Recommendation per case
What the assistant recommends, from which article, and what the agent did.
7 rows · Case · Subscriber · Category · Article
Bought by Customer Care and Service Operations
Resolve more contacts at first touch, with the agent in control and every action evidenced.
Repeat contacts, long handling times, inconsistent resolutions, and no defensible record of why an agent did what they did.
Demonstration available — A journey runs end to end on sample data, including its refusal and failure states. Runs today on a sample of your market; its workspace opens with Customer Care and Service Resolution Intelligence.
The result, on the sample
7 cases: 4 resolvable from approved knowledge, 2 to escalate, 1 in categories no article covers; 3 repeat contacts within seven days; 150 units credited above ceilings.
Sample. Choose your market to see this page, the workbook and the review in your currency and language.
Customer-care AI Resolution AssistantYour market · units
7 cases: 4 resolvable from approved knowledge, 2 to escalate, 1 in categories no article covers; 3 repeat contacts within seven days; 150 units credited above ceilings.
Cases7cases
Resolvable from knowledge4cases
First-contact resolution42.9%
Credit recommended470units
Decision Approve the 4 recommendations within bounds; escalate 2; add an article for ROAMING_BILL; review 1 credit(s) above ceiling. If nobody acts: Agents resolve from memory; credits are given without a ceiling and closures without a recorded reason.
The other 5 measures the run holds
- 3 cases — Repeat within 7 daysprevious case ≤ 7 days ago
- 470 units — Credit recommendeddisputed amount bounded by the article's ceiling
- 670 units — Credit givenfrom the agent decisions
- 150 units — Given above ceilingcredit given minus the article's ceiling where positive
- 40 % — Recommendations acceptedaccepted over recorded decisions
See the product
The workspace, screen by screen.
The 7 screens a demo walks through, captured from the built product on sample rows. Sign in to run it on your own market and your own numbers.
1 of 7 · Overview The workbench: the tiles, the run, the ladder, the last run and the next step. Captured on the built product, signed in as an analyst, on 2026-09-30. Sample rows. 
2 of 7 · Sample The result on the sample: the headline and the tiles, each with its basis. Captured on the built product, signed in as an analyst, on 2026-09-30. Sample rows. 
3 of 7 · Sample The decision the engine proposes, and what happens if nobody acts. Captured on the built product, signed in as an analyst, on 2026-09-30. Sample rows. 
4 of 7 · Sample The checks the engine refused, each with its reason and its owner. Captured on the built product, signed in as an analyst, on 2026-09-30. Sample rows. 
5 of 7 · Your data Your data: the sheets, the template to download, and the review of what comes back. Captured on the built product, signed in as an analyst, on 2026-09-30. Sample rows. 
6 of 7 · Runs Every run, as it was computed, with the version it ran on. Captured on the built product, signed in as an analyst, on 2026-09-30. Sample rows. 
7 of 7 · Decide The proposal to an approver, and what the approver sees. Captured on the built product, signed in as an analyst, on 2026-09-30. Sample rows.
Three things it decides
Each on your rows, each with its reason written.
Closure
Cases by the state your ticketing holds.
- open
- 2
- resolved
- 3
- escalated
- 1
What it needs from you, and how a pilot proves it
First-contact resolution, and repeat contacts within seven days.
cases, knowledge and remedies as three sheets: every case names a category the knowledge sheet covers; every remedy names a case and a permitted remedy with a value bound; an agent decision is present or the case is open. A subject column holds your own pseudonymous reference; a phone number, IMSI, IMEI or account number in it is refused, and the refusal names the column.
- The data you supply
- Knowledge, Cases, Agent_Decisions — your own rows, on your own subject keys, through the same review the sample passes. Bring your data — the workbook out and back, after sign-in. Download the sample workbook
The 3 sheets, column by column
- Knowledgethe approved articles: what each category's resolution is, which remedy it permits and up to what value. Columns: article_id, category, resolution, remedy, remedy_ceiling, needs_supervisor.
- Casescontacts as they arrived: the category, the customer's context, the disputed amount, the state. Columns: case_id, subscriber_ref, category, opened_on, tenure_months, previous_case_days, disputed_amount, state.
- Agent_Decisionswhat the agent did with the recommendation, and any remedy actually given. Columns: case_id, decision, remedy_given, credit_given, decided_on.
- The measure
- Counted once reconciled against your own records: closures against the ticketing system; credits against billing.
- The approval
- A second person approves; the person who loaded the data cannot. The decision is audited with its reason.
The products and the streams it draws on
Runs on Customer Care and Service Resolution Intelligence; draws on Decisioning and Next Best Action, Revenue, Risk and Assurance, AI and Governance, AI Assurance, Governance and Handover and the Contact Centre & Customer Care, Real-Time Revenue Moments streams of the catalogue. The lifecycle to live operation and the twelve-line contract are in the workbench, after sign-in.
See it on your own numbers.
Sign in with your work email; the sample is generated for your market and currency; change the values and run again. Or book a demo and we walk it with you.