If you've logged into HubSpot recently, you'll have noticed AI popping up everywhere. Breeze this, AI-powered that, agents for everything. HubSpot has gone all-in on AI, and honestly, some of it's brilliant. But some of it's not there yet, and nobody seems willing to say that out loud.
We're HubSpot Gold Partners and we've been using the platform for years across dozens of client accounts. We've also built a hard-won reputation for being straight with people, even when that means saying "don't bother with that feature yet." So here's our honest rundown of what's actually useful in HubSpot's AI toolkit right now, what's a bit half-baked, and what's quietly becoming very interesting indeed.
First, a quick note on Breeze
HubSpot rebranded all its AI features as "Breeze" back in 2024, and they've been expanding it steadily since. Breeze isn't one thing. It's an umbrella term covering everything from the little AI writing helper that pops up in your email editor, right through to full-blown AI agents that can handle customer service tickets on their own. Some of these tools have been around for a while under different names, and some are new.
The trouble is that HubSpot's marketing doesn't make it easy to tell which bits are useful and which bits are still in "sounds impressive in a demo" territory. That's what we're going to try to sort out for you.
The features we'd actually recommend right now
Let's start with the good stuff, because there's quite a lot of it.
Predictive lead scoring
This is probably the single most valuable AI feature in HubSpot, and it's been quietly excellent for a while now. It looks at your existing contacts, analyses which ones have historically turned into customers, and then scores your new leads based on how closely they match those patterns. If you're a B2B company with a reasonable volume of contacts (we'd say north of a few hundred at minimum), this can seriously change how your sales team spends their time.
We've seen it work particularly well for engineering and professional services clients where sales cycles are long and leads come in from multiple channels. Instead of your sales team working through a list from top to bottom, they can focus on the leads that are statistically most likely to convert. It's not magic, and it does need enough historical data to be useful, but when it works, it works well.
The content assistant for emails and short copy
HubSpot's AI writing assistant (part of Breeze Copilot) is genuinely handy for getting past a blank page. We use it for drafting marketing emails, writing variations of ad copy, and putting together first drafts of landing page text. It's fast, it understands your HubSpot context (so it can pull in company names, product details, that sort of thing), and the output is a solid starting point.
The word "starting point" is doing important work in that sentence, though. You wouldn't publish what it gives you without editing. The tone tends to be a bit generic and overly enthusiastic (a problem with most AI writing tools, to be fair), and it doesn't know your brand voice. But as a way to get a first draft down in thirty seconds instead of staring at a screen for twenty minutes? Very useful.
AI-powered workflow actions
This one flies under the radar. HubSpot now lets you add AI steps into your workflows, things like "summarise this conversation" or "categorise this support ticket" or "extract the company name from this form submission." These are small, practical uses of AI that save real time and don't require you to trust the AI with anything high-stakes. If you're already using HubSpot's Sales Hub or Service Hub workflows, adding a few AI actions in is well worth experimenting with.
Customer Agent (for the right use case)
HubSpot's Customer Agent is an AI chatbot that can answer visitor questions using your knowledge base content. HubSpot claims it's resolving over 50% of support tickets for some customers, and we believe it, but with a caveat: it works best when you've got a solid, well-organised knowledge base for it to draw from. If your help documentation is thin or out of date, the bot will confidently give wrong answers, which is worse than not having a bot at all.
For B2B companies with a decent library of product documentation, FAQs, and support articles, it's a real time-saver. For everyone else, get the content in order first.
The features we'd hold off on
Not everything with an AI label is ready for prime time. Here's where we'd urge caution.
AI blog writer
Yes, HubSpot can now write blog posts for you. And yes, the output reads exactly like what it is: AI-generated content that's technically correct but has no personality and no real point of view. If you care about your content (and if you've read our blog about writing SEO-friendly content, you probably do), don't use this to publish finished articles. It might be useful for generating outlines or research summaries, but the finished product needs a human writer who knows your industry and has something to say.
We're perhaps biased here, given that writing good content is literally part of our job, but we've tested HubSpot's blog writer extensively and haven't been impressed enough to recommend it for any client's live blog.
Breeze Intelligence (the data enrichment credits)
Breeze Intelligence is the feature that enriches your contact and company records with third-party data (it's built on Clearbit, which HubSpot acquired). The data itself can be quite good, filling in job titles, company sizes, industry classifications, that sort of thing. The problem is the pricing model.
It runs on credits, and those credits disappear fast. We've heard horror stories of people accidentally enriching a large contact list and burning through thousands of pounds worth of credits in minutes. There are no spending caps or guardrails built in, and because HubSpot uses annual contracts, one expensive mistake gets multiplied across your whole subscription. If you're going to use it, set very clear internal rules about who can trigger enrichments and on which lists. Better yet, start with a small test batch and see whether the data it adds is actually useful for your specific contacts before committing budget.
Prospecting Agent
HubSpot's Prospecting Agent is designed to act as an always-on sales development rep, researching accounts, spotting buying signals, and reaching out to prospects. In theory, it's brilliant. In practice, the outreach it generates can feel templated and impersonal, which is exactly the opposite of what good B2B sales prospecting should be. We think this one will get better (HubSpot is clearly investing heavily in it), but right now we'd keep a human firmly in the loop rather than letting it send messages on its own.
Where MCP makes things more interesting
Here's where things get properly exciting, and where we think the real future of AI in HubSpot lies. It's not in HubSpot's built-in AI features at all. It's in connecting HubSpot to external AI tools through something called MCP.
We wrote a plain English guide to MCP (Model Context Protocol) recently, and if you haven't read it, it's worth a look. The short version: MCP is an open standard that lets AI tools like Claude and ChatGPT talk directly to your business systems, including HubSpot, in a secure, structured way.
HubSpot was actually the first CRM to launch an MCP server, which means you can now connect AI assistants to your HubSpot data and ask them things in plain English. "Show me all the deals closing this month over £50k." "Which contacts at this company have we spoken to in the last 90 days?" "Summarise the recent activity on this deal." That kind of thing.
Why does this matter? Because HubSpot's built-in AI features are limited to what HubSpot has decided to build. MCP opens the door to using whatever AI tool you want, connected to your real CRM data, without waiting for HubSpot to add a specific feature. And because MCP is an open standard (not proprietary to any one company), the same approach works with other systems too. Your AI assistant could pull data from HubSpot, cross-reference it with your project management tool, check your calendar, and draft a prep document for tomorrow's client meeting. All through one conversation.
We're already experimenting with this for clients and for our own business, and honestly, it feels like a much bigger deal than any individual AI feature HubSpot has shipped. The built-in features are convenient, but MCP is what makes AI properly powerful for companies that use HubSpot as part of a wider set of tools (which is everyone, really).
So where should you start?
If you're a HubSpot user wondering what to try first, here's what we'd recommend based on what's actually working across our client base.
Turn on predictive lead scoring if you haven't already. It's available on Professional and Enterprise tiers, it requires very little setup, and it gives your sales team an immediate, practical advantage. If you've got enough contact data for it to work with, it's the single best return on zero effort.
Start using the content assistant for email drafts and short-form copy, but always edit the output before it goes anywhere. Think of it as a fast first-draft tool, not a replacement for a copywriter.
Experiment with AI workflow actions for the small, boring, repetitive stuff: data formatting, ticket categorisation, conversation summaries. These are low-risk, high-value uses that save real hours over a month.
If you've got a solid knowledge base, try the Customer Agent for front-line support queries. If your documentation is patchy, sort that out first.
And keep an eye on MCP. It's early days, but it's the piece that turns HubSpot from "a CRM with some AI bolted on" into "a CRM that your AI tools can actually work with properly."
The honest summary
HubSpot's AI features are a mixed bag, and that's fine. It would be more worrying if every single feature were supposedly amazing, because that's never how new technology works. The predictive scoring, content assistant, workflow AI actions, and Customer Agent are all useful tools that can save you real time and help your team focus on the work that matters. The blog writer, data enrichment pricing, and fully autonomous prospecting aren't quite there yet.
The really interesting development isn't a HubSpot feature at all; it's MCP and the way it lets external AI tools connect to your CRM data. That's where we think the biggest shift is happening, and we're actively exploring it with clients right now.
If you'd like to talk about how to make HubSpot's AI features work for your business (or whether they're even the right thing to focus on), we're always happy to have a chat. No sales pitch, just an honest conversation about what makes sense for your situation. Drop us a line or book a 15-minute call, and we'll see if we can help.
Frequently asked questions
What are the best HubSpot AI features to use right now?
The most useful HubSpot AI features right now are predictive lead scoring, the Breeze Copilot content assistant for drafting emails and short copy, AI-powered workflow actions for automating repetitive tasks such as ticket categorisation, and the Customer Agent chatbot for companies with a solid knowledge base. Predictive lead scoring is available on the Professional and Enterprise tiers and requires minimal setup.
Is HubSpot Breeze AI worth it?
Parts of HubSpot Breeze are well worth using, particularly predictive lead scoring and AI workflow actions. However, some features, like the AI blog writer, produce generic content. Breeze Intelligence's data enrichment uses an expensive, credit-based pricing model with no spending caps, and the Prospecting Agent's outreach can feel impersonal. It's best to evaluate each Breeze feature individually rather than treating them as a single product.
Is HubSpot's AI blog writer any good?
HubSpot's AI blog writer can produce technically correct content, but it lacks personality, brand voice, and genuine insight. It's useful for generating outlines or research summaries, but the output needs significant editing by a human writer before publishing. For businesses that care about content quality and SEO, it shouldn't be used as a replacement for proper content creation.
What is MCP, and how does it work with HubSpot?
MCP (Model Context Protocol) is an open standard that lets AI tools like Claude and ChatGPT connect directly to business systems, including HubSpot. HubSpot was the first CRM to launch an MCP server, meaning you can connect external AI assistants to your CRM data and query it in plain English. This lets you use any AI tool with your HubSpot data rather than being limited to HubSpot's built-in AI features.
How does HubSpot predictive lead scoring work?
HubSpot's predictive lead scoring uses machine learning to analyse your existing contacts and identify which ones have historically converted into customers. It then scores new leads based on how closely they match those successful patterns. It works best for B2B companies with at least a few hundred contacts and is available on HubSpot Professional and Enterprise tiers.
