If you're a marketing agency, founder, or RevOps engineer trying to implement AI inside HubSpot, you’ve probably felt this already:

AI sounds powerful… but the results? Meh.

Workflows break. Data gets messy. Outputs feel generic. And suddenly, AI feels like more work instead of less.

We recently sat down with Trisha Merriam, a marketing expert who has worked extensively with HubSpot and AI integrations, to unpack where things go wrong — and more importantly, how to fix them.

TL;DR (For Busy Teams)

• AI fails when your data and processes aren’t documented
• CRM hygiene alone is not enough for AI success
• Poor prompts = poor outputs (and sometimes hilarious disasters)
• Start with HubSpot’s native AI (Breeze) before jumping to third-party tools
• Use AI daily + learn in cohorts to improve faster

The #1 Reason AI Fails in HubSpot: You’re Not Ready for It

Most teams think they are ready for AI because their CRM is “clean.”

They’ve got:

• First names ✔
• Email addresses ✔
• Company names ✔

But according to Trisha Merriam from SWOPtimize, that’s only scratching the surface.

AI doesn’t just need clean data. It needs context.

That means documenting:

• Your Ideal Customer Profile (ICP)
• Your sales process
• Your marketing workflows
• Your product messaging
• Your business summary

If this isn’t documented, AI simply doesn’t have enough information to work with.

And when AI lacks context, it doesn’t fail loudly, it fails quietly with mediocre results.

“AI Didn’t Work for Us,” Here’s What That Really Means

You’ve probably heard this before:

“We tried AI… it didn’t really work.”

In most cases, the problem isn’t AI.

The problem is missing inputs.

As Trisha explained, when teams don’t provide enough context, AI outputs feel generic, irrelevant, or just plain wrong.

So instead of fixing the inputs, teams abandon the tool. That’s like blaming Google Maps… without entering a destination.

A Real Example: When AI “Fixes” Your Data (And Breaks It)

One of the most interesting moments in our conversation was when Trisha shared a real-world example. She used AI (Claude) to clean up a messy contact list before importing it into a CRM. Sounds smart, right?

Until this happened:

• “Jose Diego” → became “Jerry Douglas”

Yes. AI literally changed people’s names. Why? Because there were no guardrails defined.

Once she added a simple rule:

“Never change first name, last name, job title, or email”

The issue was fixed in seconds.

Lesson: AI is powerful, but without constraints, it can confidently do the wrong thing.

Guardrails > Tools

Most teams jump straight to tools.

But the real order should be:

1. Documentation
2. Prompts
3. Guardrails
4. Then tools

If you skip the first three, the tool won’t save you.

Some simple guardrails to start with:

• Never modify critical CRM fields (name, email, company)
• Always define output format
• Provide clear context about your business
• Validate outputs before importing into CRM

The Most Underrated AI Skill: Using It Every Day

Here’s something most people don’t talk about:

AI is a skill, not a switch.

According to Trisha, the teams that succeed are the ones that:

• Use AI daily
• Experiment constantly
• Learn from failures quickly

Because honestly?

A lot of things don’t work at first.

But over time, you start understanding:

• What to ask
• How to ask
• What works (and what doesn’t)

Why Learning Alone Slows You Down

One of the most practical insights from this conversation:

Don’t learn AI alone. Trisha recommends creating small “learning cohorts,” groups of people at the same stage in their AI journey.

Why this works:

• You learn from others’ mistakes
• You avoid wasting time on things that don’t work
• You discover better workflows faster

Think of it as shared experimentation.

HubSpot Breeze AI vs Third-Party Tools (Like ChatGPT & Claude)

Let’s address the big question.

Should you use HubSpot Breeze or external AI tools?

Here’s the honest answer: Start with Breeze.

Why?

• It already has access to your CRM data
• Better for compliance and security
• No setup required
• Quick wins for internal insights

For example, Breeze can:

• Identify cross-sell opportunities
• Summarize product information from your CRM
• Answer questions about your data instantly

That said… it’s not perfect.

It’s improving fast, but still a bit of a “mixed bag.”

Best use case: Understanding your data

Not ideal for: Complex workflows or advanced automation

When Should You Use Third-Party AI Tools?

Once you:

• Understand your data
• Have documentation in place
• Know how to write prompts

Then tools like ChatGPT or Claude become incredibly powerful.

But jumping to them too early?

That’s where most teams go wrong.

A Smarter Way to Implement AI in HubSpot

If we had to simplify everything into a process, it would look like this:

1. Document your business (ICP, processes, messaging)
2. Clean your CRM data
3. Define guardrails
4. Start with HubSpot Breeze
5. Use AI daily and build internal habits
6. Collaborate and learn in groups
7. Then scale with third-party tools

FAQ: AI in HubSpot CRM

Why is my AI output poor in HubSpot?

Most likely because your data and processes are not properly documented. AI needs context, not just clean fields.

Is HubSpot Breeze enough for AI implementation?

It’s a great starting point. Especially for data insights and quick wins. But you may need third-party tools for advanced use cases.

What are the biggest mistakes teams make with AI?

• Skipping documentation
• Not setting guardrails
• Using AI inconsistently
• Jumping to tools too early

How do I improve AI results in my CRM?

Provide better inputs: clear documentation, structured prompts, and defined constraints.

Final Thoughts

AI in HubSpot isn’t magic. It’s leverage. And like any leverage, it only works when the foundation is strong. If your processes are messy, AI will amplify the mess. But if your foundation is solid?

AI becomes a force multiplier for your entire revenue engine.

Big thanks to Trisha Merriam for sharing her real-world insights and experiences.