Ramp has raised fresh capital while assembling AI functions across several parts of financial operations rather than adding a single assistant to its expense product. The company now combines embedded agents, a shared automation workspace, a payment card for agents and model-routing services, according to TechCrunch’s report on the financing and product expansion.
The funding round brought in $1.85 billion and valued Ramp at $60 billion. It followed a $750 million raise at a $44 billion valuation four months earlier. Ramp also said its annualized revenue had crossed $2 billion and its customer count had reached 85,000, as reported by TechCrunch.
For SME leaders, the relevant development is not the valuation itself. It is the breadth of the operating layer Ramp is trying to create around business spending: capturing reusable employee automations, enabling software agents to make purchases, selecting AI models and monitoring the resulting token expenditure.
The expansion reaches beyond expense management
Ramp began as an expense-management company but has widened its product range into payments, fraud detection, procurement, vendor management and accounting. It has embedded AI agents in procurement, expenses, accounting, budgeting and other products, TechCrunch reports.
That breadth changes the role of AI within the platform. A standalone assistant typically helps a user complete one task or retrieve information. Ramp’s direction connects AI activity to financial processes where a generated recommendation can lead to a procurement action, an accounting workflow or a payment.
For a commercial operations team, this creates a more specific evaluation question than whether an assistant writes a useful response. The team needs to identify where an automation begins, which financial product it touches and whether its output remains a suggestion or becomes an executable action.
Glass turns individual AI work into shared infrastructure
Ramp recently launched Glass, a workspace that collects and remembers the AI automations, prompts and skills created by employees so colleagues can reuse them, according to the TechCrunch account. The new mechanism is organizational memory: useful AI instructions are retained in a shared environment instead of remaining in personal chat histories or private documents.
Consider a hypothetical sales-operations employee who develops an automation for checking a proposed software purchase against budget information. In a personal AI tool, the prompt and working method may stay with that employee. In the model described for Glass, the automation could become available for others to use, reducing the need to recreate the same instructions independently.
The source establishes that Glass stores automations, prompts and skills, but it does not explain how conflicts between versions are resolved, how usage is measured or how a stored skill behaves when an underlying financial process changes. Those details matter because remembering an automation is different from proving that it remains correct.
Conceptual illustration of Glass collecting reusable AI work; not a photograph of the service.
Agent spending meets model routing
Ramp has also launched a corporate credit card designed for AI agents. Separately, it entered model routing, offering existing customers a service tied to AI token-usage monitoring and token-spend management, TechCrunch reports.
These products address two different layers of an agent workflow. Model routing concerns the AI computation used to perform a task, while the agent card concerns a financial transaction that may follow from that task. Token monitoring then makes the cost of model usage part of the same broader spending environment.
A hypothetical procurement agent illustrates the distinction. It could use a routed model to interpret a purchase request, consume AI tokens while processing it and then use an agent-specific card for an approved transaction. This is an explanatory scenario, not a documented Ramp customer deployment. The source does not provide the workflow rules, approval sequence or transaction controls needed to confirm that exact implementation.
Conceptual illustration of agent payments and routed AI usage; not a photograph of a live Ramp workflow.
Funding validates reach, not workflow performance
Ramp says it facilitates more than $275 billion in annualized purchase volume, up 175% from a year earlier. It also reported that annualized revenue had risen from $1.5 billion in June to more than $2 billion, with revenue growth running 25% faster than at the same point last year, according to figures reported by TechCrunch.
Those figures attest to the scale and growth of the company, not to the accuracy or operational value of its AI agents. The report supplies no comparative results for task completion, accounting quality, procurement outcomes or savings produced by Glass, model routing or the agent card.
The immediate decision for an SME is therefore narrow: assess whether consolidating reusable AI workflows, model consumption and agent spending solves a real fragmentation problem in its finance operations. The element to watch is evidence that these connected components improve a defined workflow, rather than merely placing more AI-related spending inside one platform.
