Skip to main content

​
Automation opportunities

The Automation opportunities surface (dashboard → Agents → Automation opportunities) answers the question the conversation-intelligence scores leave open: once every inbound conversation is classified and priced, which contact reason should you hand to an agent next — and what would that save per month? This page is the concept model behind that recommendation feed: what qualifies a conversation family as agentable, what a recommendation carries, and why nothing ever launches itself.

​
The premise

The conversation-intelligence pass classifies each inbound conversation into a contact reason (“refund”, “account access”, “delivery status”) and the insights rollup family turns those classes into economics: volume, handle time, human-handling cost. That is the conversation P&L. Automation opportunities is the read-time layer that takes that P&L and flags the families where humans still do work an agent could absorb — like the rest of the insights family, it is a read-time rollup over conversation records you already have, not a new pipeline.

​
Interaction-level signals vs family-level verdict

Two levels of decision are at play.
  • Interaction level — each conversation contributes per-interaction signals: its channel, its classified topic, whether it was resolved, how long resolution took, and — critically — whether it resolved without a human assignee (that counts as already deflected) or with one (still human-handled). A conversation still open counts as unknown.
  • Family level — those interactions aggregate into a topic family verdict: the family’s current-window volume, its measured self-service deflection rate, its average handle time, and its trend versus the prior equal window.
A family flips into an opportunity only when two thresholds hold at once: its current-window volume clears the min_volume floor (so long-tail noise never surfaces), and its deflection headroom — target rate minus today’s measured rate — is positive. A family already deflecting at or above your target is deliberately excluded: the surface answers “what to automate next”, not “what is already automated”. An unknown deflection rate counts as zero headroom consumed, because “we never measured deflection here” is itself a strong automate-me signal.

​
The recommendation payload

Each row in the ranked list is one topic family, and carries:
  • the family label and its volume, trend, and busiest channel;
  • its measured deflection rate and average handle time;
  • the automatable volume — the incremental contacts deflection headroom implies;
  • a 30-day projection — monthly deflections and monthly savings, priced from your human-handling cost minus the automated per-contact cost;
  • a relative opportunity score (0–100) that ranks the set; the biggest win scores 100, and ranking falls back to automatable volume when no ROI prices out yet;
  • a scaffold suggestion — the Orbit primitive best suited to deflect the family: a self-serve flow for structured transactional reasons (refunds, cancellations, account access), a knowledge-base article for purely informational ones (delivery status, product support), and an AI agent for conversational, judgement-heavy ones.
Money is priced from your own assumptions — labour cost per agent-hour, automated per-contact cost, target deflection rate, fallback handle time — which you can read and update on the surface’s config endpoints. Until you price them, defaults kick in and the ranking falls back to raw volume.

​
From offer to agent

Clicking a scaffold does not launch anything — it creates a draft. For an ai_agent scaffold the click deep-links you into the draft agent’s page; for a flow scaffold it opens the flow builder with the draft loaded; a knowledge-article scaffold is presented as guidance rather than a draft. From there the normal authoring loop applies — ground the agent, review it, and activate it yourself, or assemble several scaffolds into a squad; the creating-agents walkthrough covers that loop. The hand-off contract is exactly this: the surface offers a pre-titled draft targeting the family’s busiest channel; the decision to activate stays with you.

​
Worked example

Suppose your 30-day P&L shows a cancellation family with 400 contacts, a measured deflection rate of 10%, and a 6-minute average handle time. With a target deflection rate of 60%, headroom is 50 points, so roughly 200 of those contacts are automatable this window — about 200 deflections a month. Priced at a 240-second handle time against your labour rate, minus a few cents of automation cost per deflection, the family projects a four-figure monthly saving and lands at the top of the list with a self-serve flow scaffold. A sibling family with the same volume but a measured deflection rate already above target does not appear at all — there is nothing left to win there.

​
Tenant-owned decision

The surface is strictly advisory. It never creates an active agent, never routes traffic, and never changes a write path — it reads conversation records and your own ROI assumptions, and it only ever produces drafts when you click. Tuning the assumptions, ignoring a recommendation, or deleting a scaffolded draft are all tenant-owned choices; the flagged options wait for you.
Projected savings are estimates computed from your own cost assumptions, not a billing commitment. Treat the ranking as a triage aid and validate a scaffolded draft on real traffic before activating it.