Wherever work follows a procedure, Floxar turns it into a flow that people and AI agents execute step by step, with a trail of every run. Choose a team to see what that looks like.
IT operations
Runbooks your engineers and your agents can both run.
Runbooks are written after an incident and are out of date before the next one. AI agents could take on much of the work, but an agent given shell access and a wiki will improvise, and nobody wants that in production. On Floxar, the runbook is a flow: the agent works within it, hands over before the risky parts, and every run is on the record.
Example: an incident
Automatic
An alert starts a run
Your alerting starts the incident flow through Floxar's API. A duplicate alert carrying the same key returns the same run instead of opening a second one.
AI agent
Collect diagnostics
The agent works through the step's checklist with the tools and credentials your team gave it, and records what it finds in the step's fields.
AI agent
Match a known cause
A known issue leads to its documented fix, and anything else leads to escalation. The agent chooses between the paths the runbook defines.
Handoff
Hand over before the risky action
Before a restart or rollback, the step tells the agent to hand the run to the on-call engineer, with everything it found already on the page.
Engineer
Resolve, close and improve
The engineer fixes it and closes the run with a reason. Trails show which branches fire and where time goes, and the team improves the runbook from them.
Where it fits
Incident response
Access requests
Onboarding and offboarding
Changes and deployments
Routine maintenance
Service desk triage
People operations
Every joiner, mover and leaver handled the same way.
People processes cross HR, IT, facilities and managers, and they break at the handovers: an account left open, a laptop never returned, a policy step skipped for one hire and not the next. On Floxar, each process is one flow that everyone involved works from, with agents taking the repetitive steps and every run recorded.
Example: an employee leaves
HR
Start the offboarding run
HR starts the flow for the leaver and fills in the step's form: last day, role, manager, equipment on record.
AI agent
Remove access
An agent works through the access checklist for that role with the credentials your team gave it, and records each system as it goes.
Handoff
Return equipment
The run passes to IT for the equipment return, with the checklist and the agent's notes already on the page.
Payroll
Final pay and benefits
Payroll picks up the final step and closes the run. The trail shows who did what and when, ready for any later question.
Where it fits
Employee onboarding
Offboarding
Role changes
Leave requests
Policy acknowledgements
Employee questions
Finance and back office
Controls that are followed, and can show it.
Back-office work is full of procedures that must be followed exactly and exceptions that eat the team's week. On Floxar, the procedure is a flow with the checks built into each step. Agents clear the routine cases, people handle the judgement calls, and the trail shows how every item was handled.
Example: an invoice exception
Automatic
The exception joins a queue
A run is created through the API for each invoice that fails matching, and waits in the team's queue.
AI agent
Investigate
An agent claims the run, compares the invoice with the purchase order and receipt, and records the differences in the step's fields.
AI agent
Apply the policy
A difference within tolerance follows the documented path to approval; anything else goes to a specialist. The agent chooses only between those paths.
Handoff
Specialist review
An accounts payable specialist picks up the run with the investigation already done, decides, and closes it with a reason.
Team
Fix the cause
Closing reasons across trails show which suppliers and causes produce the most exceptions, so the team fixes the source.
Where it fits
Invoice exceptions
Vendor onboarding
Month-end close
Expense review
Compliance procedures
Audit requests
Customer support
The right answer on every case, from people and agents alike.
Support teams work under pressure from knowledge bases that are hard to search, and new hires take months to handle cases confidently. On Floxar, each type of case is a flow that guides the person or agent handling it one step at a time, and trails show leaders how cases are actually handled, not only how they end.
Example: a billing dispute
AI agent
Gather the facts
An agent starts the dispute flow, collects the account details and billing history the step asks for, and records them in the step's form.
AI agent
Apply the refund policy
A refund within policy follows the documented path; anything outside it goes to a person. The agent chooses only between those paths.
Handoff
Exception to a team lead
A team lead picks up the run with the facts already gathered, and does not have to ask the customer again.
Team lead
Resolve and close
The lead decides, closes the run with a reason, and can leave feedback on any step that was unclear, so the flow improves.
Where it fits
Billing disputes
Returns and refunds
Account changes
Troubleshooting
Escalations
New-hire ramp-up
Building AI agents
Keep the business logic out of your agent's code.
Point an agent at a whole knowledge base and its accuracy suffers, and every business-rule change becomes a code change. With Floxar, your agent connects over the Model Context Protocol and follows the procedure one scoped step at a time. The rules live in the flow, where the people who own the process can change them without a deploy.
How an agent works with Floxar
AI agent
Find the procedure
Given a task, the agent searches the account for the flow that covers it, or reports that none exists so the gap is visible.
AI agent
Run it step by step
It starts a run and receives one step at a time, with only the instructions, fields and references that step needs.
AI agent
Decide within bounds
At each decision it chooses among the paths the flow defines, and uses credentials from the Secrets Vault by name.
Handoff
Hand over when the flow says so
Where the flow calls for a person, the agent hands the run over, and the person continues from the same record.
Process owner
Change behaviour without a deploy
The process owner edits and approves the flow, and the agent follows the new version on its next run.