All solutions
Functional team · Support and helpdesk

Answers from the company's own pages, with a person for the rest.

The same questions reach the same few people, and the policy pages go unread.

For IT service teams and the staff of every department they answer.

How the example fits together: tools and an agent built on Unflood, with a person approving before anything is written.

Made from a conversation in Claude, released to the AI platforms your team already uses, running in your own environment.

Claude
Turn this conversation into a helpdesk agent that answers staff from our IT policy pages.
Unflood created the agent from this conversation, ran its multi-turn tests and published it to the IT repository once they passed.
  • Unflood: Helpdesk agent created from this conversation 3m ago
  • Unflood: Multi-turn tests run, including a hand-over 2m ago
  • Unflood: Published to the IT repository once its tests passed 1m ago
  • Unflood: Shared with the IT team: staff use it, only the IT lead changes it Just now

See it happen in a chat

A person on the team asks for an agent. Unflood does the building in the background and comes back for the decisions that are yours.

  • Unflood tests them.
  • A person approves each release.
  • They are shared with the right people.
  • They are kept up to date, so you step in only when a decision is yours.

Examples for Support and helpdesk, to learn from

Example 1

Answer questions from the policy pages, citing the page

On Unflood: An agent bound to the policy pages; answers cite their source.

Example 2

Look up the person's open tickets

On Unflood: A read-only ticket-system tool.

Example 3

Hand a person what it cannot answer

How a helpdesk answer fits together

Knowledge
IT policy pagesThe agent is bound to the company's own pages.
Tool
Ticket systemRead only, no approval needed.
Agent
IT helpdeskA question in, an answer with its sources out.
Person
A person takes the restWhat the agent cannot answer goes to a person.

Six moments that make the difference

Each one is shown on sample data, with this team's own example where there is one.

Build it from a conversation

Describe the agent in a Claude chat. Unflood records it, its tools and its tests.

In Support and helpdesk: Answer questions from the policy pages, citing the page

  • KnowledgeThe policy pages, bound by reference.
  • ToolsThe ticket system, read only.
  • Builds itThe support team lead.
Recorded from the conversation IT helpdesk
Policy pagesTicket systemSupport lead
Drawn on sample data.

Test it

Suites and a judge run it; each failed case becomes an item to fix.

In Support and helpdesk: Multi-turn conversations, including a question it must hand over.

  • TestsMulti-turn conversations, including a question it must hand over.
Test lab
  • A case failsJudged against the version tested.
  • It becomes an item to fixNothing is lost between runs.
  • Retest, then release
Drawn on sample data.

Release it with approval

Package a version. Nothing reaches people without a person's approval; the builder cannot approve their own.

In Support and helpdesk: The service owner.

  • Approves itThe service owner.
Release waiting for approval
  • Packaged byThe support team lead.
  • Needs approval fromThe service owner.
The builder cannot approve their own release.
Drawn on sample data.

Share it

Choose who can use it, person by person or team by team, where the team already works.

In Support and helpdesk: Every member of staff.

  • Uses itEvery member of staff.
Who can use it
  • The Support and helpdesk teamEvery member of staff.
  • A named personChoose person by person or team by team.
Drawn on sample data.

Keep it up to date

Change it, retest it, release the next version. Every version can be rolled back.

  • Learn from feedbackEach failed answer becomes an item to fix.
Versions
  • Current versionLive.
  • Next versionChanged, retested, waiting for approval.
  • Roll backEvery version can be rolled back.
Drawn on sample data.

The organisation stays in control

Members, roles and a record of every change belong to the organisation.

In Support and helpdesk: Every change to the agent is one entry in the ledger, with who made it.

  • A workspace for the organisation, on its proven domains
  • Members and roles: who uses an agent and who changes it
  • Release approval by a second person
Record of every change
  • UnfloodConnects the ticket system, read only, no approval needed.
  • UnfloodDrafts the helpdesk agent: a question in, an answer with its sources out, bound to the IT policy pages.
  • UnfloodWrites a three-turn test case: a VPN failure, error 809, the policy that covers it.
Drawn on sample data.

How the team gets there faster

Your people describe the work and approve it. Unflood does the building, testing and publishing.

Today

The same questions reach the same few people, and the policy pages go unread.

With Unflood

Answers from the company's own pages, with a person for the rest.

Your people do
  • Builds itThe support team lead.
  • Approves itThe service owner.
  • Uses itEvery member of staff.
Unflood does
  • Connects your tools and data
  • Writes and runs the tests
  • Packages each version
  • Publishes it where your team works
  • Retests every change before the next release
  • Keeps the record of who changed what

In the reference journey's recorded replay: 6 steps, 6 by Unflood and 0 by your people. Counted from the journey fixture, on example data.

Talk to us about your team

Tell us the work your team wants an agent for. We set up a workspace with you and build the first one with the people who know the work.

Get in touch