Boost your customer service with HubSpot's Customer Agent – what to consider before implementation?

HubSpot is now seeing a steady stream of new AI features being released. In this blog series, we'll show you, through concrete examples, what you can do with the different features in practice, which organisations they are suitable for, and what you need to consider to get the most business value out of them. In this blog, our expert Antti Nevalainen explains what Customer Agent is, what you should use it for and what to consider before implementing it.

What is Customer Agent?

I've done a few Customer Agent deploymentsnow and thought I'd share what I've learned in the process. This is not a sales pitch, but a practical, hands-on look at how the tool works in real customer environments and what you should be prepared for to get the most business sense out of it.

Customer Agent is HubSpot's new AI-based customer service tool. It's not a traditional chatbot that provides customers with predefined response options, but instead uses generative AI to read your company's documents, websites and knowledge base articles to answer questions in the most customer-centric way possible.

In a nutshell: the customer asks a question, the Customer Agent looks up the answer in the Knowledge Base or other sources you specify and formulates the answer for the customer. If it cannot find a sufficiently reliable answer, or if the situation requires human judgment, it will redirect the conversation to the Customer Agent.

Nice and simple on paper. In practice, however, functionality and customer experience will depend on how well you do the groundwork.

Closing tickets up to 40% faster? Yes, if the implementation is properly invested in.

According to HubSpot, their customers use Customer Agent to handle more than 50% of support requests and close tickets almost 40% faster. Sounds great, butremember that these numbers come from organisations that have invested in proper implementation - processes, content and continuous optimisation.

I've also seen deployments where the agent has mostly frustrated customers. In practice, the difference comes down to how much time and care you take to prepare. I'll talk more about deploying and configuring Customer Agent in part two of this blog series.

Technical requirements for deploying a Customer Agent

Before you can deploy Customer Agent, you need:

On the license side:

• HubSpot Professional or Enterprise license (Customer Agent works across all Hube: Marketing, Sales, Service, Content, Operations).
• Sufficient HubSpot Credits balance (we'll come back to this later).

In the technical setup:

• At least one connected channel: Facebook Messenger, WhatsApp, live chat or Conversations Inbox.
• If you are using Customer Agent on pages that are not in HubSpot's CMS, you will need HubSpot's tracking code installed on your site.

On the content side:

• Knowledge Base articles, FAQs or other documentation where the agent can read up-to-date and reliable information.
• At a minimum, an initial plan for the types of questions you want the agent to answer and the processes you want it to support.

So nothing overly complex is needed, but it's worth putting these in place before you start building and training the agent. This will ensure that you get business value and benefits from the start.

How does a Customer Agent work?

A Customer Agent uses three main sources to find answers to questions:

1. Knowledge Base articles: if you have a HubSpot Knowledge Base in place, the Agent Agent will automatically tap into it and retrieve answers directly from your up-to-date support content.
2. Web pages: you can give the agent permission to crawl your company's web pages (max 1000 pages). This is useful if service and product information is scattered across multiple pages.
3. Documents: PDF files, Word documents and other files that you upload as data sources for the agent (e.g. user manuals, service descriptions, process documentation).
   

When a customer asks for something, the agent:

• Understands the context of the question (not just picks up individual keywords).
• Checks the customer's details in the CRM, if they can be found.
• Seeks the answer from the specified sources.
• Formulates the answer in their own words, but in your brand tone of voice, based on the information found.
• Redirects the conversation to a customer service representative if unable to find a sufficiently reliable answer or if the situation requires human judgment.

Can Customer Agent be used in Finnish?

Yes. I have tested Customer Agent in English and it works surprisingly well.

If you are planning to use Agent in English, I recommend that you pay particular attention to the following points:

• Write your source materials in clear Finnish (not too formal and not too mundane).
• Agent also understands spoken Finnish, but responds in the style of the source material. Dialects can also cause problems, so prefer general language.
• If your clients use a lot of specialised industry terms, define them clearly in the source documents (glossaries, term lists, examples) so that the agent uses the terms consistently and correctly.

Credit model and costs

The Customer Agent consumes HubSpot Credits. This is important to understand because it directly affects how much it costs to use the Customer Agent and at what volume the solution makes economic sense. The cost of a single conversation is approximately 100 credits. According to HubSpot, 100 credits are roughly equivalent to 1 USD, or about 0.95 EUR per conversation.

A practical example:

• Your company receives 500 support requests per month
• A Customer Agent handles 50% of these (250 conversations).
• Monthly cost = 250 x €0.95 = €237.50

Compare this to the typical hourly cost of a Customer Agent, including incidentals, of €30-50. If the Customer Agent saves even 10 hours per month, the solution will already pay for itself. And often much more than that, when you factor in reduced response times and improved scalability.

Important considerations:

• One "conversation" = one continuous chat session, regardless of the number of messages a customer sends.
• If a customer leaves the page and comes back after, say, an hour, it counts as a new conversation.
• Preview mode does not consume credits, so you can test and pilot the tool at your leisure before rolling it out more widely.

Is Customer Agent right for your organisation?

ROI calculator

Below is a simple way to calculate whether Customer Agent is right for your organisation and your current situation.

Current costs:

• Support requests per month: ___
• Average processing time: ___ minutes
• Hourly wage for a customer service agent (incl. incidental expenses): ___ €
• Monthly cost = (support requests x processing time / 60) x hourly wage

Customer Agent costs:

• 50% of support requests are handled by an agent
• Agent cost = (support requests x 0.5) x 0.95 €
• Remaining human handled = current cost x 0.5

Savings = Current costs - (Agent cost + Remaining human resources)

Practical example:

• 500 support requests/month
• 15 min processing time
• 40 €/hour wage cost
• Current costs = (500 x 15 / 60) x 40 = 5000 €/month

With a Customer Agent:

• Agent handles 250 requests = 250 x 0.95€ = 237,50 €
• Human handles 250 requests = 2500 €
• Total = 2737,50 €

Savings = 2262,50 €/month or about 27 000 € per year

The additional benefits are not just in euros. Customers get answers faster, customer service staff can focus on more complex and value-added cases, and you can provide support 24/7 without night shifts or significant additional resources.

When is a Customer Agent the right solution and when is it not?

In my experience, however, Customer Agent is not right for every organisation or every situation. You should not implement an agent if:

• You do not have any documented and up-to-date information about your products or services.
• Your customer service relies heavily on personal, consultative expertise.
• Your customers' questions are almost always unique, complex and require in-depth background knowledge.
• You cannot commit to systematically creating, updating and continuously improving your content.

Conversely, a Customer Agent adds significant value if:

• You get a lot of repetitive, structurally similar questions.
• You already have documented knowledge (or are willing to build it up in a controlled way, for example in a Knowledge Base).
• You want to provide 24/7 support without night shifts and increase service availability in a cost-effective way.
• Your customer service team is a bottleneck for growth and you want to free up capacity for more complex cases.

If you have questions about Customer Agent or want to spar about how the solution would work for your organization and your current HubSpot environment, please contact me or any of our other experts.

In the next part of this blog series, I'll take you step-by-step through how to successfully design, build and deploy Customer Agent so that you get value and business benefits from day one. Also, in the first part of this blog series, learn about HubSpot's AI Agents and what you can do with them.