The business world is falling head over heels for AI—and who can blame it? With promises to reduce grunt work, uncover insights, and turbocharge growth, artificial intelligence seems like the missing piece in every growth strategy.
But there's a catch.
No matter how clever your prompts, how embedded your AI agents are in your processes, it’s only as good as the data on which it feeds.
AI doesn’t work miracles. If your CRM is riddled with missing fields, duplicate records, disconnected tools, and unreadable notes, no amount of automation will save you. As Harvard Business Review puts it:
“AI is not magic. It’s math. And like all math, it’s only as good as the numbers you feed it.” — Harvard Business Review.
In other words, if your data is garbage, your results will be too—only faster.
Andrew Ng, co-founder of Google Brain, famously said:
“Data is the new oil—but like oil, it has to be refined.
In AI, the 'refinery' is your data design. Your CRM must be clean, structured, and regularly maintained.Tomasz Tunguz, venture capitalist and data-driven business writer, explains:
“The biggest bottleneck to AI adoption is not algorithms—it’s the lack of usable data.”
Most companies are sitting on piles of unstructured content—emails, call notes, PDFs, Slack threads—that their AI tools can’t interpret.
MIT Sloan found that:
“Companies that integrate AI into business processes only see results when they first improve data access, quality, and integration.” (MIT Sloan Management Review, 2023)
Your CRM might look “full” on the surface. But if the data inside it is:
Poorly labelled
Missing key properties
Disconnected from other tools
Stored in free-text notes...
… then your AI can’t make sense of it.
Equally important is unstructured data. Much of the power of AI lies in its ability to interpret unstructured information, such as call transcripts and email chains with prospects. However, for many organisations, that type of information lies scattered often across different applications. Exposing AI to all information, structured AND unstructured, gives it a much richer source to pull from.
| Type | Examples | AI Usability |
|---|---|---|
| Structured Data | CRM fields like industry, company size | Excellent for automation & insights |
| Unstructured Data | Emails, meeting notes, documents | Difficult to use—unless connected and indexed |
When AI tools like OpenAI’s GPT-4o and HubSpot Software can access both kinds of data—clean CRM fields AND connected unstructured content—they can deliver compelling, real-time recommendations, sales insights, and marketing suggestions that start to provide real benefits to your organisation.
The team is relieved of repetitive, busy work and can focus instead on higher-order problems, more creative problem-solving and connecting with their prospects and customers in a more trusted and authentic manner. The results are two-fold:
1) More work can be done in the same amount of time or by the same number of people.
2) The results from that work are better.
At G2M Solutions, we’ve seen first-hand how companies struggle to make AI useful—because they skip the hard part: cleaning up the data.
We believe an effective AI transformation happens in layers, not leaps and our HubSpot AI optimisation service, uses the following methodology:
Data Layer:
Start by cleaning and structuring your CRM data. Fix naming conventions. Fill in the missing fields. Remove duplicates. And equally importantly, start a regular maintenance regime.
AI loves clean inputs.
Connect Layer:
Bring in unstructured data sources—like meeting transcripts, documents, notes and email threads—so your AI has context.
Context = smarter insights.
Accelerate Layer:
With quality data, now start layering in AI and automation: lead scoring, predictive workflows, next-best-action nudges.
AI earns its keep.
Enablement Layer:
Finally, get your humans working with AI in their daily work. That means training, tracking use and adoption, and building new playbooks. There is a huge gap between AI's rapidly expanding capabilities and an organisation's ability to change to take advantage of it. Do not underestimate the sense of overwhelm your team will be feeling. Mining the power of AI is a long-term strategic commitment.
Human intelligence is augmented by artificial intelligence.
This approach doesn’t just make AI “work.” It makes it worth it.
If your sales reps are still switching between five tools to prep for a call, AI isn’t helping them.
If your marketing team can’t trust the numbers in your reports, AI can’t optimise campaign performance.
If your service team logs notes in ways AI can’t read, it won’t identify upsell triggers or churn risks.
AI thrives on clean inputs and it dies in silos.
Q: What’s the #1 reason AI projects fail in sales and marketing?
A: Bad data. Without consistent fields and naming conventions, your AI tools can’t find patterns.
Q: Can’t AI tools just “figure it out” with machine learning?
A: No. Machine learning isn’t mind-reading. It needs clear, structured examples to learn from.
Q: What kind of unstructured data should I connect to my CRM?
A: Emails, meeting transcripts, support tickets, proposal documents—anything that holds business context.
Q: Is this just a problem for big companies?
A: No. In fact, small-to-midsize businesses often suffer more because they have fewer resources to manage their data
Before you start investing heavily in AI and automation, ask yourself:
Can my CRM data be trusted?
Is my unstructured content accessible to AI tools?
Does my CRM have built-in AI and automation tools out of the box, such as HubSpot CRM?
Get those foundations right, and AI won’t just impress your board—it’ll transform your business.
Harvard Business Review – “If Your Data Is Bad, Your Machine Learning Tools Are Useless”
A practical deep dive into why data readiness trumps algorithm complexity in AI projects.
https://hbr.org/2021/02/if-your-data-is-bad-your-machine-learning-tools-are-useless
McKinsey & Company – “The data-powered enterprise of the future”
A comprehensive look at how data governance, architecture, and quality underpin successful AI implementation.
https://www.mckinsey.com/capabilities/mckinsey-digital/our-insights/the-data-powered-enterprise-of-the-future
MIT Sloan Management Review – “Building the AI-Powered Organization”
Especially useful for understanding how businesses can effectively integrate structured and unstructured data to make AI more practical.
https://sloanreview.mit.edu/article/building-the-ai-powered-organization/
IBM – “Structured vs. Unstructured Data: What's the Difference?” (2024)
A plain-English breakdown of what each data type is, and how AI can (or can’t) work with them.
https://www.ibm.com/topics/structured-unstructured-data
Gartner – “CRM and Customer Experience Primer for 2024”
Tracks how CRMs are evolving to support AI natively, and what foundational work (especially data-related) is required first.
https://www.gartner.com/en/insights/crm-customer-experience
HubSpot – “AI in CRM: How to Use Artificial Intelligence for Better Customer Relationships”
HubSpot’s official guide to integrating AI features like predictive lead scoring and email automation.
https://blog.hubspot.com/marketing/ai-crm
Andrew Ng – “Data-Centric AI” (Short Lecture & LinkedIn Post)
Ng’s key insight: improving data quality beats fiddling with models.
https://www.linkedin.com/pulse/data-centric-ai-andrew-ng/
Microsoft – “The Cloud-Scale Data Strategy”
For companies scaling AI: how to structure your data lake and data governance models.
https://techcommunity.microsoft.com/t5/azure-architecture-blog/cloud-scale-analytics-data-governance/ba-p/3892337