry” is blank on 40% of records, what does your AI-powered segmentation actually segment?
These are not technology problems. They are judgment problems — decisions about definitions, ownership, and standards that someone who understands the business has to make. And they are exactly the problems that quietly sabotage AI initiatives that looked great in the demo.
Why this work doesn’t get done
Rangan is right that this is where most companies fall short, and we’d add a reason why: data quality is invisible work. Nobody gets promoted for deduplicating a database. The work doesn’t demo well, it doesn’t make the board deck, and it never feels urgent — until an agent sends a pricing email to a churned customer, or routes your biggest inbound lead to the wrong rep.
There’s a second reason, one we see constantly: companies treat data quality as a project instead of an operation. They run a one-time cleanup, declare victory, and watch the database degrade again within two quarters. B2B contact data decays at 25–30% per year on its own. Every new list upload, every event import, every integration adds fresh inconsistency. Clean data isn’t a milestone. It’s a discipline.
What data readiness actually looks like
Rangan closes by advising business leaders to focus on data readiness first, and notes that an ecosystem of partners exists to help. This is precisely the work RightWave has done for B2B companies for over two decades, and it breaks down into three layers:
Definitions before automation. Before any AI touches your funnel, the human decisions get made: what an MQL means, how lifecycle stages are defined, which system is the source of truth for each field, who owns each definition. This is governance, and it’s the unglamorous prerequisite for everything else.
Cleansing at the source, not the symptom. Deduplication, normalization, enrichment, and standardization — applied where data enters your systems, not patched downstream after the damage spreads. A bad record fixed at upload never becomes a bad email.
Continuous operations, not one-time projects. Ongoing monitoring, automated hygiene workflows, and a marketing operations team that treats the database the way finance treats the ledger: something that must reconcile, always.
Companies that invest in these layers don’t just avoid AI failures. They get to outcomes faster, because their agents start from truth instead of inheriting two decades of accumulated inconsistency.
The boring work is the competitive advantage
The most important shift in Rangan’s post is one of framing. Data quality used to be a hygiene issue — important, ignorable. AI has turned it into a readiness issue. The companies winning with AI over the next few years won’t be the ones with the cleverest prompts or the most agents. They’ll be the ones whose data the agents can actually trust.
That work is human, it’s specific to your business, and there’s no shortcut for it. But it is entirely doable — and you don’t have to do it alone.
If you’re planning AI initiatives on top of a marketing database you don’t fully trust, start with a data readiness assessment. That’s the conversation RightWave has been having with B2B companies for 20 years — talk to us before your agents inherit the mess.
Reference – https://www.linkedin.com/posts/yaminirangan_every-customer-conversation-i-have-about-share-7470486525885976576-7XCq/
