S16.4 · Corporate & Business Travel Management
A sub-segment of Corporate & Business Travel Management. Market structure, and how AI is reshaping it.
Corporate Booking & Expense Software is the application layer of corporate travel and expense: online booking tools, policy and approval engines, expense capture, reconciliation and reporting, and integrated T&E suites. SAP Concur, Navan's software product, TravelPerk, Coupa Travel & Expense, Expensify, Emburse, the T&E modules inside Brex and Ramp, Center, Fyle, Mesh, Pleo, Payhawk, Soldo and the booking tools embedded in TMC platforms all sit inside the boundary; human agent servicing sold on a contract, traveler risk services, and the underlying card-issuing and interchange rail sit outside it. The segment owns the application layer on the payment rail, not the rail itself. It also owns the booking action and the expense-capture moment directly — which is exactly why it is more exposed than most enterprise software to an AI agent bypassing the interface entirely, though who captures the commission on an agent-executed booking remains an open industry question.
T&E software market-size estimates diverge sharply by scope, and neither of the leading figures holds up as a ceiling. Research and Markets puts the base at $3.77bn in 2025, rising to $6.1bn by 2030 on a 9.7% CAGR; Fortune Business Insights puts it at $4.08bn in 2025, rising to $12.98bn by 2034 on a 13.8% CAGR. Summing disclosed revenue for just five participants — Navan's $765m TTM, Ramp's roughly $1.0-1.5bn run-rate, Expensify's $138m, plus Brex and Coupa's T&E slice — already approaches or exceeds the smaller estimate. The likely explanation is classification: most market-sizing houses count card-led players as "fintech" rather than "T&E software" and undercount accordingly.
Geographically the segment is mature and developed-market-weighted: North America 38.9%, Europe 27.1%, APAC 16.3%, Latin America 11.9%, Middle East/Africa 5.8% (Research and Markets, 2026). SAP Concur remains the largest single incumbent by enterprise seat count, but SAP discloses no standalone Concur revenue or customer count today — Concur is folded into a broader "Intelligent Spend and Business Network" segment, a genuine information gap rather than an estimate.
The split that matters is subscription versus usage-based revenue, because the two carry materially different margins and AI exposure. Navan is the sharpest disclosed example: 90% of FY2025-26 revenue is usage-based, mostly a 7% take rate on travel bookings, against only 10% from SaaS subscription — and the SaaS line is growing faster (52% year over year) than the usage line (32%). Gross margin reached 71% LTM, up from 60% in FY2024, still short of pure-SaaS norms above 80% because usage revenue carries network and processing cost; net revenue retention runs above 110%. Switching cost bifurcates by tier: SAP, Oracle and Workday ERP/HRIS integration locks in the enterprise base, while the SMB tier switches cheaply and shows thin CAC discipline — Expensify's multi-year revenue stagnation is the evidence.
Adjacency traffic runs in both directions, and neither direction is a defensible moat by itself. Card issuers are moving into T&E software to own the transaction-capture point and deny it to competitors — "card and expense management should be deeply linked," as one platform's CEO put it (Center CEO, March 2025) — while software platforms expand the other way, into cards and fintech, to capture take-rate economics they would otherwise leak to card-led competitors.
Start with what AI does not threaten. Per-employee seat pricing — used by Concur, Expensify, Pleo and Soldo among others — is a headcount proxy, not a work-volume proxy: a seat is issued to every employee who might travel or expense, independent of how many expense reports AI generates on their behalf. Because the buyer is finance or T&E administration, not the line employee, AI automating receipt capture and reconciliation is a straightforward cost-side gain the buyer keeps; it does not by itself shrink the licensed seat base. The seat becomes unnecessary only if an agent bypasses the booking and expense interface entirely — a separate and more consequential risk.
The two revenue types then read genuinely differently. Navan's usage-based revenue is mostly travel-booking commission — a travel-agency-style take rate that requires no regulated position to earn, which is exactly why automation can compete for it. True interchange revenue fires on a purchase happening rather than on work being done, so its exposure is AI-driven travel-volume compression rather than labor automation — and interchange models are increasingly being absorbed into regulated depository institutions, folding that exposure into a genuinely regulated position rather than a purely competitive one.
At the regulated enterprise tier, the compliance case for a human-governed system wins decisively for now. SOX, audit and expense-substantiation requirements mean CFOs will not let an ungoverned agent book or file outside a system that produces an approvable record, and incumbents are racing to become that policy layer — Ramp has built an AI-agent corporate card, and TripGain has extended a Model Context Protocol server into agentic booking approvals (August 2026). On undifferentiated point-to-point bookings, though, commission capture is genuinely unresolved: who gets paid on an agentic booking is a live, unanswered industry question (Hospitality Net, 2026). That makes the booking-commission line, not the subscription or controls line, the segment's real point of AI exposure. The compliance stack itself mostly favors automation rather than blocking it — SOX audit trails, IRS accountable-plan substantiation and VAT reclaim documentation all require a record to exist, not a human signature, and receipt OCR is already table stakes. The one human checkpoint that survives, European payment-authentication rules requiring human-initiated confirmation at payment execution, sits on the payment rail rather than this segment's core software licensing. On balance, the compliance stack does not protect the T&E software seat from AI bypass.