After Ramp’s corporate customers adopted its spend management service, their median expenses fell 7.5% year over year. The company’s annual revenue has surpassed $2 billion, and its growth over the past year reached 25%. Our judgment is that competition in enterprise software is shifting from “providing data” to “doing the work for people,” with spend management potentially serving as a mature entry point.
What this is
Ramp initially offered corporate cards and expense management, then expanded its software into budgeting, reimbursements, payments, and vendor management. In this interview, co-founder and CEO Eric Glyman went further into Ramp’s AI transformation: hiring was halted for a quarter to improve internal efficiency, a testing framework was built for running models, and the company developed an “enterprise brain” that understands company data.
This enterprise brain is not an ordinary chatbot. It is an AI Agent—a program capable of autonomously completing multistep tasks—designed for internal workflows. It can combine company policies, historical transactions, and specific circumstances to determine whether an expense is reasonable, identify who should approve it, and help complete subsequent processing. Ramp also argues that organizational information should be readable by the system, rather than existing only in employees’ heads.
Industry view
This approach has practical advantages: financial processes have clear rules, high repetition rates, and verifiable outcomes, while companies can readily calculate the savings. By placing AI together with payments, cards, and expense data on one platform, Ramp can form a closed loop more easily than vendors that merely provide a chat interface.
But the risks are equally clear. Once the enterprise brain participates in approvals and payments, it must explain the basis for its decisions, manage the boundaries of its permissions, and avoid amplifying incorrect rules across many transactions. Companies may also be reluctant to hand budget, procurement, and vendor information to a single platform. Glyman’s mention of two extreme views on AI shows precisely that management teams have yet to reach a stable consensus. Whether software companies can replace some positions affects not only model capabilities, but also their internal hiring and customers’ trust.
Impact on regular people
For enterprise IT: Budgeting, reimbursement, and procurement systems will gradually connect with AI capable of executing tasks. Companies must first organize permissions, data, and audit records before deciding which parts of these workflows can be automated.
For individual careers: Repetitive workflows in finance and operations may require less manual data entry, but exception handling, communication, and business responsibility will still fall to people. Companies may also redirect the labor they save into other roles.
For consumer markets: Greater transparency in corporate spending may squeeze suppliers’ promotions, rebates, and pricing margins. Whether costs ultimately fall will still depend on whether companies pass those savings on to consumers.