Ramp has launched Stack, an AI-native platform for accountants, designed for firms that manage work across multiple client systems.
The platform lets firms deploy software agents for tasks including reconciliations, schedule updates, journal entries and flux analyses. Firms can set the rules that agents follow, decide when work runs automatically and determine when human review is required.
The launch comes as accounting firms face rising client demand and a tighter supply of qualified accountants. Ramp says it works with more than 4,500 accounting firms, including 92 of the top 100 CPA firms.
Stack is intended to replace the patchwork of general AI tools, prompt libraries and spreadsheet exports that many firms use to automate parts of their workflow. Ramp argues that the approach creates auditability gaps and makes it harder to connect work directly to accounting systems and client data sources.
How it works
Firms can add clients to the platform whether or not they already use Ramp. Pre-built integrations connect accounting systems, file repositories, external bank accounts and cards, giving agents access to operational and financial data across different tools.
Ramp says the agents come with accounting concepts already built in, including capitalisation rules, revenue timing and depreciation schedules. Firms then define their own internal processes and client-specific requirements in plain language, setting instructions for how reconciliations, allocations and exception handling should be managed.
The agents confirm instructions, ask follow-up questions when processes are unclear and show the reasoning behind each decision. Repeatable processes can be saved as reusable skills and applied across multiple clients, with firms able to adjust those skills for individual circumstances.
Workflow management is also part of the product. Firms can build or import close checklists for each client and assign tasks to staff members or Stack agents, while tagging approvers and reviewers in the system.
As work progresses, Stack generates workpapers in real time, allowing accountants to review formulas, structure and logic within the platform. When a journal entry is required, the system prepares it and synchronises it with the accounting system, Ramp says.
Benchmark claim
Ramp says Stack was built using leading AI models, but argues that general-purpose models alone are insufficient for accounting work that requires a clear audit trail and high accuracy. The product adds system connectivity, action-level audit records and a model trained to handle a broader range of accounting work, according to the company.
Ramp also cited an independent benchmark of more than 200 accounting tasks built and graded by working accountants. In that test, Stack outperformed every general-purpose model evaluated, the company says.
Client firms using the system are already reporting time savings.
"For me the big thing is consistency. Being able to hold our standard across clients, month after month. When an agent handles the repeatable steps the same way every time, there's just less room for human error," said Calli Hayes, Account Lead, Decimal.
Another user said the tool has reduced the time spent on monthly closing work for some customers.
"It is eliminating or reducing the amount of time we spend on a month-end close by 50% on some clients," said Tyler Otto, President & Owner, Specialised Accounting.
Firm oversight
Ramp says Stack includes an Advisor Console that gives firms a view across their client books, showing which tasks are complete, in progress or waiting for approval. The console also lets firms centrally manage their library of skills and apply them across clients while tailoring workflows as needed.
Ramp positions the product as a way for firms to increase capacity without extending staff hours. It says firms using Stack are closing books faster and taking on clients they would previously have struggled to serve.
Its aim, Ramp says, is to give accountants a system in which processes, controls, and data connections are all in one place, with a record of every action taken by an agent or reviewer.