The HubSpot Agentic Platform: Where It's Been, Where It Is, and Where I Think It's Heading

Let me ask you something before we go one paragraph further.

Are you making decisions about AI agents in your HubSpot portal based on what's actually available today, or based on a keynote slide that said “over 20 agents” and made your stomach drop?

Because here's the thing. There's a metric butt ton of noise around the HubSpot agentic platform right now, and most of it is built to impress you, not to inform you. I want to do the opposite. I want to walk you through where this thing came from, where it honestly is today, and where I think it's heading, with my name on every prediction so you can hold me to it.

I'm not here to be the hero of this story. I'm here to help you become one. So let's do this in three acts.

And one promise up front: I run my own AI agents inside HubSpot. I moved my own digital clone into Customer Agent in about two hours one April afternoon. So this isn't theory I read on a vendor page. This is proof from the work.

Act I: Where We've Been

Before Breeze had a name, HubSpot was already quietly buying the pieces.

In 2023, HubSpot acquired Clearbit, a data enrichment engine sitting on a database of what HubSpot says is over 200 million buyer and company profiles. Around the same time, HubSpot co-founder and CTO Dharmesh Shah was building ChatSpot, a conversational AI layer for the CRM. Two ingredients. Nobody outside the building was calling it a platform yet.

Then INBOUND 2024 happened.

The INBOUND 2024 Launch (September 18, 2024)

On September 18, 2024, HubSpot launched Breeze, the brand for its entire agentic platform, alongside hundreds of product updates. Breeze arrived with three parts, and this is the framing that still holds today.

1. Breeze Assistant. An AI companion that makes the work easier. This is ChatSpot, grown up and folded into HubSpot. (It launched under an earlier name and is now simply Breeze Assistant, with web search, memory, file upload, and connections to the apps your team already lives in.)
2. Breeze Agents. Autonomous, drop-in agents that actually do work instead of just chatting about it. At launch there were 4: Content Agent, Social Media Agent, Prospecting Agent, and Customer Agent. Several shipped in beta or private beta.
3. Breeze Intelligence. Data enrichment and buyer intent, built on that Clearbit engine. Enrich a record, spot buying intent, shorten a form so a known visitor doesn't have to retype what HubSpot already knows.

I remember exactly where I stood on AI agents back then, because we'd just recorded our 100th HubHeroes episode about HubSpot Breeze, AI, and the co-pilot mindset. My take then is my take now: this stuff is only as good as the human strategy wrapped around it. A smart agent pointed at a messy system just makes mess faster, and now it makes it faster in real time.

Spring 2025: The Lineup Gets Refined (April 10, 2025)

At the Spring 2025 Spotlight on April 10, 2025, HubSpot tightened the story to four agents on a page literally titled “Four AI agents that help teams scale”: Customer Agent, Content Agent, and Prospecting Agent (all available globally), plus a new Knowledge Base Agent in private beta.

Notice what's missing? The Social Media Agent from the launch lineup. That's the first clue to something we'll come back to in Act III. The agent count on a slide and the agent count in production are two different numbers, and that gap matters more than any feature.

INBOUND 2025: The Blueprint (September 3, 2025)

On September 3, 2025, HubSpot went big. The headline was a “blueprint for building hybrid human-AI teams,” wrapped around 200+ product updates. HubSpot announced “over 15 Breeze Agents across marketing, sales, and service,” with 18 more in beta on top of that. It shipped a Breeze Marketplace (pre-built agents you can grab), Breeze Studio (where non-developers customize agents), Custom Assistants for specialized experts, and reframed the whole platform around three jobs: connect your data, enable your people, build your AI team.

This is also where HubSpot started calling itself, in its own investor language, “the agentic customer platform for scaling businesses.” Read that again. The HubSpot agentic platform isn't a feature HubSpot ships. It's how HubSpot now describes what HubSpot is.

So that's where we've been. A data acquisition, a companion, a launch, a refinement, and a blueprint. Three years, one clear direction of travel.

Now let's talk about what's actually real today, because that's where it gets useful.

Act II: Where We Are Right Now

Here's the move that separates the humans who get value from the HubSpot agentic platform from the humans who get oversold: learn to read the map.

HubSpot's AI offerings confuse people because the names blur together. So let me give you the clearest version I've got, the same one we taught on HubHeroes when we mapped Breeze versus Customer Agent.

Breeze AI (the Assistant and the internal agents) helps you, the human working inside HubSpot. Customer Agent faces outward, helping the humans you serve. One looks in. One looks out. Get that straight and 80% of the confusion evaporates.

The Agents That Are Actually Generally Available

As of June 2026, three agents are truly generally available, which is the industry's way of saying “shipped, supported, and ready for everyone,” not “announced.” Three are real and ready.

1. Customer Agent. HubSpot calls it your “AI front office concierge.” It answers questions, resolves issues, and turns conversations into opportunities. This is the outward-facing one.
2. Prospecting Agent. A 24/7 BDR that monitors prospects for buying signals, researches accounts, and personalizes outreach.
3. Data Agent. Researches and answers custom questions about your records, like which prospects are using a competitor's product.

There's a longer list in beta: Customer Health Agent, Company Research Agent, a Closing Agent for CPQ, plus Knowledge Base, Personalization, ABM Landing Page, and Blog Research agents. Those are coming. They're just not here-here yet. Hold that thought, because it's a whole prediction later.

Breeze Studio, the Marketplace, and the HubSpot CLI

If you don't want to wait for a pre-built agent, you've got two on-ramps.

For the non-developers, there's Breeze Studio and the Breeze Marketplace, where you can customize an existing agent or grab a community-built one for a common task. No code required.

For the developers and the vibe coders, HubSpot ships two MCP servers (Model Context Protocol). One is a remote server that gives any MCP-compatible AI tool secure read and write access to your CRM data. The other is a local developer server. And the front door to that builder world is the HubSpot CLI, the hs command line tool (version 8.8.0 as I write this). You install it with npm install -g @hubspot/cli, run hs init to connect your account, and run hs mcp setup to wire the local Developer MCP server straight into tools like Claude Code or Cursor. Translation for the non-coders: a developer on your team can now talk to HubSpot in plain English from inside their code editor and have it scaffold, test, and ship real projects.

That layer matters more than it looks. It means your agents don't have to live inside HubSpot's walls. They can reach out and do things. Which is exactly the upgrade I felt when I moved my own clone.

The Pricing Model That Actually Changed Everything

I'll be real honest with you. The single most important thing that happened to the HubSpot agentic platform wasn't an agent. It was a price tag.

As of April 14, 2026, HubSpot moved Customer Agent and Prospecting Agent to outcome-based pricing, settled in HubSpot Credits ($10 per 1,000 credits). HubSpot's own line: “Now you pay when the task is complete.” HubSpot's Chief Customer Officer Jon Dick put it even flatter: “You pay when it works, full stop.”

And here's a piece most people miss: you don't start from zero. Every plan comes with a monthly bank of credits included before you ever buy more. As of June 10, 2026, Starter includes 500 credits a month, Professional includes 3,000, and Enterprise includes 5,000. Burn through the bank and you add capacity packs at $10 per 1,000 credits per month. So for a lot of teams, the first chunk of agentic work each month is already paid for.

Here's the per-action math, as of June 10, 2026 (and yes, it has already changed once and it will change again, so check it before you budget):

• Customer Agent: $0.50 per resolution. A “resolution” means the agent handled it and a human didn't have to step in for 72 hours.
• Prospecting Agent: $1.00 per lead recommended for outreach.
• Data Agent: $0.10 per answer.

How HubSpot Stacks Up (The Competitive Reality)

HubSpot isn't the only one selling outcome-based agents. Here's the landscape as of April 2026, mostly per SaaStr's reporting, cross-checked against the vendors.

The Honest Skeptic's Note (Because I Won't Sell You a One-Sided Story)

If I tell you this is all upside, I'm selling you something, and I don't do that.

The sharpest pushback comes from Jason Lemkin at SaaStr, and it's worth borrowing. His argument: HubSpot is following on outcome-based pricing, not leading. Sierra baked it in from day one back in early 2024. Intercom built Fin on $0.99 per resolution before HubSpot moved.

And here's the part every HubSpot Super Admin needs to internalize: the gap between “pay per resolution” and “pay per conversation” shrinks as the AI gets better. When agents only resolved 25% of conversations, paying per resolution was a real discount. But when resolution rates climb toward 90% (where Sierra's best deployments already sit, and where Gartner projects 80% of common issues will get resolved by 2029), per-resolution and per-conversation pricing become almost the same thing. The discount quietly evaporates.

There's also the CFO headache. Variable bills sound friendly until they show up. A real HubSpot deployment of 5,000 conversations a month at a 65% resolution rate runs around $1,625 a month. Push resolution to 90% and it climbs past $2,250 and keeps climbing with volume. “You pay when it works” also means “you pay more when it works more.”

But credit where it's due, and Lemkin gives it too: at $0.50 per resolution, HubSpot is genuinely the cheapest option in the market right now. For a small or mid-size team that got burned by a dumb chatbot two years ago, that price plus a monthly credit bank plus a 28-day trial isn't really about saving money. It's about removing fear. That's a trust play dressed up as a pricing play, and it's a smart one.

My Lived Proof: Digital George and the FEAM

Let me get specific, because abstract is easy and specific is honest.

I ran an AI clone of myself on Delphi.ai for over a year. It answered HubSpot questions in my voice, it took phone calls, it coached humans through my frameworks. Good workhorse. But it was an island. It couldn't see my contacts. It couldn't tell whether the human typing was a paying partner or somebody who downloaded a PDF five minutes ago. Every conversation started from zero.

In April 2026, I moved Digital George into HubSpot Customer Agent. About two hours of config work. And the upgrade wasn't the chat box getting smarter. It was the chat box getting connected. Now the brain lives where the business lives, with the content, the CRM, and the customer context all in one house. The Delphi clone answered questions. Customer Agent answers questions and acts: it can book a call, pull an article, log the whole conversation to the contact timeline. Smart chat box became smart teammate.

That move taught me something that shapes my whole take on this platform: the HubSpot agentic platform is only as good as the system it's plugged into. Garbage in, garbage out. Always has been, always will be, now just faster.

And here's the proof that I actually believe the hybrid human-AI thing instead of just repeating it: at Sidekick Strategies, we run a FEAM, our Family plus tEAM, where the team is humans and agents. The agents handle the repeatable production work. The humans handle judgment, relationships, and the work only a human can do. The agents draft. The humans decide. Nobody on my human team got replaced. They got amplified. That's the model I'm betting the future on, which is a perfect handoff into Act III.

Act III: Where I Think It's Heading (The Crystal Ball)

Okay. Crystal ball time.

What follows is my point of view, not fact. These are George's predictions, with my name on them, made on June 10, 2026. I could be wrong on any of them. But I've been the HubSpot guy for 12-plus years and I've watched this company telegraph its moves for over a decade, so I'm willing to go on the record. Each prediction is tied to a real signal from Act II, and each one runs through my Four-Part Philosophy: Value-First, Human-First, Content-Driven, AI-Assisted. Because that's the lens I judge every tool through. I'm not optimizing AI to replace humans. I'm optimizing it to augment them.

Here's where I think the HubSpot agentic platform goes. Seven predictions.

Prediction 1: Credits Become the New Seat (And Super Admins Become the Stewards of Spend)

The signal: HubSpot's move to outcome-based Credits on April 14, 2026, with a monthly credit bank baked into every plan.

The prediction: Within 18 months, the most valuable governance skill for a HubSpot Super Admin won't be building workflows. It'll be forecasting and governing agentic spend. Credits behave like cloud-consumption billing, monthly bank and all, and that's a brand-new mental model for HubSpot teams who've only ever budgeted in seats. The admin who can answer “what's our cost per resolution this quarter, are we using our monthly bank, and is the trend going the right way” becomes the most important human in the room.

Through the philosophy: This is Human-First. The point of governing spend isn't to penny-pinch the robots. It's to protect the humans whose budgets and jobs are tied to that number. Somebody has to be the steward, and it should be a human with judgment, not an auto-renew toggle.

Prediction 2: The “20+ Agents” Number Collapses Into the Handful That Matter

The signal: “Over 15 Breeze Agents,” 18 more in beta, but only about 3 truly generally available today. And the Social Media Agent that quietly vanished between launch and Spring 2025.

The prediction: The marketing count keeps climbing toward “30+ agents.” The number that actually ships to everyone grows slowly and deliberately. HubSpot will keep some agents in beta for a long time, kill a few quietly, and consolidate others. The teams who win will treat “announced” and “production-ready” as two completely different words, and they'll plan around the second one.

Through the philosophy: This is Content-Driven and Value-First. My job, and your job if you lead a team, is to cut through the count and teach people what's real. Clarity is a gift. Hype is a tax. Give the gift.

Prediction 3: “Pay When It Works” Gets Re-Priced Within 12 to 18 Months

The signal: Lemkin's point that per-resolution and per-conversation pricing converge as resolution rates climb toward 90%.

The prediction: HubSpot's $0.50-per-resolution headline is a 2026 wedge, not a forever price. As resolution rates rise across the customer base, the economics shift under HubSpot's feet, and they re-price again, probably toward tiered bundles, volume floors, or a blended per-conversation model. Whatever the cheapest-in-market story is today, re-pull your spreadsheet every two quarters. The math that says “yes” this month might say “renegotiate” next year.

Through the philosophy: This is AI-Assisted done with eyes open. AI is an amplifier, and amplifiers get cheaper and then get re-priced. Don't fall in love with a number. Fall in love with the outcome and keep checking the receipt.

Prediction 4: agent.ai Stays Independent, But Becomes HubSpot's R&D Radar

The signal: Dharmesh Shah's agent.ai network passed 2 million humans by INBOUND 2025, with 26,000 of them building their own agents. ChatSpot got absorbed into HubSpot and became Breeze Assistant. agent.ai didn't get absorbed.

The prediction: agent.ai stays a separate company, not a HubSpot product. But HubSpot uses it as a free, live signal source: when an agent type gets popular with 2 million-plus humans, HubSpot builds its own version into Breeze. One Dharmesh project (ChatSpot) got built in. The other (agent.ai) becomes the radar that tells HubSpot what to build next. Watch what trends on agent.ai to predict the next Breeze Agent.

Through the philosophy: This is Value-First thinking about build-versus-bet. HubSpot doesn't have to be first. It has to be useful to a 10-person team. agent.ai tells them what “useful” looks like at scale before they spend a dollar building it.

Prediction 5: Brand Voice Becomes the Most Important Field in Your Portal

The signal: When I moved Digital George into Customer Agent, the single highest-impact setting wasn't the agent config. It was Brand Voice, the upstream setting every HubSpot AI product reads.

The prediction: As agents multiply across marketing, sales, and service, the Brand Voice setting goes from a nice-to-have to the governor of your entire AI output. One field, filled in once, shaping how a dozen agents sound to your humans. Teams that nail it sound like themselves at scale. Teams that skip it sound like a generic robot wearing their logo. I think HubSpot leans into this hard and makes Brand Voice the control tower for agentic tone.

Through the philosophy: This is Human-First and Content-Driven at the same time. Your voice is how humans recognize you. Protecting it across a swarm of agents is how you stay human at scale instead of disappearing into the sea of beige AI everyone else is producing.

Prediction 6: Identity-First Agents Are the Next Frontier (And This One's an Open Invitation)

The signal: Brand Voice is a great start, but a few sliders and a paragraph of tone notes can only take an agent so far. When I rebuilt Digital George, the thing that made humans say “that actually sounds like George” wasn't a setting. It was an identity.

The prediction: The next leap for the HubSpot agentic platform isn't more agents. It's deeper agents. We move from brand voice to full identity: not just “sound upbeat and use our words,” but a structured, persistent sense of who this voice is, what it believes, the stories it tells, the things it would never say. The platforms that crack identity will make agents that feel like a real human echo or mirror instead of a polite stranger reading your FAQ.

Here's where I'll be vulnerable with you, and honestly, HubSpot, if you're reading this, consider it an open invitation. We've built an identity architecture off-platform at Sidekick. We call the heart of it ECHO, and it runs on a system of written identities: voice, values, signature moves, stories, the banned words and the favorite ones. It's the reason Digital George sounds like me and not like a chatbot cosplaying as me. I'm not going to hand over the secret sauce in a blog post. But I'd love to sit down with the HubSpot team and talk about Identities and the power of that thinking, because I think it's the missing layer between “AI that uses your words” and “AI that carries your soul into rooms you can't physically walk into.” Call me, HubSpot. Let's build the human-sounding part together.

Through the philosophy: This is Content-Driven and Human-First taken all the way down. Brand voice protects how you sound. Identity protects who you are. If we're going to put agents in front of the humans we serve, they'd better carry more than our vocabulary. They'd better carry our heart.

Prediction 7: The Winning Teams Won't Have the Most Agents, They'll Keep the Most Humans in the Loop

The signal: HubSpot's own framing. A “blueprint for hybrid human-AI teams,” not a blueprint for human-free teams. And my own FEAM, where the agents amplified my humans instead of replacing them.

The prediction: Two years from now, the case studies that actually matter won't be “we replaced our support team with AI.” They'll be “we kept our team, gave each human an agent, and served more humans better than we ever could before.” The teams that treat agents as drop-in human replacements will quietly bleed trust and customers. The teams that treat agents as sidekicks for their humans will flourish.

Through the philosophy: This is the whole thesis. The human side of business is the side that wins. AI handles persistence, memory, and speed. Humans bring purpose, empathy, and care. Put them together and you create experiences that feel personal at scale. Pull the humans out and you've just automated mediocrity faster.

That's my crystal ball. Seven predictions, my name on all of them, made June 10, 2026. Screenshot it. Come find me in two years and tell me where I was right and where I was a guy who guessed wrong. I'm good with either, because being willing to be wrong out loud is how you earn the right to be trusted.

So Here's the Bottom Line

The HubSpot agentic platform is real, it's moving fast, and it's pointed in an obvious direction: HubSpot wants to be the place where small and mid-size teams run hybrid human-AI work without needing a data-science department to do it.

But a platform is just a playground. What you build on it is up to you. The agents are only as smart as the system, the strategy, and the humans behind them. That's the part no keynote will tell you, and it's the part we obsess over.

So here's your next step. Don't go add ten agents to your portal this afternoon. Go pick the one task that drains your team the most, the repetitive thing everybody complains about, and ask whether a single, well-governed agent could carry it. Start there. One trigger, one outcome, one win. Earn trust, then expand. (If you want the tactical version of that play, here's the practical guide to less busywork and more human work.)

And if you want a human to think through your agent strategy with you, that's literally what we do. Book a free strategy call and let's map it together. Not to sell you a platform you already own. To help you make it work for your humans instead of the other way around.

One question to sit with before you go: a year from now, do you want your team buried under AI you don't understand, or amplified by agents you actually trust? You get to choose. Choose like it's yours.