How a Luxury Hotel Group Cut Rate Loading to Four Minutes, and Brought Two Revenue Controls In-House
Powered by Lemma Series
Thread AI Editorial
September 28, 2026
Two Lemma Workers that put a global luxury hospitality brand's rate loading and partner availability checks on infrastructure it owns: rates enter Sabre SynXis from a form in Microsoft Teams with retail pricing gated by a named human approver, and the weekly partner availability report is rebuilt from the group's own data with every flag traceable to a written rule.
4 min
Per rate load, measured on a live submission
3 min
To generate the partner availability report, previously a weekly vendor delivery
1 ticket
Per rate request, down from two or three
The Challenge
Two Revenue Controls, Neither Under the Group's Control
For a luxury hotel group, the rates loaded into the central reservation system are the revenue line, and the weekly check that hotels have not closed global partner programs while keeping their own retail offer open is the control that protects it. Both ran on a third-party vendor's RPA bots outside the company. Rate requests arrived on a macro-heavy Excel form hosted on the vendor's SharePoint and raised two or three ServiceNow tickets each. The partner availability report came once a week, a week behind, with flagging logic the group could neither read nor download. The distribution lead's brief was blunt: high adoption, high usability, no errors.
The Solution
Governed by a Named Human Approver, Auditable to the Rule
Two Lemma Workers replaced the bots and moved both processes onto systems the group controls. Rate loading starts from a form inside Microsoft Teams, holds retail rates for a named regional revenue director, and drives SynXis through a browser agent because the group does not leverage SynXis’ API. The partner availability report is rebuilt from a SynXis restriction subscription joined to mapping files on the group's SharePoint. What makes both acceptable in a process that sets prices is accountability: every rate run leaves one ticket, a result table and a recording, and every flag traces to a rule the group owns and can change.
What it does
1
Starts where hotel teams already work
A validated form inside Microsoft Teams. Submit starts the Worker, assigns one run ID per property, and opens a single ServiceNow ticket that the completion email updates on the same ID.
2
Gates public pricing behind a named human approver
Negotiated rates load directly. Retail rates wait in a Handoff until the regional revenue director named in the group's approval file signs off, so no unapproved price reaches the booking engine.
3
Loads the rate and keeps the evidence
The browser agent creates the rate and its seasons in SynXis, retries login failures, and writes status and failure reasons to a SharePoint list the group owns, with a recording of every run.
4
Rebuilds the partner availability report on the group's schedule
Pulls the SynXis restriction subscription, joins the group and rate-template mapping files, recomputes the flags, writes the Excel to a SharePoint folder and emails the regional revenue directors. New hotels join through chain-level subscriptions rather than a vendor change request.
The Results
Minutes per Run, One Record per Request, and the Logic Back In-House
A rate load completes in about four minutes, timed by the group's distribution lead on a live submission. Both workflows were running end to end in production roughly fifteen weeks from the first scoping call. Each request raises one ticket instead of two or three, so ticket volume once again reflects real work. The partner availability report generates in about three minutes against a vendor delivery that arrived weekly and a week behind; it reached the first production batch of hotels in February 2026 with the flagging logic documented on the group's SharePoint. The forms, mapping files and run history now sit in the group's own tenant.
Does This Sound Familiar?
A revenue-critical process runs on a vendor's bots, and the vendor holds the forms, the data and the logic. Your prices are keyed into a system with no API. A control report arrives a week behind, and nobody can produce the rule that raised a flag. Requests reach the team by email because the form is harder than the inbox.