Floworbiter Floworbiter
Open menu

Service · AI agents

AI agents for ticket handling and quoting, supervised by your team

We deploy AI agents that analyse incoming tickets and draft replies to customers and quotes. Your team approves them in Slack or another messaging tool, and the agent follows rules agreed with the process owner. Implementation takes 3 to 12 weeks.

What the agent does in support and sales

  • Replies to tickets

    The agent analyses every new ticket and drafts a reply to the requester: a proposed fix, or questions about the details still needed to move forward.

  • Draft quotes

    From an incoming enquiry, the agent drafts a quote based on company data such as price lists. A salesperson reviews it and sends it while the enquiry is still current.

  • Customer self-service

    The agent can handle requests the company considers safe on its own, such as ticket status, instructions or answers from the knowledge base. It works from your documentation and company data.

  • Questions to the team

    When the agent is not sure, it does not guess. It asks a team member in the messaging tool and picks the case up again once it has an answer.

  • Memory and procedures that grow over time

    The agent remembers agreements, team members' decisions and earlier cases. Corrections can become procedures, but only an authorised person who is responsible for answer quality can save them. The agent applies approved procedures to similar tickets, so over time it needs fewer corrections.

Supervision: from approving every message to agreed autonomy

The agent's autonomy grows in stages. At every stage it follows rules approved by the process owner on your side.

  1. Every message approved

    At first, every reply and every draft quote goes for approval in the messaging tool. A team member approves, edits or rejects it. Nobody has to watch the ticket queue all day, because cases that need a decision come to the team.

  2. Rules of conduct

    Based on approved and edited replies, we define rules. They set which cases the agent closes on its own, when it asks a team member and when it hands the case over entirely.

  3. More autonomy

    The agent handles cases covered by the rules on its own. Everything else still goes through approval. Every action is logged together with the data the agent relied on.

In production

Supervised AI agents are in use today at two companies, in two separate deployments.

  • Field service
  • IT support

At both companies, the agent analyses new tickets, drafts replies to requesters, including questions about missing details, and turns sales enquiries into draft quotes. The team approves replies and quotes in Slack and answers the agent's questions there.

  • hours to minutes

    time to first response to a ticket

  • days to minutes

    time to prepare a draft quote

  • over 50%

    of draft replies approved without edits

Figures from the deployment at the field service company.

What changed for the teams

  • A draft quote is ready as soon as the enquiry arrives, while the customer is still ready to talk.
  • Team members no longer have to watch the queue. Cases that need a decision reach them in the messaging tool.
  • The team spends its time on cases that need its expertise, not on rewriting similar replies.
  • Requesters get a first response and contact with the company sooner.

How AI agent implementation works

Implementation takes 3 to 12 weeks. The timeline depends on the number of processes, the systems to connect and the approval rules.

  1. Process review and baseline

    We talk to the people who handle tickets and quotes today. We pick a first process, such as a single ticket queue, and measure the starting point. That means time to first response, time to prepare a quote and the share of cases closed without handover to a person.

  2. Integrations and knowledge sources

    We connect the agent to your helpdesk, CRM, messaging tool and knowledge sources such as documentation, ticket history and price lists. The agent can only access the sources agreed at this stage.

  3. Go-live with full approval

    The agent works on real tickets, and every message to a customer goes through your team's approval.

  4. Rules and wider scope

    Based on the results, we set rules for autonomy, compare the measures with the baseline and add further processes.

  5. Maintenance

    We monitor the agent, refine the rules, update the knowledge sources and change the model when needed.

AI models, data and costs

  • Any language model

    The agent is not tied to a single provider. Current deployments use commercial models. We can also use a model running on your own infrastructure when data must not leave it. The agent's memory and procedures are stored separately from the model, so changing the model does not erase them.

  • Your company data

    The agent bases its replies on the sources you choose: knowledge base, documentation, ticket history, price lists and data from your business systems.

  • Your team's messaging tool

    Approvals and the agent's questions happen in Slack or in any other messaging tool your team already uses.

  • Model costs

    Model usage can be paid by Floworbiter as part of our work, or by your company on its own account with the provider. The choice depends on your preference and your data requirements.

How we work together and pricing

We start with a fixed-price discovery sprint. Then we implement the first process, and maintenance and further development of the agent run as a monthly subscription.

  • Discovery sprint

    Process review, choice of the first queue or case type, a baseline for measurement and a fixed-price implementation plan.

    from EUR 3,000 (net), fixed price

  • First process implementation

    Agent configuration, integrations with your helpdesk, CRM and messaging tool, knowledge sources, go-live with every message approved.

    from EUR 12,000 (net)

  • Agent maintenance and development

    Monitoring, rule refinements, knowledge source updates, further processes and a model change when needed.

    from EUR 2,000 per month (net)

The amounts are indicative, net of VAT and valid as of November 2026. Every quote is prepared individually. It reflects the expected effort, the number of processes and integrations, the scope of data to be consolidated and the number of people to be trained. AI model usage is either included in the subscription or paid by your company on its own account with the provider.

All engagement models

Questions about AI agent implementation

Do we need our own AI team to use AI agents?

No. We configure the agents, connect them to your helpdesk, CRM and messaging tool, and maintain them afterwards. On your side, a process owner decides what the agent may do on its own and what needs approval.

How do you control what the agent sends to customers?

At first, every reply and quote is approved by a team member in the messaging tool. The agent's autonomy is extended only through rules your company approves. When the agent is not sure, it asks a team member, and every action is logged with the data it relied on.

Which AI model does the agent use?

The one that best fits the task, the budget and your data requirements. Current deployments use commercial models, but the agent can also work with a model running on your own infrastructure. The model can be changed without rebuilding the process.

How long does AI agent implementation take?

Between 3 and 12 weeks, depending on the number of processes, the systems to integrate and the approval rules. We start with one process, measure the results and only then extend the scope.

Who pays for AI model usage?

It depends on your preference. Model usage can be paid by Floworbiter as part of our work, or directly by your company on its own account with the model provider.

How do you measure results?

Before launch we set a baseline: time to first response, time to prepare a quote and the share of cases closed without handover to a person. We compare the same measures after go-live.

We also build systems for service companies: work orders, inspections, quotes and reports for technicians in the field. Case study: a system for field service

Which process should come first?

Tell us how ticket handling or quoting works in your company today. We will suggest where to start and how long implementation could take.

Talk to us about your process