The DigiKat Learning Center

The Support Agent: Help Desk to Portal Chat | DigiKat

Written by Mick Goman | Aug 17, 2026, 8:30:00 PM

Your customers don't care which door they knocked on. They raised a ticket on Monday, asked the website chat on Tuesday and typed into the portal on Wednesday, and to them it's one question that still hasn't been answered. Most businesses run those three channels as three separate problems. We run them with one Support agent that reads the same record every time.

The Support agent is the third agent in our series on the management agent team we build on HubSpot. The Sales agent works the sales pipelines and the Service agent keeps delivery moving. The Support agent is the one your customers actually talk to, so it's the one people judge you by.

One agent, three front doors

The Support agent covers three channels:

  • The help desk. Tickets that arrive by email, form or the HubSpot help desk inbox.
  • Website chat. Visitors asking questions before they buy, or customers who can't find the login.
  • Chat inside portals. Patient portals, membership portals and client portals, where the person is already logged in and HubSpot knows exactly who they are.

Why one agent and not three? Because the answer to "where's my order?" or "how do I change my booking?" shouldn't depend on where the question was asked. When you set up three bots, you get three versions of the truth, three sets of rules and three places to fix when something changes. One agent means one set of approved answers, one set of hand-off rules and one scorecard.

It also means the agent sees the whole conversation. If someone opened a ticket yesterday and then jumps on chat today, the agent can see that ticket on the contact record and pick up where it left off, rather than asking them to start again.

What it reads before it answers

This is where HubSpot earns its place. A chatbot bolted onto a website only knows what's on the web page. A Support agent connected to HubSpot can look up the person and everything associated with them:

  • The contact and their company, so it knows who it's talking to.
  • Open and recent tickets, so it doesn't answer a question that's already in progress.
  • Their deals or orders, so it can tell a prospect from a paying customer.
  • Their membership, booking or service record, often a custom object or one of HubSpot's Services or Appointments objects.
  • The approved knowledge base, which is the only place it takes answers from.

Because those records are linked by associations, the agent follows a clear path: find the contact, fetch the associated records, check the knowledge base, answer. That structured flow is the whole reason we build on HubSpot. An agent given a messy inbox and a stack of PDFs has to guess. An agent given a clean record with the right associations doesn't. If you want the plain-English version of how those links work, our guide to pipelines, objects and associations covers it.

It's also the difference between a chatbot and an agent. A chatbot matches a question to a scripted reply. An agent looks things up, decides what applies to this person and takes a next step. We wrote more on that in chatbot or AI agent.

Answer it, or hand it off with the whole story

A good Support agent is judged as much by what it hands off as by what it answers. Every agent we build has written hand-off rules. They usually look something like this:

  1. Answer when the question matches an approved knowledge base article and nothing on the record contradicts it.
  2. Ask one clarifying question when the request is ambiguous, then try again.
  3. Hand off to a person when the question isn't covered, the customer is upset, money is involved, or the topic is on the "always a person" list.

The hand-off is where most chat tools let customers down. The person gets "someone will be in touch" and the team member gets a transcript with no context. Our Support agent creates or updates the ticket with a short note that covers:

  • What the customer asked, in one or two lines.
  • What the agent already checked and told them.
  • The associated records it looked at (order, membership, booking).
  • Why it's handing off, and a suggested next step.

The ticket goes to the right pipeline and owner, so the person picking it up starts with the full picture. The customer doesn't have to repeat themselves, and the team doesn't have to dig.

Portal chat in health and membership businesses

Portal chat is where the rules matter most. In a patient portal, the Support agent handles admin: booking changes, where a form is up to, how to upload a document, when a script was sent. It never gives clinical advice. Any question about symptoms, medication or results goes straight to the clinical team's queue with the full record attached. In a membership portal, it can explain freeze rules and booking policies, but a freeze that changes billing still goes to a person to approve. We covered the membership side in the member portal where an agent handles freezes, bookings and billing questions.

What it doesn't do

The "always a person" list is short, and we write it down before the agent goes live:

  • Refunds, credits, payment plans or anything else that moves money.
  • Clinical, legal or financial advice.
  • Changes to contracts, plans or account ownership.
  • Complaints and anything that sounds like one.

Everything else starts in draft. For the first weeks, the agent drafts replies to tickets and a person approves them. Chat answers are limited to topics where the knowledge base is solid. If the approval queue gets backed up, the agent doesn't help anyone, which is why we set up the approval queue before the agent goes live.

How we switch one on

We use the same method for every agent: Context, Action, Coordination, Result.

  • Context: we check the data first. Are tickets in the right pipelines? Are contacts linked to their orders or memberships? Is there a knowledge base, and is it current? Then we pick one job, for example "answer booking change questions", agree what done means and measure what it costs you today.
  • Action: HubSpot's own Customer Agent first, because it already works across chat and the help desk. We build custom only where it stops, such as reading a custom object in a portal. We test it on real past tickets before any customer sees it.
  • Coordination: people approve, agents act. The agent earns more freedom once 95% of its replies have been approved unchanged for two weeks. Money, signatures and clinical or legal advice stay with a person, always.
  • Result: a monthly scorecard. What it answered, what it handed off, what it saved, what it cost and the net return. HubSpot's own Customer Agent performance reports give us a lot of this out of the box.

Support is often the first agent we recommend, because the questions are repetitive, the knowledge already exists in your team's heads and the hand-off to a person is easy to define. It also feeds the rest of the team. Every question it answers or can't answer is a signal about what your customers need, and the other agents use that.

If your help desk, website chat and portal chat are three separate headaches, we can show you what one Support agent would look like on your HubSpot. Build my solution to get a plan for your business, or start with an AI Readiness Audit to check whether your data is ready. You can see more about how we build agents into customer portals too.