What Spend Management Vendors Announce, and Why Their Customers Actually Leave

Two independent datasets, assembled separately, point in opposite directions.

On the vendor side, IndustryLens tracked 238 announcements over the last 90 days. Of those, 58 are AI-flavoured — they mention AI, agent, agentic, autonomous, copilot, LLM, Joule, MCP or intelligence. That is 24.4% of everything the category said publicly in a quarter. By contrast, 14 announcements were support-themed (5.9%) and 7 were UX or usability-themed (2.9%). Forty were product updates.

On the buyer side, 371 switching mentions in the review corpus carry a stated reason for the move. Automation or AI accounts for 3 of them — 0.8% — with a single instance in 2026. Usability or dated UX accounts for 102 mentions, 27.5% of stated reasons, with 51 of those in 2026. Support accounts for 37 mentions, 10.0%, with 21 in 2026.

The asymmetry is the finding. The theme the category invests the most announcement volume in is the one buyers almost never name when explaining a move. The theme buyers name most is the one that receives the least announcement volume.

Two cautions before this is over-read. First, announcements and switching reasons are not the same kind of object: a vendor may improve an interface without announcing it, and roadmap work often precedes public messaging by quarters. Second, an AI capability could well be driving purchases without buyers describing it in those words — reviewers write in the language of their own frustration, not in category taxonomy. What the data supports is narrower and still substantial: in the stated language of people who moved, AI is not the reason, and interface quality is.

It is also worth noting that the usability and support figures are not a legacy artefact. Half of the usability mentions (51 of 102) and more than half of the support mentions (21 of 37) are from 2026. The complaint set is current.

The full taxonomy of 371 stated reasons breaks down as follows: usability or dated UX, 102; support, 37; feature depth, 21; price, 21; integrations, 15; multi-country requirements, 4; automation or AI, 3. A further 168 fall into other or uncategorised — a large residual, and one that should temper any claim that the taxonomy is exhaustive.

Within the categorised set, the ranking is unambiguous. Usability leads by a wide margin. Support is second. Price ties with feature depth at 21 mentions each, which is itself notable in a category where discounting is a routine competitive tactic: in this corpus, price is named less than half as often as interface quality.

The verbatim comments show why the usability and support categories overlap so heavily in practice. Reviewers rarely cite one in isolation; they describe a system that is hard to use and a support function that does not resolve the resulting problems.

"Concur support failed to deal with our requests and their platform was outdated. Furthermore, it was really important to us to find an expense platform where the end users can submit their expenses with ease." — enterprise reviewer, G2, 8 July 2026, moving to Navan

"We had many connection issues with Concur and their support team was not helpful" — mid-market reviewer, G2, 22 June 2026, moving to Tipalti

"Concur does not have a clear policy setup for expenses and approvals, and routing items back to employees is a very manual process. Also, integration with Sage Intacct has to be done through a third-party application" — enterprise reviewer, G2, 8 July 2026, moving to Navan

The third quote is instructive because the reviewer describes a manual, high-friction workflow — precisely the problem category AI messaging targets — without reaching for the vocabulary of automation. The gap above may be partly a gap in language, not only in substance.

Each edge below counts reviewers who explicitly named a product as the one they left. The largest single flow in the corpus is Navan from SAP Concur: 27 mentions, 26 of them in 2026, spread across enterprise, mid-market and small-business reviewers. That breadth matters — it is not confined to one segment.

• Payhawk from Spendesk: 20
• Mesh Payments from Expensify: 18
• Payhawk from Kloo: 15
• Tipalti from BILL AP/AR: 14
• Navan from Expensify: 10
• Ramp from SAP Concur: 9 (8 in 2026)
• Brex from Expensify: 7; Payhawk from Expensify: 7; Payhawk from Pleo: 7
• Brex from SAP Concur: 6; Tipalti from SAP Concur: 6; Ramp from BILL AP/AR: 6; Ramp from Expensify: 6
• Ramp from BILL Spend & Expense (formerly Divvy): 5
• SAP Concur from Microsoft Excel: 4; Navan from Egencia by Amex GBT: 4
• Navan from Emburse Expense Professional: 3

SAP Concur is the standout node. Across the corpus, 55 reviewers named it as the product they left, 39 of those in 2026. Their destinations: Navan 27, Ramp 9, Brex 6, Tipalti 6, Mesh Payments 3, Expensify 2, Payhawk 2. Inbound flow to SAP Concur in the same corpus is 4 mentions, all four from Microsoft Excel.

Stated plainly: within this dataset, the only source SAP Concur draws named switchers from is spreadsheets, while seven different vendors draw named switchers from it. That is a directional observation about review text, not a market-share or revenue statement, and the methodology note sets out why the two must not be conflated.

Expensify is the second most frequently named origin, appearing as the departure point in flows to Mesh Payments (18), Navan (10), Brex (7), Payhawk (7) and Ramp (6). Payhawk is the most frequently named destination for European-origin flows in the set, drawing from Spendesk (20), Kloo (15), Expensify (7) and Pleo (7).

Three implications follow from the data, and only these three.

The stated reason set is stable and current. Usability and support are not residual complaints from an older cohort. With 51 of 102 usability mentions and 21 of 37 support mentions landing in 2026, the pattern is being renewed in the most recent reviews in the corpus, which runs to 2 August 2026.

Displacement is currently concentrated. SAP Concur accounts for 55 of the named departures, with 39 in 2026, and Navan alone absorbs 27 of those. Expensify is the other repeatedly named origin. If those two nodes continue to generate outbound mentions at the observed rate, the near-term competitive story in this corpus will remain a story about a small number of origins rather than a broad reshuffle.

Announcement mix and stated demand are not converging. At 24.4% AI-flavoured announcements against 0.8% AI-related stated switching reasons, and 2.9% UX announcements against 27.5% UX-related switching reasons, there is no evidence in this quarter's tracked announcements of the messaging mix moving toward the themes buyers name. Whether that gap narrows is measurable, and IndustryLens will re-run both counts.

The review corpus comprises 3,486 reviews of 13 tracked vendors, collected from G2, Capterra and Trustpilot, with the earliest review from 2020 and the latest dated 2 August 2026. Per-vendor review counts range from 371 (Navan) to 56 (Bill). Within the corpus there are 405 switching mentions in total, of which 371 carry a stated reason; only those 371 are used in the reasons taxonomy. The announcement dataset comprises 238 vendor announcements tracked over the last 90 days, classified by keyword.

• This is self-reported data from people who chose to write a review. Reviewers who have recently switched are more motivated to write than satisfied incumbent users, which biases the corpus toward switching narratives.
• A switching mention is not a churn rate. These counts measure how often a product is named in review text as the one a reviewer left. They say nothing about accounts lost, revenue affected, or share of any vendor base.
• Review volumes differ by vendor, so raw mention counts partly reflect how much each vendor's users write. Average ratings range from 4.84 (Ramp) to 2.13 (Bill, on only 56 reviews) — the latter sample is small enough that the average should not be treated as a stable signal.
• The reason taxonomy is incomplete. 168 of 371 stated reasons fall into other or uncategorised, so the categorised shares describe the classified subset, not the whole.
• Announcement classification is keyword-based, which captures messaging emphasis rather than underlying engineering investment.