Support Deflection Analyst
Finds the twenty questions behind a thousand tickets
- 8hard rules
- Things it must never do, enforced in every run
- 5handover rules
- Conditions where it stops and gives the work to a person
- 5output fields
- A structured result, not a paragraph you have to re-read
- ≤7steps per run
- Bounded work: at most 5 tool calls, enforced by the runtime
Read-only by design. This agent has no path to sending, spending, publishing or committing anything. Where that limit is the product, removing it would remove the reason to trust the output.
Overview
Support teams scale linearly with customers because nobody does the analysis that would break the link. The volume is not a thousand different problems — it is twenty problems arriving a thousand times, and each has a fix that is not "answer faster".
Helpdesk categories do not reveal them. Categories are chosen by whoever is closing the ticket, under time pressure, from a dropdown designed two years ago. "Billing" contains four unrelated problems. "Other" contains the interesting ones.
And the fix is rarely a help article. That assumption is baked into every deflection tool on the market and it is wrong most of the time. A recurring ticket usually means the product does something surprising, an error message does not say what to do, or a step in onboarding was skipped. Writing an article about a confusing button treats the symptom, and volume never actually falls.
This agent groups tickets by the problem underneath, counts what each cluster costs in tickets and handling time, and says which kind of fix would end it — naming the product change, the wording, the onboarding step or the document. It answers no tickets and writes no articles: it tells you where to spend an engineer's week so that a thousand tickets become fifty.
Benefits
Volume that actually falls
Because the cause was fixed rather than documented.
A ranked case for engineering time
With the ticket cost attached to each ask.
Support stops scaling with customers
Which is the only way the economics ever change.
What it does
Groups by problem, not by category
The dropdown was designed two years ago and is chosen under time pressure. 'Other' is where the interesting tickets are.
Names the kind of fix
Product change, error wording, onboarding gap or document. Assuming it is always a help article is why volume never falls.
Costs each cluster
Tickets, handling time, and repeat contacts. That is the number that funds the fix.
Finds the tickets that come back
A resolved ticket that reopens is a worse signal than one that takes a long time.
Separates confusion from breakage
Both produce tickets. Only one is a bug, and they are fixed by different teams.
Answers nothing
It never replies to a customer or publishes an article. It tells you where an engineer's week pays back.
What this changes
What this changes about your week
- It takes over
- Finding the recurring causes behind ticket volume.
- Done today by
- Nobody, because the ticket categories make it look already answered.
- How often
- Never, which is why support scales with customers.
- Typical manual effort
- Regrouping thousands of tickets by the problem underneath rather than the dropdown someone picked.
Why Customer support in particular. Every deflection tool assumes the fix is a help article. Usually it is a product change, an error message or a skipped onboarding step — so volume never actually falls.
Runs unattended
Started by you or by an event, and it finishes on its own. Nothing waits for someone to be at a desk.
The same standard every time
The two-hundredth item is held to the bar the first one was. Consistency is the part people cannot sustain.
It cannot act on its own
Support Deflection Analyst has no path to sending, spending or committing. That limit is why its output is safe to act on.
How it runs
This agent runs server-side through the PROMIVO runtime. Each run is logged step by step and every tool call is permission-checked before it executes.
- 1InputValidated against the agent input schema
- 2Modelclaude-opus-5 with automatic fallback
- 3Toolssearch_knowledge, extract_structured_data
- 4PermissionsRuns READ actions automatically; anything higher waits for approval
- 5OutputStructured result matching the output schema
- 6LoggingTokens, duration and every step recorded
Demo
Demo dataIllustrative sample output, abridged.
Input
{
"period": "Q3 2026",
"tickets": []
}Output
{
"clusters": [
{
"trend": "growing",
"theFix": "The export screen shows the same state for 'running' and 'failed'. A distinct failed state with the reason would end this cluster. It is not a documentation problem — an article cannot tell someone what their own export did.",
"fixType": "error-wording",
"problem": "Customers cannot tell whether an export finished or failed.",
"tickets": 214,
"examples": [
"\"Has my export worked? It has said the same thing for an hour.\""
],
"reopenRate": 18,
"handlingHours": 61,
"documentedAlready": true,
"shareOfHandlingTime": 22.4,
"cutsAcrossCategories": [
"Billing",
"Technical",
"Other"
]
},
{
"trend": "growing",
"theFix": "Apparent breakage rather than confusion — a third of these reopen, which means the workaround does not hold. Reproduce before writing anything.",
"fixType": "product-change",
"problem": "Second and third users at an account never receive their invitation.",
"tickets": 88,
"examples": [
"\"I have invited my colleague three times and nothing arrives.\""
],
"reopenRate": 31,
"handlingHours": 25,
"documentedAlready": false,
"shareOfHandlingTime": 9.1,
"cutsAcrossCategories": [
"Account",
"Other"
]
}
],
"escalate": true,
"disclaimer": "An analysis of your own ticket history. No ticket has been answered or changed, nothing has been published, no customer has been contacted, and no identifying customer detail appears in this output.",
"genuineTail": "41% of tickets are genuinely one-off and do not cluster. Any tool that tells you otherwise is manufacturing groups.",
"escalationReason": "The largest cluster is 22.4% of handling time and growing, and one cluster shows apparent breakage affecting multiple accounts with a 31% reopen rate."
}Requirements
Integrations
No integrations required.
Before you start
- Ticket history: subject, body, resolution, time spent, reopens
- Existing help documentation, where you want coverage checked
- Requires the All-Access plan or higher
Use cases
Quarterly support review
The clusters, their cost, and what would end each.
Before hiring more agents
Whether the volume is real or self-inflicted.
After a release
What the new version made people ask about.
Prioritising engineering
Product problems with a ticket cost attached.
Pricing
All-Access
Billed monthly through your PROMIVO subscription. Cancel at any time.
Runs consume your plan allowance for agent executions and tokens. See plan limits
FAQ
Will it reply to tickets or write articles?
No. It is an analyst, not an agent that talks to customers. Publishing an answer is a decision about your voice and your accuracy, and it belongs to your team.
Why not just use our ticket categories?
Because they are wrong. They are chosen under time pressure from a stale dropdown, and the clusters that matter cut across them. Regrouping is the work.
Does it read customer personal data?
It reads ticket content, which may contain personal detail. It reports clusters and never reproduces a customer's identifying information in its output.
Reviews
Reviews open once customers have run this agent.
Need this agent to work differently?
Tell us what to change and our team will scope a customised version for your business.
Customize this agent