Business travel managers expect artificial intelligence to manage or authorise an average of 37% of their organisations’ corporate travel spending by 2028, according to research commissioned by Manchester-based payments technology company Conferma.
Conferma calculates that the expected share could represent about $700bn globally when applied to the Global Business Travel Association’s 2025 forecast for business travel spending. The $700bn figure is a forecast extrapolation based on survey expectations, not a measured amount of purchases already made or contractually committed to AI systems.
The study, conducted by Sapio Research in September 2026, questioned 600 corporate travel and finance decision-makers at organisations employing more than 200 people. It included 200 UK respondents, 200 in the US and 200 across Australia and New Zealand, providing a picture of expectations in those markets rather than a direct measurement of all businesses worldwide.
More than a quarter of respondents, 28%, expected AI to manage over half of their travel budgets within two years. The expectation was more pronounced among organisations spending at least $6.8m, equivalent in the release to £5m, on corporate travel annually: that group forecast AI management of 51% of spend, against 27% among organisations spending under $136,000.
The commercial development envisaged is a shift from AI helping employees search for flights and hotels towards systems that can select suppliers and complete purchases subject to corporate policies. That creates opportunities to reduce administration and accelerate transactions, but also gives rise to new questions about delegated financial authority.
Conferma found that 81% of organisations already had some form of AI purchasing control, policy or payment safeguard, while only 36% described their controls as comprehensive. Respondents reported concerns about autonomous purchasing, including data security, and senior managers appeared more confident about delegation than travel and procurement teams.
Among organisations with safeguards in place, 55% said a human must approve purchases before AI completes them. Other measures included real-time monitoring at 51%, spending limits at 49% and virtual cards with controlled or single-use limits at 37%.
The report forecasts that use of controlled virtual cards within the safeguarded group could rise to 48% by 2028. Such cards can restrict merchant categories, spending amounts or transaction periods, potentially making a payment instrument part of the control mechanism rather than relying on the AI system to interpret every policy correctly.
Mark Ledsham, Conferma’s chief executive, cautioned that assigning AI authority over company funds is different from using it to recommend an itinerary. He argued that limits should remain fixed, traceable and accountable, with human intervention available when transactions fall outside approved parameters.
Finance departments evaluating autonomous purchasing will need clear ownership of payment approvals, audit records, exception handling and supplier disputes. Procurement teams must also consider whether AI systems follow negotiated rates and preferred suppliers, or whether convenience could lead to unauthorised purchases.
There is an important methodological limitation. The sample was drawn from selected English-speaking markets, and the projections depend on respondents’ expectations about technology adoption in 2028. The stated overall margin of error of about four percentage points does not capture uncertainty about future business behaviour or the assumptions used in the global $700bn calculation.
A payment transaction initiated by software creates several possible failure points. An agent might select a supplier that falls outside policy, misinterpret a traveler’s authorisation, or make a valid purchase at an unexpectedly high price. Controls embedded in cards or payment infrastructure can block some prohibited transactions before funds are transferred, but do not remove the need to review bookings, settle disputes and recover costs. Finance leaders also need to determine who is accountable for an agent’s action when responsibility is divided between the travel platform, payment provider and the company’s own staff.
The study nonetheless identifies a material governance issue already facing companies experimenting with AI agents. Before automated software is permitted to spend corporate money, businesses need to determine how authority is granted, transactions are restricted and accountability is preserved when mistakes occur.
You must be logged in to post a comment.