讯和数聚DATAJOINT
中/EN
Agent · Lingxi · Chengtu
GAFocus 2026Travel / Airline

Booking Agent

Hand the messiest airline ticket workflows to an auditable AI

Book a demo of this agent

LINGXI · CHENGTU

Booking Agent · Order #ED-884219

LIVE

Customer request

"Change my Apr 25 Qingdao flight to Apr 28, keep under ¥1,200, preferably after 8am."

Agent execution

Fetch travel policy · Lingxi KB

0.4s

Airline re-ticket · 3 candidates

1.2s

Rank by budget+pref · Pick A

0.3s

Await finance approval (Chengtu client #F-8821)

pending

Recommended A · rule-checked

CA1893TAOPEK

¥1,180

in budget

Policy passafter 8ame-itinerary

3.0×

Throughput

94%

Hit rate

2.1s

Avg latency

№ 01 · Scenarios

What it solves in real business

01

Peak-hour rescheduling

During international peak seasons, a single reschedule takes human agents ~12 min. With the agent, it drops below 90 sec with 8× fewer errors.

02

Multi-carrier rule differences

One order spans multiple cabins, carriers, and fuel surcharges. The agent reads NDC/GDS rules live and returns viable reschedule options with fare diffs in one pass.

№ 02 · Workflow

How the agent runs end-to-end

01

Trigger

User intent / system event / schedule

02

Retrieve

Private KB + business APIs

03

Reason

Best-in-class model + industry Skills

04

Act

System writes / human approval hook

05

Reply

Response + audit + feedback loop

№ 03 · Integrations

Plugs into the systems you already run

Amadeus
Sabre
Airline NDC
Enterprise OA / approval
Internal CRM

№ 04 · Guardrails

How we handle failure, over-reach, hallucination

01

Human-in-the-loop

If fare diff exceeds the client's threshold, the agent does not execute — a human reviewer confirms first.

02

Tenant isolation

Each client is a separate tenant; carrier pricing strategies are never cross-visible. Every order carries an operator audit log.

03

Hallucination fallback

Flight times and fares must come from live GDS/NDC APIs — never generated. On retrieval failure, the agent says "please confirm manually".

№ 05 · Detailed capabilities

Going deeper

Background

Outbound ticket businesses see steep peaks, frequent rescheduling, and heterogeneous rules. Legacy OTA systems bottleneck on manual handling — slow response, high error rate.

Core capabilities

  • Natural-language order entry: customer or agent describes the intent in one line
  • Live rule parsing: carrier fare classes and change policies interpreted in real time
  • Full-trail audit: every step from intent to ticket is traceable
  • Multilingual service: CN/EN plus minor languages auto-handled

Typical customers

Cross-border e-commerce, trading companies, travel service providers. One outbound customer lifted agent productivity 4× and cut monthly error rate from 2.1% to 0.25% in the first 3 months.

Deployment

On-prem to customer network, integrated with GDS / NDC / direct airline channels. Runs on OpenClaw / Hermes / our own orchestrator.

№ 06 · Next step

Want to see this agent run on your business?

30-min scoping → 1-week proposal → 6-week PoC. No hard sell.

Book a 30-min consultation