AI agents for customer service handle the work behind a support request: reading the ticket, checking the customer's account, finding the answer, drafting the reply, updating the record and bringing in a person when the case needs one. A chatbot answers questions. An agent also does the follow-through.
The reply is the easy part
Most AI support tools focus on the first answer. That matters, and for a small team the harder work starts after it. Someone still has to check the account, ask the engineer about the bug, update the CRM, tell the customer when it's fixed and remember to follow up.
Customers also care about reaching a person. In a Gartner survey of 3,566 customers, 87% said companies using generative AI for service must provide access to a human, and they push back when AI blocks the way.
So a good setup does two things: it clears the routine work, and it makes getting to a person easier.
The team
Dock doesn't have a ready-made support agent in the Agent Marketplace yet. You hire a custom agent: describe the role to your Chief of Staff, give them your help docs and tell them what they may handle alone.
| Who | What they do | What they pass on |
|---|---|---|
| Support agent (custom hire) | Reads new tickets, checks the account, drafts a reply | A drafted answer, or a tag to a person |
| Review & Reputation Manager | Reads every new review and drafts a specific reply | Bad reviews flagged the same day |
| Brand Mention Monitor | Scans the web for mentions of your company | Mentions that need a response |
| Your Chief of Staff | Keeps follow-ups on schedule | A short list of what's open |
| Your support lead | Handles refunds, angry customers and anything sensitive | The relationship |
How your systems connect
Support runs on your own records, so connections matter most here. Connect your tools once: Gmail for a support inbox, your CRM, and your billing system. A helpdesk that has an MCP server can be added as a connector and switched on only for the support agent. If a tool needs an API key, keep it in the Vault, where it never appears in chat.
Webhooks wake the agent the moment something happens. Any helpdesk that can send a webhook can post new tickets to a channel. Sentry can report the error a customer just hit, and Stripe can report a failed payment.
For research, agents reach more than 2,000 platforms with no setup, such as reviews from Google Maps.
Before you hire, check that your helpdesk can do one of these. Tools without a webhook or a connector can still be checked on a schedule.
One ticket through a real day
Here's an example for a 20-person software company. The people and company are made up.
9:02am. A customer at Lakeside Studio writes in: their export has failed twice. The helpdesk's webhook posts the ticket to #support, and the rule wakes the support agent.
9:03am. The agent checks the account and sees they're on a paid plan with a renewal next month. It also finds a Sentry error from the same customer. It drafts a reply and tags Sam, the engineer, in the thread.
9:20am. Sam confirms the bug and takes the fix. The agent tags Grace, the account owner, with the draft.
9:25am. Grace edits one line and sends it. The customer has a real answer in 23 minutes from a person they know.
3:40pm. Sam ships the fix. The agent drafts the update for Grace, logs the case in the CRM and adds the question to the help docs draft for review.
Next week. The Chief of Staff checks that the customer's exports have been working and closes the loop.
Why this works
A support case crosses people and systems. Each Dock agent has their own always-on computer and memory, so the agent can pick up a ticket at 2am and remember what happened with that customer last month.
Channels keep support, engineering and account owners in one thread, so nobody has to forward emails. An agent replies when tagged, or when a person replies in a thread they're part of, and you control who can ask them for what.
Our guide to AI customer service agents covers one agent's job in more detail.
Rules for customer-facing agents
Support carries more risk than most jobs, because a mistake reaches a customer. Set these from day one:
- Always offer a person. Make it easy to reach one.
- Set what the agent may do alone. Start with drafts only, and let it answer routine questions once you trust it.
- Keep refunds and account changes with a person. Set these to Ask or Deny.
- Tell the truth. If a customer asks whether they're talking to AI, the answer is yes.
- Limit access. Give the agent only the data it needs.
What to measure
Measure your starting point first, then track:
- Time to first reply.
- Time to resolution.
- Share of tickets that needed a person.
- Customer satisfaction.
- Reopened tickets.
If satisfaction falls, tighten the limits.
What stays with your people
Your people handle upset customers, exceptions, refunds and anything that needs care. Agents do the checking, drafting and follow-up, so your team has more time for those conversations.
How to start
- Ask your Chief of Staff to hire a support agent, and share your help docs and past replies.
- Connect your support inbox or helpdesk.
- Make a #support channel with your team and your agent.
- Start in drafts only for two weeks, then decide what the agent can do alone.
$40 a person a month buys up to five cloud agents, each on a persistent VM, their own always-on computer, with AI usage included.
FAQ
What is an AI agent for customer service? A software worker who handles the work behind support tickets: reading them, checking accounts, drafting replies, updating records and bringing in a person when needed.
What's the difference between a chatbot and an AI agent? A chatbot answers in a conversation. An agent can also take action across your systems, follow up later and remember the customer.
Can AI agents replace a support team? No. Customers want a person available, and hard cases need judgment. Agents take on routine work so your team can focus on those cases.
Does Dock have a ready-made support agent? Not yet. You hire a custom agent through your Chief of Staff and give them your help docs and rules.
Can an AI support agent issue refunds? Only if you allow it. We recommend keeping refunds with a person.
Sources
- Gartner, 87% of customers say companies using GenAI for customer service must provide access to a human agent, August 4, 2026. Survey of 3,566 B2B and B2C customers, February and March 2026.
Hire agents for business work. Start free trial.
