How AI assistants are quietly taking over corporate travel

Talk to anyone running corporate travel this year and the same complaints come up. Too many approvals. Systems that don’t talk to each other. An expense report that should take 10 minutes and somehow consumes 45. What’s changed is who does the work. AI assistants have quietly taken over a chunk of the travel management workflow — not with a big rollout, but by absorbing the repetitive, rules-based tasks that used to swallow a Tuesday afternoon. It’s happening in finance teams, in operations, in travel programmes across every sector. The timing is no accident. Global business travel spending is forecast to hit $1.69 trillion in 2026, an 8.1 per cent jump on the year, according to the Global Business Travel Association. In Europe alone, the same GBTA index puts 2026 spend at €389.9 billion ($448.2 billion), up 8.2 per cent on the year, with the region’s figure climbing to €414.5 billion in 2027 — and global spending on course to pass $2 trillion by 2029. Volumes are up, budgets are tighter and every trip gets more scrutiny than it did three years ago. The appetite hasn’t cooled, either: nine in 10 business travellers now call work trips an essential investment or a necessary cost, up eight percentage points in a year (Skift and Navan, 2026). That mix is what makes the administrative drag around booking, approval and reporting so costly, and it’s why companies are finally moving to automate it. Rewriting the booking process Here’s how it used to work. Someone needed a flight, opened a booking tool, searched, checked the fare against policy, submitted a request, then waited for a manager to notice. By the time approval landed, the price had usually moved. Every step was manual, and every step could break. AI assistants take most of that out. The system already knows what a traveller is allowed to book, so compliant options come up first. It picks up on preferences — seat, airline, time of day — and weights results accordingly. Most people now choose from a shortlist that was filtered before they started looking. The more interesting change happens before the booking goes through. Pre-payment policy enforcement blocks out-of-policy options at the moment of selection instead of flagging them three weeks later in an expense review. The policy stops being an audit and starts being a guardrail. There’s a good reason this matters. Skift and Navan’s 2026 State of Corporate Travel and Expense survey — which drew on 1,798 responses across the US, UK, France and Germany, including 929 business travellers and 869 travel and finance managers, collected in August 2025 — found that 80 per cent of business travellers still book off-platform at least occasionally, usually because they think they’ll find a cheaper fare or a more convenient option somewhere else. Every one of those bookings is a hole in the company’s data: a negotiated rate left on the table, a compliance gap, an employee the company can’t locate if a trip goes sideways. Pulling that spend back into the managed channel comes down to better inventory and a booking experience people don’t want to escape. Approvals: where the time actually goes Nobody who has worked in corporate travel needs this spelled out. A request goes in. The approver is stuck in meetings. The fare creeps up. The traveller either waits and pays more or books off-policy and hopes nobody checks. AI assistants deal with this by reading context. A junior employee booking one domestic night doesn’t warrant the same review as a VP asking for a last-minute business class seat to Singapore, and the system can tell them apart. Routine trips clear on their own; the exceptions get flagged. Managers stop signing off on things that were plainly fine. The problem underneath is not small. The same survey found that 71 per cent of travellers spend half an hour or more on a single expense report, and 36 per cent spend over an hour. At the same time, 29 per cent of organisations still process expenses by hand, up from 23 per cent two years earlier. Manual workload is rising in step with travel volumes, and something has to absorb it. Expense reporting is where people feel it Ask a frequent business traveller what they hate most about travel admin and you’ll hear it straight away: the expense report. Keying in receipts by hand. Digging through a bag for a taxi receipt from Tuesday. Trying to remember what that dinner in Frankfurt was for. Coding everything to the right GL account, submitting, waiting, getting it kicked back for a blank field, redoing it, waiting again. AI tools step in at the moment of the transaction rather than at the end of the trip. The receipt gets photographed and filed on the spot, the charge matches itself to the card statement, and the GL code applies based on the merchant category and the company’s own rules. Modern tools capture dozens of fields per transaction — merchant, location, cost centre, attendees, business purpose — without the traveller typing a thing. Some platforms now log more than 130 data points on a single transaction (Navan, 2026). “There’s a real split opening up right now between companies that have made this shift and companies still running manual T&E processes,” says Maria Budekhina, CEO of GetOffers.com, the AI-powered B2B platform for travel. “It’s not always about budget. We work with mid-sized companies that have implemented AI-assisted expense tools and are genuinely outperforming larger competitors on cost control and travel programme efficiency — in some documented cases, cutting travel costs by around 20 per cent after moving off manual workflows.” Trust is catching up to the capability. In the 2026 survey, 76 per cent of business travellers said they trust AI for standard travel and expense tasks, against 59 per cent two years before. In the same survey, 88 per cent of managers said New Distribution Capability (NDC) delivers real savings and better booking options. Since adoption is what feeds the system clean data, and clean data is what makes everything downstream work, that number carries more weight than it first appears. Reporting and decisions Pull the data. Build the spreadsheet. Send the report. Do it again next month. By the time a travel report reached the person who needed to act on it, the figures were weeks stale. Decisions got made on old numbers, and supplier negotiations happened without a current read on spend. This is where the survey turns up its most awkward finding. Among travel and finance managers, 80 per cent say they’re confident in their access to spending data, but only 40 per cent actually have it in real time. That gap between what people think they can see and what they can see is exactly where overspending hides, because nothing surfaces until month-end. Live reporting closes it. Plain-language queries let finance teams ask a question directly — how much did we spend on Frankfurt hotels last quarter, which cost centres are booking outside policy, where are our negotiated rates sitting unused — and get an answer in seconds instead of after an IT ticket. The level of detail would have seemed like overkill a few years ago. Capturing dozens of fields on every transaction lets you connect things that were never connected before: whether booking lead time affects compliance, which approval rules create the most friction, which destinations generate the most off-policy spend. Travel managers who used to compile reports are using the data to make arguments and win them. That appetite for one connected view is showing up in what companies say they want. In the same survey, 77 per cent of travel and finance managers said they’d prefer a single all-in-one T&E platform, up from 66 per cent a year earlier. Fragmented tools are why the visibility gap won’t go away, and buyers have worked that out. Where AI still falls short None of this is universal, and the switch is rarely clean. Legacy systems drag their feet on integration. Some finance teams have run the same workflow for 15 years and see no reason to touch it. Smaller companies often don’t have the IT capacity to set these platforms up properly, let alone tune them to their own policy rules. The resistance is real, and it isn’t irrational. There’s also work AI does badly. A client relationship that genuinely justifies an out-of-policy hotel upgrade. A reroute forced by fast-moving political trouble. A six-leg itinerary spread across four booking systems that fits no template anywhere. Those still need a person. The best travel programmes know precisely where that line sits: let the software clear the routine, and keep human attention for the calls that need judgement. What travel managers should do now Pick one thing to start with. Expense reporting is the usual entry point because the savings show up fast and you can measure them. Run a pilot with a single team, time the processing before and after, and let that number argue the case for a wider rollout. It’ll do a better job than any vendor pitch. Weigh integration over features. An assistant that plugs cleanly into your accounting system, your HR platform and your corporate card will beat one with a longer spec sheet and poor data flow every time. What you’re buying is the connections, not the features on their own. And say what you’re doing. Travellers and finance staff need to know when an AI is making calls on their behalf; being upfront heads off the resentment that builds when people find out after the fact. The companies that get through this transition with the least friction are the ones that brought their teams in early. AI assistants haven’t replaced travel managers. They’ve taken over the parts of the job that were never worth a travel manager’s time to begin with, which, in a $1.69-trillion industry where every trip has to justify itself, suits just about everyone.