A help center that answers,
instead of just listing articles
A self-service portal that only links to articles still leaves the customer to turn an article into an actual account change. This agent reads the question and finds the real answer. When the action is something it is allowed to do, like updating an address or pulling an invoice, it just does it.
Where help centers fall short
Most self-service portals are really searchable documentation with an account dashboard bolted on. A customer trying to update a billing address still has to find the right article, then separately find the right settings page, then do it themselves. The portal has everything needed to resolve the question. Resolving it is still work the customer has to assemble on their own.
From answering to acting
The agent sits inside your existing portal and answers questions straight from your documentation, citing the source. But some questions are really requests to do something: update an address, download an invoice, check a plan’s details, change a notification setting. For those, it performs the action directly, from a pre-approved list, instead of just pointing to where the setting lives. That turns the portal from a reference library into something that actually finishes the task.
For anything outside that list, it hands off cleanly to live support with the context of what the customer was trying to do. The conversation does not restart from zero. It also tracks what customers actually try to do in the portal. That is often different from what the portal’s designers assumed. It gives your team real usage data to improve both the portal and the agent’s approved list over time.
That usage data tells apart a customer who found the answer and left satisfied from one who gave up and opened a ticket anyway. That is a far more useful signal than raw portal traffic for deciding what to automate next. For businesses with tiered account types, the approved list can differ by tier. A basic account’s address change and an enterprise account’s address change might look identical on screen. They can still carry different approval rules behind the scenes.
Where your team stays in charge
Any account action outside the pre-approved list goes to live support, as does any request involving a dispute or exception. Your team decides which actions are safe enough for the agent to perform directly, and reviews usage data to expand or restrict that list. Your team also sees which actions customers attempt most often but are not yet on the approved list. That becomes a practical roadmap for where to expand self-service next, rather than a guess.
How it stays inside the lines
The agent can only perform actions on a pre-approved list your team defines. Nothing outside that list gets attempted, however the request is phrased. Every action taken is logged with a before-and-after record of the account state. The approved action list is tested in a sandbox against real account scenarios before going live, and a person can reverse any action. The embedded agent respects your portal’s existing permissions exactly. A customer only ever sees and acts on their own account data, through the same login the portal already enforces. No new access path gets added for your security team to review separately. The agent introduces itself as an AI assistant at the start of every conversation.
Price and timeline
| Option | Price | Timeline |
|---|---|---|
| Agency runs it | from $3,400 + support plan | 4 to 6 weeks |
| Full control, handover-ready | from $4,400 | 6 to 9 weeks |
“Agency runs it” keeps the agent on our infrastructure with a monthly support plan covering tuning and monitoring. “Full control, handover-ready” puts the agent on your own servers and accounts with documentation, source and credentials, so your team can run and change it without us. It costs more up front because the handover package, your-infra deployment and internal documentation are built in from day one.
Related
See the AI agents overview for how we build and guard these systems, and Development for the services that often pair with this one. Inside the agents catalogue: FAQ and knowledge base agent, Subscription management agent, Technical support tier-1 agent. For a narrower, single-process version of this work, see Internal knowledge base Q&A, RAG knowledge base with citations in the automations catalogue. For how the same guard patterns held up on real systems: ProBay marketplace AI agent team.
Want this running for your team? Get in touch and tell us where the messages pile up.
FAQ
How much does a self-service portal agent cost?
From $3,400. That covers embedding the agent in your existing portal and wiring it to the account actions you want it to handle directly. Live in 4 to 6 weeks.
How long before it is live?
4 to 6 weeks. Most of that time goes into deciding exactly which account actions are safe to automate and which should stay a request to support.
Which channels and tools does it work with?
Your existing customer portal or account dashboard, embedded as a widget. It connects to your CRM, billing platform or account database for whatever actions it is allowed to take.
What if the agent takes an action it should not have?
Every direct action is limited to a pre-approved list your team sets: address changes, invoice downloads, plan detail views. Nothing beyond that list is ever attempted, and every action taken is logged with a before-and-after record.
Is account data safe inside the portal?
The agent works inside your existing portal's login, so it only ever touches the account the logged-in customer already has access to. No new access layer gets added, and every action is logged.