Flexivity’s AI handles tickets through three composed mechanisms: it reads each ticket and classifies its attributes, consults the rules and procedures that govern what should happen, and takes action — from leaving the ticket alone to resolving the conversation and making the change in your system. Progressive Autonomy sits over all of it, letting you set how far the AI goes per class of tickets, and change your mind as you learn.
Progressive Autonomy
A password reset might get answered directly. A frustrated billing question might stay at Assist for an agent to send. A refund dispute might go straight to a person. Each rule is a setting an admin changes in minutes, so your configuration keeps up with what you’re learning about where the AI holds up and where it doesn’t.
- Multi-level — seven Answer levels and four Action levels, from Assist to full resolution
- Set per class of tickets — rules match on ticket attributes to choose a conversation level; data connector change tools choose an action level
- Promote or demote in minutes — not a project, not a vendor engagement, not a change window
Classification
Attributes come from three places: ones you define that reflect how your business thinks about tickets (customer tier, product line, billing dispute type); ones mapped to your help desk’s own fields (topic, category, priority); and ones built into Flexivity (sentiment, agent reply state, message count). Agents can correct any attribute inline, and the correction wins — the rules use it from then on.
- Flexivity attributes — define custom labels that reflect how your business thinks about tickets
- Mapped attributes — the AI reads or fills in your help desk’s existing fields (topic, category, priority)
- System attributes — sentiment, agent reply state, message count, built into Flexivity
Procedures
You write a procedure in plain language. Paste the SOP your team already follows and Flexivity drafts it for the AI, or start from a template. Before a procedure goes live, the Reading tab shows what Flexivity understood — the steps in order, the tool each step uses, the questions it may ask, its conditions, and its handovers — so you can fix anything that doesn’t match your intent before it runs. Then key steps are enforced in code: a change is prepared only when the procedure reaches it, and conditions are checked against the actual record before the change can run. Finally, automated tests exercise full conversations to ensure the desired outcome, consistently.
- Three ways to start — from a template, by pasting your SOP, or from a blank page
- See what Flexivity understood — the Reading tab shows the AI’s interpretation before anything goes live, with key steps enforced in code (not just AI prompts)
- Test full conversations — automated tests ensure the desired outcome is met consistently, including changes made and not made as expected
The three mechanisms, composed
Each piece is useful on its own, but AI Handling is the loop they form together: Classification reads the ticket. Rules consult the attributes and choose a Progressive Autonomy level. If a Procedure applies, it orchestrates what the AI does next — look things up, ask the customer a question, prepare a change, or hand over. The level governs how far the AI goes without an agent’s involvement; the procedure governs what the AI actually does within that scope.
The result: a ticket about an order that hasn’t shipped gets classified as an Order Status question with low urgency, matches a rule promoting it to Resolve, triggers the Order Status procedure, looks up the order in your system, and tells the customer where it is — end to end, grounded in your actual data, within the boundary you set.