Frequent flyer member service where every answer is pinned to the member's account and the rule in force
Every answer pinned to the member's account and the rule that decides it.
See one case, screen by screen ↓Frequent Flyer Member Service Assistant is AI for frequent flyer member service: seven agents work beside the loyalty service agent during the call, read the member's own account, find the program terms clause and partner chart in force on the date that matters, replay the points and cite every value. When two rules disagree it says so, and the service agent decides what is said and done.
How to choose AI for airlines: a buyer's guide to agentic AI solutions →
Updated
Frequent flyer member service is hardest on the calls that matter most. A Gold member asks why his partner flights did not count toward Platinum, and the answer sits across a 64-page program terms document, a partner earning chart that changed on July 1, a partner bulletin with its own effective date and 214 lines of his own activity. The service agent reads the ledger in one screen, searches the rule library in another and works out the qualifying points by hand while a high-value member waits on hold.
The risk sits where two documents disagree. The program terms say the partner chart in effect on the date of travel applies; the partner's bulletin says its fare change covers only tickets issued from July 1. Nobody owns that conflict, so goodwill miles close the call, two agents give two answers, and the Program Rules Manager hears about it weeks later from a complaint. When the rule is finally clarified, nobody knows which members were told the old answer.
Estimated split for one unclear partner earning rule, from the first member call to every agent giving the same answer, by hand and with the solution.
Seven agents read the account, find the rule in force, replay the points and check every answer; the service agent decides what is said and done.
Friday, October 16, 2026, 10:31 am. Gold member Graham Ellison — Compass 4471 2093, Raleigh, member since 2014 — calls Corvane Airways' Gold & Platinum desk. He is 1,240 Qualifying Points short of Platinum, and the answer to why sits in four documents, two of which disagree. Here is what happens next, in the working solution.
Nadia's day opens on what needs her: 23 calls on the Gold & Platinum desk, 61 of 64 answers used as given, a callback promised to Carla Mendes by 2:00 pm. At the top, Graham Ellison is on the line — the agents have already found his Aurelis flights to Athens and a rule that is not clear-cut.
“Graham Ellison is on the line · Gold 4471 2093 — The agents found the flights and a rule that is not clear-cut.”
Graham says he flew Aurelis to Athens in September and none of it counted. Nadia asks the agents why. Six agents read his ledger, §4.3 of the program terms, two chart versions and a partner bulletin in 3.4 seconds. The answer is pinned to his account: Aurelis 418 LHR–ATH on Sep 12 and 419 ATH–LHR on Sep 19, class G, 1,494 miles flown each, earning 0 QP and 374 miles each — on a ticket issued Jun 22, before the change.
“Both Aurelis flights earned 374 miles and 0 Qualifying Points each: on his travel dates the partner's chart counts fare class G as a basic fare, and basic fares earn no QP under Program Terms §4.3.”
Every value carries a numbered source. One click opens Compass Program Terms v2026.2, §4.3 Partner flights, effective Jul 1, 2026, page 18, with the passage highlighted: partner flights earn “according to the partner earning chart in effect on the date of travel”, and basic economy fares earn “redeemable miles but no Qualifying Points.”
“A read-only copy is kept with the conversation for seven years, so a supervisor, an auditor or a DOT request sees the same passage.”
The answer checker reads the answer against every source and finds the conflict. By date of travel (Program Terms §4.3), 0 QP: he stays Gold, 1,240 short. By date of issue (Aurelis bulletin PB-26-14), 747 QP per flight at 50%: 1,494 QP, 100,254 and Platinum. Confidence drops to 61%, and the conflict goes to Hannah Lindqvist, the Program Rules Manager, as unclear rule U-031.
“Two signed documents disagree, so I will not go above 61%.” — Answer checker
The next-step agent proposes what to do, and says who decides: open a status review for Owen Castillo for 1,494 courtesy QP while §4.3 is clarified (recommended); email Graham a written summary with the clauses attached; or explain the rule and close without a review (not recommended). Each one can be approved, edited or rejected.
“Next steps the agents propose — nothing happens until you approve.”
Nadia approves the status review: SR-2291 opens with Owen Castillo and Graham hears back by Tuesday, October 20. She cannot give the points herself, and the answer rules say never to promise a status outcome. The script under “What to say” is in plain words, with no internal codes.
“Aurelis says that change only applies to tickets issued from July 1 — yours was issued June 22 — so I've asked for a status review. You'll hear from us by Tuesday, October 20.”
Owen opens SR-2291: what Graham asked, both readings, the gap and the next-step agent's recommendation. The 1,494 courtesy QP are inside his 5,000 QP limit. He approves with the reason code “Partner reclassified fare — ticket issued before change” and signs. Graham moves to 100,254 QP and Platinum, and is emailed the outcome today.
“I want the rule and the account on one page.” — Owen Castillo
Hannah sees U-031 with both passages side by side and who it touches: 2,318 members flew Aurelis class G on tickets issued before Jul 1, on 3,904 segments, with 2,916,288 QP to re-credit; 37 would move up a tier and 64 past answers said the opposite. She publishes §4.3.2 — when a partner changes how it classifies a fare, the partner's stated basis applies — in force Oct 16, and signs.
“When two documents disagree I hear about it three weeks later, from a complaint.” — Hannah Lindqvist, before
Rule update RU-118 replays the clarification on every member and past answer: 2,318 members to re-credit, 37 moving up a tier, 64 to call back because they were told “0 QP is final” — Samuel Fox this morning, now +1,494 QP. The call-back script and email are drafted; nothing is credited or sent until a supervisor approves.
“Nothing is credited or sent until a supervisor approves.”
Every read, answer, check, edit, flag and decision is on the audit trail with the record it touched: the account reader's 214 lines, the answer checker's conflict and 61%, Nadia's approval, Hannah's signed clarification. It is kept seven years, and an evidence pack for one member or one rule exports from the same screen.
“Show me what the member was told and why.” — Celia Grant
The same call, seen by the four people who carry it — what their day looked like, and what it looks like now.
Seven agents read the account, find the rule in force, replay the points and check every answer; the service agent decides what is said and done.
Turns the service agent's words into the question, the member and the activity in scope, from the call transcript and the CRM contact.
Reads the member's activity ledger, bookings, partner credits, card points and status history.
Finds the clause and the version in force on the date that matters — program terms, partner earning charts, promotions and bulletins — and compares new documents with the ones they replace.
Replays Qualifying Points and miles for each activity under each reading of the rule, line by line, by partner, chart version and fare class.
Reads every answer against every source before it is shown, sets the confidence, finds conflicts between documents and opens an unclear rule for the Program Rules Manager.
Proposes steps inside the person's authority — a status review, a written summary, an explanation — and drafts the script, the email and the review.
Traces wrong-answer flags to their cause and finds every conversation with the same answer; when a rule changes, replays it on affected members and past answers and lists who to call back.
Decide every answer, review and rule. The agents propose; a named person decides.
Follow-ups stay in the same conversation, and every answer points to the ledger line, clause or policy it came from.
What does he need for Platinum by Dec 31?
Without the review he needs 1,240 QP by Dec 31. One Corvane round trip RDU–DEN in fare M (base fare $412) earns about 2,060 QP at 5 QP per dollar. If SR-2291 is approved he reaches 100,254 and needs nothing more.
Can I give him the points?
No — service agents cannot add courtesy QP. A supervisor can approve up to 5,000 QP per review, and members hear back in 2 business days. That is why the status review is the proposed step.
Why is the confidence only 61%?
Four checks passed: the flights, the chart version, the miles and the partner file. The fifth failed: §4.3 counts the date of travel, the partner's bulletin the date of issue. Two signed documents disagree, so I will not go above 61%.
Which answers were flagged wrong today, and why?
12 of 47 flags were confirmed wrong. The biggest: F-0927 — a lounge answer used policy v4 instead of v5; 38 conversations had the same answer and 21 members are being called back.
13 screens from the working solution, on its sample data. Pick one to see it large.
The thresholds the agents work inside and the rules every answer is checked against are settings, saved and on the audit trail.
| Setting | Default | Choose from |
|---|---|---|
| Confidence below which an answer is marked “not clear-cut” | 85% | Set in percent; applied to every answer |
| Supervisor approval limit per review | 5,000 QP | Set in QP; above it the review goes to the Head of Loyalty |
| Courtesy QP a service agent may give | 0 QP | Set in QP; at 0 every courtesy credit needs a review |
| Status review answer time | 2 business days | Set in business days; each review shows its clock, and late ones are marked overdue |
| Every value cites its source | On | On · Off |
| Never promise a status outcome | On | On · Off |
| Use only signed, live versions | On | On · Off |
| Offer a written summary on status calls | Off | On · Off |
Every figure is labelled: a target the solution is built to, an estimate, a typical published result, or a proven one.
“demo” = seen in the working solution, on its sample program data · “target” = the design goal, measured in the live solution · “estimated” = our estimate · “typical” = published figures (Reuters review of airline filings, March 2026 — co-brand cash about four times adjusted operating income at one US network carrier in 2025; Deloitte 2025 Consumer Loyalty Survey — about 1 in 5 of the most loyal flyers feel recognized; an agent-assist vendor's published customer story — a US carrier's 48-agent pilot cut handle time 10%). DOT opened its inquiry into the four largest US loyalty programs on Sept 5, 2024, under 49 U.S.C. 41712. Corvane Airways, Corvane Compass, Aurelis Airlines and every person named on this page are fictional — characters and sample data in the working solution.
Frequent Flyer Member Service Assistant is AI for frequent flyer member service: seven agents work beside the loyalty service agent during the call, read the member's own account, find the program terms clause and partner chart in force on the date that matters, replay the points and cite every value. When two rules disagree it says so, and the service agent decides what is said and done.
Usually the partner's fare class earns miles but no qualifying points under the chart in force. The earning calculator replays each partner flight by chart version and fare class — in the demo, class G at 0% QP and 25% miles — and pins the answer to the member's own ledger lines and the clause, so the agent can show exactly which flight did what.
The agent asks in plain words during the call and gets a short answer first, then the member's activity, the rule as it read on the date, the calculation and a script of what to say. Follow-ups stay in the same conversation. In the demo the agents answered from 11 sources in 3.4 seconds.
The answer checker reads every answer against every source before it is shown. When two documents disagree it marks the answer “not clear-cut”, shows the outcome under each reading, lowers the confidence and sends an unclear rule to the Program Rules Manager with both passages side by side.
When the Program Rules Manager publishes a signed clarification, the feedback and rule-update agent replays it on every affected member and every past answer, and lists the re-credits, tier moves and call-backs. In the demo one clarification found 2,318 members and 64 past answers; nothing is credited or sent until a supervisor approves.
No. The agents are read-only on the ledger, never post a credit and never promise a status outcome. The next-step agent only drafts; a service agent approves each step, a supervisor approves courtesy points up to the limit with a reason code and signature, and the Head of Loyalty decides above it.
Every read, answer, check, edit, flag and decision is on the audit trail with the record it touched, kept seven years, with the passage each answer relied on. An internal auditor can export an evidence pack for one member or one rule — what the member was told, and why.
The Agentic Solution Engine builds and deploys it from your requirements and documents — your program terms, partner earning charts, status review policy and the systems your desk uses — and it goes live once every quality gate has passed. A first-time setup walks through sources, roles and a test question before the desk goes live.
Stolen miles stopped before the award is flown, and honest members never locked out.
1-min demoMissing Miles Claims ProcessingMissing miles credited in minutes, each one tied to the flown coupon and its record.
1-min demoLoyalty Partner Billing ReconciliationEvery partner statement tied out line by line, every missed mile billed or credited.
1-min demoWe'll run Frequent Flyer Member Service Assistant on your own program terms, partner charts and a sample of past member questions.