TL;DR
In GBTA's March 2026 buyer research, 58% of corporate travel buyers said AI has had little or no impact on their program to date, even though interest in specific AI capabilities runs from 83% to 95%. The gap is not appetite. For travel management companies (TMCs), the money is in unglamorous back-office work such as disruption servicing, policy enforcement, data consolidation and expense reconciliation, not in the customer-facing chatbot most pilots start with.
The tension: high interest, low impact
The most useful number in 2026 corporate travel research is a gap. According to GBTA research reported by Business Travel Executive (a March 2026 survey of corporate travel buyers in the US, Canada and Europe, run with Spotnana, Marriott International and Direct Travel), 58% of buyers report AI has had little or no impact on their program so far. Yet the same buyers say they want it: 95% are interested in AI-driven policy recommendations, 89% in automated disruption management and rebooking, 85% in AI-powered traveler support, and 83% in conversational booking.
Supply-side data shows the same shape. Phocuswright's February 2026 research found that over 60% of surveyed travel businesses are experimenting with or scaling agentic AI, but only 6% are actively scaling it and 22% are beginning to scale. Most of the industry is in the middle: lots of pilots, little production.
For a mid-market TMC, that gap is the opportunity and the trap. The opportunity is that buyers are asking for specific, operational capabilities. The trap is that the capabilities buyers rank highest sit on top of data and workflows that most TMCs have not cleaned up.
Where AI delivers value today in a TMC's back office
1. Disruption servicing and rebooking triage
Automated disruption management is the second most requested capability in the GBTA data (89%). It is also a classic back-office problem: when a flight is cancelled, agents spend time reading PNRs, checking policy, finding alternatives across content sources and contacting travelers. AI is well suited to the first 80% of that, namely classifying the disruption, drafting options inside policy, and queueing them for an agent to approve. The human stays in the loop for the booking change itself, which matters because only 57% of buyers say they are comfortable with AI making booking changes or cancellations.
2. Policy interpretation and exception handling
Policy work is text-heavy, rule-heavy and repetitive: is this fare within policy, does this exception need approval, which approver applies. Language models handle the interpretation layer well when paired with deterministic rules for the final check. This is an area where buyers are comfortable: 95% are comfortable with AI recommending negotiated rates and 92% with custom report generation.
3. Data consolidation and reporting
Only 12% of buyers in the GBTA survey say they have consolidated data from a single source, and 63% lack consolidated reporting. For TMCs serving clients across multiple regions, stitching together GDS, NDC, hotel, card and expense data is a large manual cost. AI-assisted mapping and anomaly detection can shorten the work, but the underlying fix is an integration layer, not a model.
4. Expense and reconciliation workflows
The Skift and Navan State of Corporate Travel and Expense 2026 reports that 71% of travelers spend 30 or more minutes on expense reporting and 77% of managers say their expense platform falls short. The same report found 76% of travelers trust AI for straightforward travel and expense tasks, up from 59% in 2023. Matching receipts, itineraries and card feeds is exactly the kind of bounded task where AI can propose and a person can confirm. (Note that Navan is a vendor in this space, so treat the survey as directional.)
5. Leakage detection
Skift and Navan report that 80% of travelers say they sometimes book off-platform, and GBTA found 72% of buyers cite travelers finding cheaper options outside managed channels as their top hotel pain point. Spotting leakage patterns and feeding them back into policy and supplier negotiations is analytical work that AI can do continuously rather than at quarterly reviews.
Where it is overhyped or premature
Fully autonomous booking changes. Comfort drops sharply when AI acts rather than advises: 57% for changes and cancellations, and 64% for calendar access. Agents that act on a traveler's behalf need audit trails and rollback that most TMC stacks do not yet have.
Customer-facing chatbots as the first project. The most visible pilot is also the riskiest. In Moffatt v. Air Canada (2024), a British Columbia tribunal held the airline responsible for incorrect bereavement-fare information given by its website chatbot, awarding roughly CAD 650 plus interest and fees and rejecting the argument that the chatbot was a separate entity. The sum was small; the precedent is not. A TMC is liable for what its assistant tells a client's traveler. We covered the pattern in why customer service rollouts in travel get pulled back offline.
"Agentic" as a label. Phocuswright's numbers show only 6% actively scaling. Many deployments described as agents are scripted workflows with a language model on the front. That is not a criticism, as scripted workflows are often the right answer, but it should shape budget and expectations.
Implementation risks and mitigations
| Risk | What it looks like in a TMC | Mitigation |
|---|---|---|
| Fragmented data | Content from GDS, NDC, direct connects and hotel sources with inconsistent identifiers; 63% of buyers lack consolidated reporting | Build a normalized trip and booking data layer before any model work |
| Hallucination in client-facing answers | Wrong fare rules or policy statements given to travelers | Retrieval from approved policy documents only, cite the source in the answer, human review on anything that changes money |
| Action risk | AI rebooks or cancels incorrectly | Propose-then-approve flow, idempotent operations, full audit log |
| Integration cost | Each supplier and client tool needs a connector | Prioritize two or three high-volume workflows; budget for integration at least as much as for the model |
| Privacy and compliance | Passport data, payment data, traveler location | Data minimization, regional processing, PCI DSS scoping for anything touching payment |
| Change management | Agents distrust or work around the tool | Involve servicing agents in design; measure handle time and error rate, not just usage |
Is your TMC ready? A checklist
- Can you list your top five servicing workflows by volume and average handle time?
- Is booking data from all sources available in one normalized store, queryable by an engineer?
- Do you have a machine-readable version of each client's travel policy?
- Can every automated action be traced, reversed and attributed?
- Have you defined which actions AI may propose and which it may execute?
- Do you have a baseline (handle time, error rate, cost per trip) to measure against?
- Is there a named owner on the operations side, not only in IT?
If you cannot answer yes to the first three, the highest-return project is probably data and integration, and the model comes second.
Where custom software fits a mid-market TMC
Large TMCs and platforms such as the big GDS and online booking vendors will keep shipping generic AI features. A mid-market TMC rarely wins by out-building them on a chatbot. It wins by encoding its own servicing knowledge, client policies and supplier relationships into tooling that a generic product cannot see. In practice that means API development and integration work to connect booking, card and expense data, and targeted custom software development for the workflow layer, such as a disruption queue with policy checks and approval steps. Our write-up on multi-supplier travel aggregation covers the data side. More broadly, the travel industry is a place where the integration work is usually harder than the AI work.
Sources
- GBTA research shows gaps in AI adoption, Business Travel Executive (March 2026 survey)
- 61% of travel businesses surveyed experimenting with or scaling agentic AI, Phocuswright
- The State of Corporate Travel and Expense 2026, Skift (vendor-sponsored by Navan)
- BC Tribunal Confirms Companies Remain Liable for Information Provided by AI Chatbot, American Bar Association
Next step
If you run a travel management company and want to find out which of your servicing workflows are ready for AI and which need data work first, we offer a 30-minute AI-readiness conversation.