AI & Agent Hub · Diamond Solutions Partner

HubSpot agents your team will actually trust.

HubSpot AI that works: Customer, Prospecting, Data and custom agents, launched on clean data.

Insight Sales specialist beside an Agent Hub panel showing Customer Agent live, Prospecting Agent testing, Data Agent paused and credits at 42% of a 3,000 cap, with the HubSpot Diamond Solutions Partner badge

Sound familiar?

The agent is on. Nobody trusts it yet.

AI in HubSpot fails for three reasons, and none of them is the model.

  1. 01

    Wrong answers from bad data

    Empty properties and duplicates in, confident wrong answers out.

    How we handle it

    Data model and Context Home fixed first; agents go live on a portal that scores well.

  2. 02

    The credit bill surprised finance

    Agents bill per outcome, custom ones per action; nobody set a cap.

    How we handle it

    Caps, run limits, alerts and simulation set before day one.

  3. 03

    Switched on, never tuned

    Launched in a week, abandoned in a month; nobody owned it.

    How we handle it

    Thirty days of tuning on real cases, then handover or ongoing care.

Agent Hub

Every agent, what it needs, what it costs to run.

Names and prices change often; this is the lineup as of this month, and we keep it current.

  • Illustration of a HubSpot chat window with customer and agent messages and a Resolved tag

    Customer Agent

    Resolves support and sales questions on chat, email and voice from your knowledge base.

    Needs
    Service or Sales Hub Pro+
    Runs on
    Credits per resolved conversation
    We set up
    Knowledge, channels, handoff, approval gates
  • Illustration of a HubSpot prospecting record with Funding, Hiring and Visit signal tags and a Send button

    Prospecting Agent

    Researches accounts, watches buying signals and drafts outreach for reps.

    Needs
    Sales Hub Pro+
    Runs on
    Credits per recommended lead
    We set up
    ICP, signals, sequences, send rules
  • Illustration of a HubSpot record table with two rows tagged Enriched

    Data Agent

    Enriches and cleans records, answers questions about your data.

    Needs
    Any Pro+ hub
    Runs on
    Credits per answer
    We set up
    Enrichment, dedupe, property mapping
  • Illustration of a HubSpot content editor with a Brand voice tag and a Review button

    Content and Campaign Agents

    Drafts pages, emails and campaign plans in your brand voice.

    Needs
    Marketing or Content Hub Pro+
    Runs on
    Credits per asset
    We set up
    Brand voice, templates, review flow
  • Illustration of a HubSpot deals board with a card moving between stages

    Smart Deal Progression

    Moves deals and flags risk from activity on the record.

    Needs
    Sales Hub Pro+
    Runs on
    Core included, credits for actions
    We set up
    Stage rules, triggers, rep alerts
  • Illustration of an agent builder canvas on a dotted grid with connected steps and a Simulation tag

    Custom agents (Agent Builder)

    Your own agents on your data and tools: triage, routing, research, reporting.

    Needs
    Starter+ Builder, Pro+ triggers
    Runs on
    Credits per action unit
    We set up
    Scope, tools, guardrails, run limits

Breeze Assistant comes with every seat and uses no credits; we switch it on and train your team as part of any engagement. See HubSpot's current agent list →

Last reviewed: November 2026

Before anything runs

Caps, gates and an audit trail.

Every agent we launch ships with a monthly credit cap, run limits, alerts at 75 and 90 percent, human approval on actions that touch customers or money, and a log of what it did. Finance sees the number before the agent sees a customer.

See how we scope an agent
Dark HubSpot Credits panel for this month showing 42 percent used, 2,520 of 6,000 credits, usage per agent, an approval needed alert for the Prospecting Agent and an audit log line

The method

Ready, pilot, then live.

  1. Readiness

    Context Home score, data model, knowledge sources, permissions.

    What you get: a readiness report and a go/no-go.

  2. Scope and cap

    Use case, channels, guardrails, credit budget, simulation.

    What you get: a one-page agent spec with a monthly cost.

  3. Pilot

    Live on one channel or one team for 30 days, tuned weekly on real cases.

    What you get: resolution and accuracy numbers.

  4. Scale or stop

    Roll out, add agents, or switch it off and keep the data clean.

    What you get: a documented agent your admin can run, or a Managed Services plan.

Track record

AI in HubSpot trusted, kept.

  • 300+

    HubSpot implementations

  • 500+

    Projects delivered

  • 13 yrs

    On HubSpot

  • 25

    HubSpot certifications

  • 5.0

    Partner rating

  • “Extremely qualified professionals. We worked side by side with very positive results. Versatile, agile and up to date with best practices.”

    Malena Figueiredo

    Senior Product Manager at eureciclo

  • “Excellence in strategy and full HubSpot implementation, far beyond simple configuration. A human, collaborative approach that delivers concrete results quickly.”

    Tulio Ferreira

    Head of Sales at Estuda

  • “They navigate seamlessly between theory, practice and technical depth. Solutions that fit perfectly with our business needs. A truly hands-on company.”

    Raphael Santo

    Head of Sales & Customer Success B2B at Quero Educação

  • “Technical knowledge, accountability and lots of experience. Implementing HubSpot with them was easy. Professionalism above the average. Highly recommend!”

    Hari Aizemberg

    Marketing Manager at Koper

  • “A partner with complete command of the platform. A year after implementation, the partnership and support remain at the highest level.”

    Pedro Silva

    Technology Manager at Tambasa

  • “Insight showed genuine interest in understanding our product and our challenges. The quality of automations and integrations made all the difference. A truly top-tier partner.”

    Lucas Rocha

    Co-founder at Data Facts

  • “Nothing but great feedback! My biggest regret? Not implementing a CRM sooner. Insight Sales made all the difference.”

    Fabricio França Lima

    CEO & Founder at Power Tuning

  • “Super proactive support. We solved our challenges and have been partners for 5 years! Exceeding expectations from day one.”

    Henrique Vieira

    Managing Partner at Colonial Netimóveis

  • “It was a great experience. They helped us with everything from choosing the right plan to overcoming the challenges we faced and completing the onboarding process.”

    Felipe Dias

    Head of Sales at Teachy.ai

  • “Great experience, attentive and competent professionals. Highly recommend hiring them!”

    Pedro Del Negro Tayer

    GTM Data Engineering at Netlex

Common questions

HubSpot AI, answered.

What is Agent Hub, and what happened to Breeze Agents?

Agent Hub is the name HubSpot gave in July 2026 to the place where you find, switch on and manage the AI agents in your portal; what used to be called Breeze Agents now live inside it. Nothing you had configured was lost: the Customer Agent, Prospecting Agent and the others kept their settings and moved under the new name. Breeze Assistant, the in-app copilot every seat gets, kept its name and still uses no credits. Agent Builder, where custom agents are made, sits in the same Hub. We keep the catalogue on this page current as HubSpot renames things.

Which HubSpot plan do I need for agents?

Most agents need the Professional or Enterprise tier of the hub they belong to. The Customer Agent needs Service or Sales Hub Pro+, the Prospecting Agent and Smart Deal Progression need Sales Hub Pro+, the Content and Campaign Agents need Marketing or Content Hub Pro+, and the Data Agent runs on any Pro+ hub. Agent Builder is available from Starter, but most of the triggers it relies on, such as workflows, need Pro+. If you are on Starter, the readiness step tells you whether an upgrade pays for itself before you buy one.

How do HubSpot Credits work, and how do we stop overspending?

HubSpot Credits are a prepaid pool that agents draw from per outcome: a resolved conversation, a recommended lead, an answer, an asset, or an action unit for custom agents. Every portal gets an included monthly allowance and you buy more in blocks. We stop overspending with four controls set before launch: a monthly cap per agent, run limits per workflow, alerts at 75 and 90 percent, and simulation mode so a new agent shows what it would have spent before it spends anything. Finance gets the estimate in the agent spec, and the audit log shows where every credit went.

Which agent should we start with?

The one tied to a task with real volume, a clear owner and data you already trust. For most teams that is the Customer Agent on one support channel, because the knowledge base exists and resolution rate is easy to measure. Sales teams with a defined ICP usually start with the Prospecting Agent. If your properties are half empty, start with the Data Agent, since every other agent depends on what it fixes.

Can you build a custom agent for our process?

Yes, in Agent Builder, on your data and your tools. Typical builds are ticket triage, lead routing, account research before a call and weekly reporting summaries. We scope the trigger, the tools the agent may call, the guardrails and the run limit, then test it in simulation mode on real records before it acts on anything. Custom agents bill per action unit, so the spec states a monthly cost at your volume.

How long until an agent is live?

Two to four weeks to a live pilot on one channel or one team, followed by 30 days of tuning. Readiness takes about a week, scope and cap another, and the agent goes live as soon as the spec is approved. The variable is your data: a portal that scores well in readiness moves faster, one that needs a data model fix adds a sprint. Rollout to more channels or teams follows the pilot numbers.

Do agents need clean data first?

Yes. An agent reads the same properties, associations and knowledge your team does, and it answers with confidence either way. Context Home, HubSpot's view of what the portal knows about your business, is the first thing we check, alongside duplicate rates, required properties and association rules. If the score is low we fix the data model before any agent goes live. That is the readiness step, and it ends with a go or no-go.

Will the agent talk to customers without a human?

Only where you decide it should. The Customer Agent can resolve a conversation end to end, hand off below a confidence threshold, or draft for a rep to approve. We set approval gates on anything that touches money, contracts or an at-risk account, and the agent always identifies itself as AI. During the pilot every conversation is reviewed weekly, and handoff rules are tightened on what we see.

How much does it cost to have you set it up?

We price per agent: readiness, scope and cap, pilot and 30 days of tuning come as one package, with the credit estimate stated separately so you see both numbers. Custom agents in Agent Builder are scoped per build. After handover your admin runs the agent, or it moves into a monthly plan for ongoing tuning and cap reviews.

HubSpot Managed Services →

Can we use agents from outside HubSpot (n8n, Make, our own) alongside Agent Hub?

Yes, and many teams should. Agent Hub covers tasks inside HubSpot's data and channels; anything that needs your warehouse, another system of record or a model you choose runs better outside and writes back through the API. We build those with the same caps, logs and approval gates, so one governance view covers both.

GTM Engineering →

Talk to us

Which job should an agent take?

Tell us the task, the hub you're on and how many conversations or records a month. A specialist replies with a readiness check and a credit estimate.

Prefer to talk first? Book a 20-min call

Average reply time under 2 business hours.

All fields required