Desk
AvailableThe AI-native service desk
Resolves routine tickets end to end, escalates the rest to a human with full context, and proves the deflection rate with analytics your finance team can audit.
The problem
Support teams at growing companies spend most of their day on the same forty questions. Hiring scales the cost linearly; a generic chatbot deflects the easy half and infuriates everyone else. Neither gives you a number you can take to a board.
Who it’s for
Operations and support leaders at 50–1,000 person companies running a ticket queue that grows faster than headcount.
How it works
- 01
A ticket arrives and is triaged against the policy you set per category.
- 02
The agent drafts or resolves, working only from your knowledge base and connected systems.
- 03
Confidence below your threshold, a VIP sender, or a negative-sentiment signal routes it to a human — with the reasoning attached.
- 04
Every resolution is metered to an append-only ledger, so the deflection figure survives an audit.
What it does
Auto-resolve policy you control
Per category: on, assist-only, or always-human. Set the confidence threshold and the maximum attempts before a human takes over. VIP sentiment overrides the policy automatically.
SLA engine with business hours
Live countdowns on every ticket — ok, at-risk, breached, paused, met — computed against per-priority policies and a business-hours calendar, with escalation chains and VIP routing.
Knowledge base that fills its own gaps
Clusters the questions the agent couldn't answer, drafts an article from the gap, and routes it for approval. Resolutions feed back into the library.
Billing you can defend
A ticket counts as AI-resolved only if it closed with no human reply and no reopen within 72 hours. Reopens post a compensating reversal. Rows are never mutated or deleted.
Bring your own model key
Per-tenant model allow-list and your own API key. If the key is absent, the desk degrades to templates rather than going down.
Two-way with Command
An escalated ticket can start a governed multi-agent workflow in Command, with the run linked back onto the ticket and audited on both sides.
Built, not mocked up
- Productised from a post-production support system already running in production
- Append-only meter ledger and audit log — enforced by database trigger, not convention
- Role-based access (manager / agent), session auth, every settings change audited
- Deflection, resolution mix and usage all computed from the ledger — no hardcoded dashboard figures
Where it sits in the method
Desk is what step M looks like when the specific workflow is “repeat inbound support.” It ships with the baseline measurement built in, because step M requires one.
Read the OUTCOME MethodNext step
See Desk against your own workflow.
We will run it on a real example from your business rather than a canned demo, and be straightforward about where it does not fit.