Every travel technology vendor now describes their product as AI-powered. The label has stopped carrying information, which makes evaluation harder rather than easier.
This is a framework for cutting through it — the categories that actually exist, the questions that separate capability from marketing, and an honest account of where different approaches fit.
First: what problem are you solving?
The most common evaluation mistake is shopping for “AI” rather than for a specific bottleneck. Before looking at any vendor, measure where your time actually goes.
For two weeks, capture volume, human touch time, and handoffs for each transaction type: new bookings, changes, cancellations, refunds, quotes that do not convert.
Most TMCs are surprised. The assumption is that new bookings dominate. The reality is usually that changes, cancellations and rebookings consume disproportionate time — lower volume, much higher touch, arriving at the worst moments.
That finding should drive your shortlist. Buying booking automation when your bottleneck is servicing is the most expensive mistake in this category.
The four categories of AI for TMCs
1. Booking and request automation
Parses travel requests, searches fare sources, ranks options, and in some cases books.
Best for: TMCs where request handling and fare search consume the most agent time.
Watch for: whether it actually books or only recommends. Most tools in this category recommend.
2. Policy and compliance automation
Encodes client travel policy as rules, evaluates transactions against it, routes exceptions and approvals.
Best for: TMCs with varied client portfolios where compliance is currently manual or retrospective.
Watch for: whether evaluation happens during search or after selection. After-selection flagging is much less useful.
Usually the best first automation — rules are already written down, and results show within one reporting cycle.
3. Finance and reconciliation automation
Three-way matching across bookings, supplier invoices and payments; commission tracking; period close.
Best for: TMCs losing days to month-end close.
Watch for: handling of partial refunds with residual value and aggregated card postings. Ask for a live demonstration of both.
Often self-funding — recovers uncollected commission, so it can pay for itself in recovered revenue rather than saved hours.
4. Servicing and disruption automation
Autonomous changes, cancellations, rebookings and disruption handling.
Best for: TMCs whose senior agents are absorbed by servicing.
Watch for: escalation quality. This is the highest-risk function; a system that fails to escalate appropriately is worse than none.
Highest value, highest risk — automate it after the other layers are reliable.
The autonomy test
This is the single most useful evaluation lens. Adapted from the self-driving vehicle levels, which map unusually well:
| Level | Description | Reality |
|---|---|---|
| 1 — Assisted | Shortcuts and saved profiles | Legacy TMC desktops |
| 2 — Augmented | Recommends; human actions everything | Most “AI booking” tools today |
| 3 — Conditionally autonomous | Completes routine transactions end to end; escalates the rest | Where capacity actually changes |
| 4 — Highly autonomous | Handles most transactions including servicing; learns continuously | Leading edge |
| 5 — Fully autonomous | No human ever | Does not exist, and should not |
The question to ask every vendor: “What proportion of our transactions would this complete end to end, with no human, on day one — and after six months?”
A vendor who cannot answer with a number is selling Level 2. That may still be useful, but it will not decouple your volume from your headcount, which is usually the actual objective.
Seven questions for any vendor
- What proportion of transactions complete with no human touch? (The autonomy test.)
- Walk me through a partial refund with residual value and a change fee. Best single test of finance capability.
- What happens when the only available option is out of policy? Reveals operational understanding.
- Can policy differ per client, per traveller grade, per trip type? Single-tenant policy models do not survive a real TMC portfolio.
- Does it integrate with my specific GDS and mid-office, or require replacement? Rip-and-replace is where TMC projects die — you cannot suspend operations to migrate.
- Does it learn from how my team resolves exceptions? Static rules plateau; learning systems improve.
- What does implementation actually require from my team, in hours? The honest answer is rarely “almost nothing.”
What to be sceptical about
“AI-powered” with no specified function. Ask which model does what. Vagueness usually indicates a rules engine with a new label.
Demos on vendor data. Always demand a demonstration against your own transaction patterns. Demo environments are curated.
Auto-match or accuracy rates without context. “95% accuracy” is meaningless without knowing the transaction mix and whether it is at go-live or after six months of tuning.
Anything requiring you to replace working systems. Occasionally justified. Usually a sign the vendor cannot integrate.
Headcount-reduction framing. Most TMCs do not have a demand problem; they have a capacity problem. The realistic gain is handling more volume with the team you have, not cutting the team. Vendors leading with redundancy are usually overselling.
A sensible sequence
Whatever you buy, sequence by rule clarity and volume, divided by exception rate:
- Compliance — rules already documented, results in one cycle
- Finance — binary success, often self-funding
- Booking — needs the policy layer first, or you just make non-compliant bookings faster
- Servicing — highest value, automate once the layers beneath are trusted
Run each in advisory mode alongside your existing process for one full cycle before switching over. This surfaces configuration errors safely and — more importantly — builds agent trust, which is usually the real constraint rather than the technology.
Where VantisCorp fits
To be direct about our own position rather than pretending this page is neutral:
VantisCorp covers all four categories with specialised agents — Booking, Compliance, Finance and Servicing — layered onto your existing GDS and mid-office rather than replacing them. Across TMCs on the platform: 40% more bookings with the same team, 60% faster turnaround, 98% policy compliance, 3x faster reconciliation.
We are likely a good fit if your bottleneck is post-booking — servicing volume or reconciliation — or you want one platform across all four workflows without a development team.
We are likely not the best fit if you only need booking automation and already have compliance, finance and servicing well handled. In that case a specialist booking tool may serve you better, and we would rather tell you that than sell you four agents you do not need.