AI in marketing sounds simple, plug it in, and watch productivity skyrocket. Reality? It’s a lot messier.

I recently sat down with a HubSpot consultant to unpack what’s really happening inside teams trying to adopt AI. From data chaos to tool limitations, this conversation revealed a pattern: AI doesn’t fail because of the tech. It fails because of the system around it.

TL;DR

• AI works best when your data is centralized inside HubSpot.
• Most teams struggle with AI adoption due to lack of understanding and poor data hygiene.
• Custom AI integrations offer more power, but come with higher cost and maintenance.
• HubSpot’s native AI (Breeze) is improving but still lacks speed and accuracy.
• Third-party AI tools like Claude are gaining traction for better performance and control.
• Bad data + automation = faster mistakes (and broken trust).

The Big Insight: AI Is Only as Good as Your Data

One theme came up repeatedly: context is everything.

AI inside HubSpot performs best when all your data, sales, marketing, product, lives in one place. When that happens, AI can actually understand your business and generate meaningful outputs. But that’s not how most companies operate.

In industries like manufacturing or SaaS with complex systems, critical data often sits outside HubSpot. Product data, pricing models, or CPQ systems are disconnected.

The result?

• Incomplete AI outputs
• Weak personalization
• Poor decision-making

This is where custom AI integrations come in, connecting external systems back into HubSpot to create a fuller picture.

More context → better AI → better ROI.

Why AI Adoption Fails Inside Teams

Even when the tech is ready, teams often aren’t.

Adoption tends to split into two groups:

• Early adopters who experiment and push boundaries
• Resistant users who don’t trust or understand the tools

And here’s the real issue: most people don’t understand how AI actually works with their data.

That leads to mistakes like:

• Broken workflows due to missing data fields
• Incorrect personalization in emails
• Over-reliance on AI outputs without validation

As highlighted in the conversation, “garbage in, garbage out” becomes “garbage in, amplified out” when AI is involved. The fix isn’t more AI, it’s better education and cleaner data.

The Hidden Cost of Custom AI

Custom AI sounds like the ideal solution and in many ways, it is.

You get:

• More flexibility
• Better integrations
• Richer insights

But there’s a tradeoff most teams underestimate: Maintenance. Custom integrations create what we like to call “information debt.”

• Data needs to stay accurate in real-time
• Systems need ongoing monitoring
• Teams need technical expertise to manage everything

If you don’t have the resources, things break, and when AI breaks, it breaks quietly. That’s why many teams still lean toward HubSpot’s native AI despite its limitations.

Where HubSpot’s AI (Breeze) Falls Short

Let’s be honest, HubSpot’s AI is improving, but it’s not there yet.

The biggest concerns we’re seeing:

• Slow response times
• Inaccurate or generic outputs
• Failure to fully utilize available context

Even when connected to knowledge bases or CRM data, results can feel… off.

And the risk is bigger than just inefficiency:

Most users won’t realize when the AI is wrong.

That creates a dangerous loop where bad outputs lead to bad decisions. This gap is exactly why more teams are exploring third-party AI tools.

The Rise of Third-Party AI in HubSpot Ecosystems

Instead of relying solely on native tools, teams are starting to integrate external AI platforms.

Popular choices include:

• Claude (for better reasoning and privacy controls)
• Notion (for content and project workflows)

What makes these tools appealing?

• More accurate outputs
• Faster innovation cycles
• Better control over data sharing

But again, these only work if implemented correctly.

Before You Add AI: Do This First

Most teams jump straight into tools.

The smarter approach is to prepare your system first:

• Understand your data model inside HubSpot
• Map what data each tool can access
• Define privacy boundaries (especially for sales data)
• Clean your CRM data before automation

Skipping this step is the fastest way to fail with AI.

Marketing Ops Still Matter (More Than Ever)

One underrated takeaway from the conversation:

AI doesn’t replace marketing operations, it makes them more important.

Successful teams are:

• Mapping workflows visually
• Aligning sales and marketing regularly
• Using tools like whiteboards and process diagrams

AI sits on top of these systems, it doesn’t fix them.

Final Thought: AI Is a System Problem, Not a Tool Problem

If your AI isn’t working, the issue usually isn’t the AI.

It’s:

• Your data
• Your processes
• Your team’s understanding

The companies winning with AI aren’t just adopting tools.

They’re fixing the foundation those tools rely on.

FAQs

1. Should I use HubSpot’s native AI or a custom solution?

Start with native AI if you want simplicity and low maintenance. Move to custom AI when you need deeper integrations and more control, but be ready for higher costs and complexity.

2. Why is my AI output inaccurate?

In most cases, it’s due to incomplete or poor-quality data. AI relies on context, if your data is fragmented or messy, outputs will suffer.

3. Is custom AI worth the investment?

Yes, if you have the resources to maintain it. Otherwise, it can quickly become a burden rather than an advantage.

4. How do I improve AI adoption in my team?

Focus on education. Help your team understand how AI works, how it uses data, and where it can fail.

5. What’s the biggest mistake companies make with AI?

Automating bad data. AI doesn’t fix problems, it scales them.