HubSpot's Customer Agent:
a new colleague

HubSpot's Customer Agent draws on your knowledge base to answer some customer service requests automatically and hand the rest off to a human.

At Guépard, we deploy this agent to get measurable results.

What is HubSpot's Customer Agent?

It's an agent that handles certain customer service requests based on their nature and on criteria you define. Think of it like training a new employee and defining their role.

It relies on your content

The agent handles everyday requests using your knowledge base and points each person to the right information. It's configured to avoid hallucinations.

  • Since it's connected to your CRM, it can recognize the person and answer based on their record.
  • It can query other software through APIs, depending on your ecosystem (e.g., Shopify).

It follows your processes

The conditions under which a request should be taken on or taken over by a person are built into the agent's settings: humans stay at the heart of sensitive conversations.

  • It knows your organization's nuances and can tell when a customer wants to talk to a human.
  • You set the rules (e.g., refunds and complaints always go to the team).

L'agent IA :
un exosquelette tentaculaire

Bon, d'accord : l'image d'un exosquelette tentaculaire est plutôt cauchemardesque. L'idée, c'est qu'il s'agit d'une extension permettant aux employés de faire plus de choses en même temps, et avec plus de force : un plus grand nombre de suivis plus personnalisés et plus serrés, une information obtenue en un instant plutôt qu'en des heures de recherche, et ainsi de suite.

Bref, un bon agent IA décuple votre potentiel.

Delegate to your agent flexibly and strategically

The AI agent

"Where's my order?" Live chat
"How do I set up feature X?" Email
"I can't decide between plans" Email

The person

Refund request → Team X
Angry customer who wants to be heard → Team Y
Complex file to investigate → Team Z
Assistance — your store Answers in a few seconds
Hello, I placed my order 3 days ago (C-4812). Where is my order?
Checking Shopify…
Hi! Your order C-4812 left the warehouse yesterday and will be delivered Thursday.
Resolved without human intervention · 12 s
LOGO-GUEPARD-Favicon-RGB-1
Customer assistance to Sophie Belanger — RE : configuration

Hi Sophie,

Thank you for your message! To activate this functionality: go to Settings → Integrations, then click on Connect.

It should take around two minutes to activate and won't affect your existing data. Here is an illustrated guide if you would rather follow there.

Let me know if anything is unclear, I'm available!

Answer sent based on knowledge base Resolved
LOGO-GUEPARD-Favicon-RGB-1
Customer assistance to Marc Tremblay — RE : what plan to choose

Hi Marc,

The Growth plan includes multi-channel tracking; the Pro plan adds advanced reporting and API access. For a team your size, most of our clients start at Growth and then scale as needed.

The best option is typically to talk about it for 15 minutes. Would you want me to reserve a demo with one of our team members this week?

Qualified lead sent to the sales team Demo proposed

How Customer Agent works,
step by step

Analysis and understanding

As soon as a request comes in by chat, email or phone, the agent grasps the context and what's being said to pin down the real need behind the question.

  • It picks up on intent, even when it's awkwardly worded.
  • It recognizes when someone wants to talk to a human.
  • It's connected to the CRM, so it knows who it's talking to.

Applying the established process

If the request falls within its scope, the agent replies in your brand's voice. If not, it hands it off cleanly to the right person.

  • It can instantly answer everyday requests.
  • It can transfer to a human based on your rules (e.g., refunds or complaints).

Writing a reply, if applicable

The agent draws on your knowledge base and, when needed, queries external tools to give an accurate, specific answer.

  • It's programmed to avoid hallucinations.
  • It can query an external tool via API.
  • It can enrich the customer's record in your CRM.

Concrete result:
most requests are handled
automatically

Real-world usage data currently puts autonomous resolution at around 65% to 70% of conversations. That automation gives teams time back for the cases that truly need a human.

65-70 %

OF REQUESTS RESOLVED ON THEIR OWN

-54%

FIRST RESPONSE TIME

L'agent IA :
un exosquelette tentaculaire

Bon, d'accord : l'image d'un exosquelette tentaculaire est plutôt cauchemardesque. L'idée, c'est qu'il s'agit d'une extension permettant aux employés de faire plus de choses en même temps, et avec plus de force : un plus grand nombre de suivis plus personnalisés et plus serrés, une information obtenue en un instant plutôt qu'en des heures de recherche, et ainsi de suite.

Bref, un bon agent IA décuple votre potentiel.
Sources: performance data published by HubSpot (2026) and customer case studies. Outcome-based pricing since April 14, 2026: you pay when the agent resolves the conversation (about $0.50 per resolved conversation), with a 28-day free trial.

Customer Agent:
myth and reality

Whether at SMBs or larger companies, there are fears and preconceived notions about AI-assisted customer service. Here are a few clarifications to clear things up.

Evenement-Franck-Aide
"AI agents dehumanize service"

The fear is that AI keeps replacing human connection with a cold machine, pushing customers even further away from a real person when they need one.

The agent frees up time for the human side

In reality, teams are often buried under administrative requests. By clearing out that noise, the AI agent frees up time to listen to customers and respond when it really counts.

1 (1)
"The agent could say anything"

A common fear is that the agent starts to hallucinate, meaning it makes up a wrong answer and confidently sends it to the customer, which can create new problems.

The agent has clear guardrails

The agent answers from your knowledge base and your rules, without trying to invent something just to please. If it can't find the answer in your documentation, it passes the ball to a human.

Who and what is this agent ideal for?

Any business that talks with its customers is a good candidate, in both B2B and B2C. Here are three situations where it comes in especially helpful.

Overwhelmed teams

Customer expectations keep rising: people want an answer right now. The agent absorbs the first wave of requests so the team can breathe.

After-hours requests

A customer writing at 9 p.m. may mostly want reassurance, and fast. The agent can reply overnight and on weekends, or hand things off the next day if needed.

Several questions before a purchase

When someone has a question about your product, the agent can answer it, help them compare, and even offer a demo to create a qualified lead.

How Guépard deploys HubSpot's Customer Agent
with you

As a HubSpot Platinum Partner agency, Guépard follows a proven approach to prepare your AI agent the same way you'd train a new employee.

1

Analysis of available content

We take stock of everything that can teach the agent: knowledge base, FAQs, product docs, and so on. If there's too much missing, we assess the "homework" and help you get it done.

2

Configuration and rules

We set up your brand voice, the planned scenarios and, above all, the handoff rules to a human. When needed, we connect your other tools through APIs, and your CRM to verify the customer's identity.

3

Testing begins

We start gently, often outside business hours. The idea is to review and validate the agent's answers before expanding. Each channel we add (chat, email, phone) has its own subtleties.

4

Team training

At Guépard, we see the agent as a new colleague, so your team needs an adjustment period to understand what it does. In short, the human-machine handoff is learned on both sides.

justine-portrait-2025

"I like to think of these agents as exoskeletons for teams: they let people do much more with the time they have, instead of adding to their workload. By removing some of the noise, we give time back for real human connection."

Justine Gaignard-Parent

President and founder of Guépard

2 to 3 weeks

From first meeting to agent live

When your content is already in place, that's all it takes to launch your Customer Agent and ease into testing. If there's a knowledge base to prepare, we'll plan the "homework" together. Let's see what's realistic for you.
Book a meeting

Technical questions for pros who want to dig deeper