A practical guide to HubSpot AI prompted actions

If you're using HubSpot, you've probably noticed AI features popping up everywhere. One of the most interesting is HubSpot AI Prompted Actions, which lets you plug AI directly into your workflows. The idea is simple: automate all those little tasks like summarizing tickets or drafting follow-up emails.

But how well does it actually work? This guide will give you an honest look at what these AI actions can do, where they struggle, and whether they're the right tool for your team. We’ll also touch on another option if you find yourself hitting a wall.

What are HubSpot AI Prompted Actions?

Basically, HubSpot AI Prompted Actions are little AI helpers you can add to your workflows. You give them a simple text instruction (a "prompt"), and they perform a task on your contacts, deals, or tickets. It's all powered by HubSpot's own AI, called Breeze.

When you're building a workflow, you'll find them under the "AI" section. The main ones you'll see are:

• Data Agent: Custom prompt: This is the most open-ended one. You can feed it info from a record (like a deal or ticket) along with a custom instruction, and then use whatever the AI spits out in the next steps of your workflow.

• Data Agent: Research: This action is a bit more focused. It helps the AI pull out specific information from properties or even call transcripts tied to a record.

• Data Agent: Fill Smart Property: This one uses AI to fill in predefined "smart properties" by looking at the context of a record.

These tools are built to live inside the HubSpot ecosystem, using the data you already have to handle tasks that a person would otherwise have to do manually.

HubSpot AI Prompted Actions in the workflow builder.

Key features and capabilities

So, what can you actually do with these AI actions? Let's get into the specifics of how they help teams smooth out their daily grind.

Using custom prompts to manage data

The "Data Agent: Custom prompt" action is your go-to for custom jobs. You can tell it to look at the data in a record and generate something new. For example, you could ask it to:

• Give you a quick summary of the last customer email.

• Figure out if a support ticket is about billing or a technical problem.

• Pull a company name out of a messy block of text.

You then take that AI-generated text and use it to update a property, create a new task, or send the record down a different path in your workflow. It's pretty handy for standardizing data or adding a bit of context automatically.

Researching records and filling properties

The "Data Agent: Research" and "Data Agent: Fill Smart Property" actions are more structured. They’re less about free-form instructions and more about targeted data grabbing. You might use them to find a contact's industry from their email signature or pull key topics from a sales call transcript. The main goal here is to keep your CRM data clean and complete without anyone having to do it by hand.

HubSpot's Data Agent in action.

A typical workflow

Here’s what this might look like in the real world: a new support ticket comes in, which kicks off a workflow. The "Data Agent: Custom prompt" action reads the ticket's subject and description to guess if it's a "Billing," "Technical," or "Sales" question. Based on the AI's answer, an if/then branch sends the ticket straight to the right team.

graph TD  

A[New Support Ticket Arrives] --> B{Use 'Data Agent: Custom prompt' to analyze ticket content};  

B --> C{AI determines category};  

C -->|Billing| D[Route to Billing Team];
C -->|Technical| E[Route to Technical Team];
C -->|Sales| F[Route to Sales Team];

While these tools are a solid starting point for some basic automation, they have one major catch: they can only work with data that's already inside HubSpot. For many teams, that's a big problem.

Common use cases

When you use them for the right things, HubSpot's built-in AI actions can definitely save you some time. Teams are using them to add a bit of smarts to their sales, marketing, and service routines.

Here are a few common examples.

Use Case	Sample Prompt for "Data Agent: Custom prompt"	Why It's Useful
Qualifying Leads	"Based on this form submission, is this lead 'High,' 'Medium,' or 'Low' priority? Just give me one word."	Helps your sales team focus on the best leads right away instead of digging through a list.
Summarizing Deal Status	"Read the last three email notes. In 10 words or less, what's the buyer's next step?"	Gives sales managers a quick snapshot of the pipeline without reps having to write up manual summaries.
Formatting CRM Data	"The following text is a state name: "[State Property]". Please change it to its two-letter abbreviation."	Cleans up messy data from different sources so your reporting and lists are more reliable.
Drafting Initial Responses	"Write a friendly, 100-word outreach email based on this contact's industry ("[Industry Property]") and job title ("[Job Title Property]")."	Gives your reps a starting point, which can speed up how quickly they respond to common questions.

These examples show what's possible when you put AI right inside your CRM. But they also hint at the system's limits. The results are only as good as the data in your HubSpot properties and the AI's ability to understand what you're asking without any outside context.

Limitations and challenges

HubSpot's AI actions are a decent start, but it doesn't take long for teams to hit some pretty big roadblocks, especially when they try to do anything complex. A quick look at Reddit or HubSpot's own community forums shows a common theme: people need more.

1. Your knowledge is siloed

The biggest catch is that HubSpot's AI can only see what's inside HubSpot. But where does your team actually keep its important info? Probably scattered across Confluence pages, Google Docs, Notion databases, and random Slack threads. If the answer to a customer's question isn't in a CRM property, the HubSpot AI is stuck. This means it can't solve tricky problems and ends up escalating tickets that an AI probably should have been able to handle.

An infographic showing how unifying knowledge from different sources overcomes the limitations of siloed HubSpot AI Prompted Actions.

2. Lack of robust testing and simulation

How do you know if your brilliant AI prompt will actually work on thousands of real tickets? With HubSpot, you kind of just have to build it and hope for the best. You can test it on one record, but there's no way to see how it'll perform at scale. This can get risky. You might end up sending weird responses to customers or failing to automate tickets you thought were covered, all without a good way to see it coming.

3. Inflexible, black-box AI models

You're stuck with HubSpot's AI model, Breeze. You can't swap it out for a more powerful one like GPT-4o or Claude if you need it to handle more complex reasoning. You’re locked into their system, which might not be strong enough for what you need to do.

4. Confusing and unpredictable pricing

HubSpot uses a "Credits" system for its AI features, which can be a headache to track. As you automate more, you burn through credits faster, and you might get a surprise bill at the end of the month. The lack of a simple, flat rate makes it tough to figure out if your automation efforts are actually saving you money.

These issues often push teams to look for tools that are specifically built for AI automation.

Pricing for HubSpot AI Prompted Actions

Getting access to HubSpot AI Prompted Actions involves navigating specific subscription plans and that credit system, which is worth understanding before you go all-in.

• Plan Requirements: Workflows, which are necessary for AI actions, are mostly available on the Professional and Enterprise plans for Marketing, Sales, Service, and Operations Hubs.

• HubSpot Credits: Many of the more useful AI actions, especially the custom prompt one, use up HubSpot Credits. You get some with the higher-priced plans, but you’ll likely need to buy more as you use them.

• Operations Hub: If you need to do more advanced data formatting or want to run custom code, you'll probably need a subscription to Operations Hub Professional or Enterprise, which is another added cost.

All told, the pricing can get complicated and expensive, particularly if you're trying to automate a large chunk of your team's work.

A better alternative for advanced AI workflows: eesel AI

If you're nodding along to those limitations, it might be time to look at a tool built specifically for this stuff. That's where a platform like eesel AI comes in. It plugs right into HubSpot but gets around the common problems.

Here’s how eesel AI tackles the main challenges of HubSpot's native tools:

• Unify all your knowledge: Remember how HubSpot's AI is stuck in its own world? eesel AI connects to all your company knowledge in one click, including Confluence, Google Docs, Notion, and your past ticket history. This allows it to answer much tougher questions and automate a higher percentage of your tickets.

• Test with confidence: With eesel’s simulation mode, you can test your AI on thousands of your old tickets before it goes live. You get accurate predictions on how many tickets it will solve and can see exactly how it will respond, taking the guesswork out of the equation.

The eesel AI simulation feature, which addresses the lack of testing in HubSpot AI Prompted Actions by allowing users to test workflows on historical data.

• Total control and flexibility: eesel AI has a straightforward workflow builder that lets you choose exactly which tickets to automate, tweak the AI’s tone and personality, and even create custom actions that hook into other tools. You get full control without needing to write any code.

• Transparent and predictable pricing: No more confusing credits. eesel AI has clear, flat-rate plans with no hidden fees. Your costs are predictable, so you can scale up your automation without worrying about the bill.

A view of eesel AI's clear pricing page, an alternative to the credit system used for HubSpot AI Prompted Actions.

You can get eesel AI up and running in minutes, not months. It’s a self-serve platform that fits right into the helpdesk and workflows you already have.

This video demonstrates how to build powerful automated workflows using AI in HubSpot, similar to the concepts discussed.

Getting more from your HubSpot workflows: The verdict on HubSpot AI Prompted Actions

So, what's the verdict on HubSpot AI Prompted Actions? They're a handy tool for tidying up data and automating simple, internal tasks. If you just want to categorize a ticket or format a property, they can save you some time.

But if you're hoping to automate real, customer-facing conversations or handle tricky support questions, you're going to feel the constraints pretty quickly. By only using data inside HubSpot and missing key features like large-scale testing, they just don't offer the power needed for serious customer support automation.

If you're ready to do more than just basic data cleanup and want an AI that can learn from all your company knowledge and deliver real results, it's time to see what a dedicated platform can do.

See how eesel AI can boost your HubSpot workflows and automate your support. Start your free trial today.