Ramp Stack is a skills-based AI close tool. Agents run the month-end close by following “skills,” which are SOPs the firm writes, tests, and maintains. The main alternative is a pre-built close platform, where the accounting workflows already exist and the firm’s work starts at review.
Meridian is in a category of its own. It is the only tool on this list that ships with the skills and automations to run the month-end close already built, meaning your firm’s work starts at review instead of with building from scratch. That difference shapes everything else in the comparison.
This piece covers what Ramp Stack does, why firms look at other options, and how to tell which model fits your practice.
What Ramp Stack is
Ramp launched Stack in June of 2026. It is Ramp’s first product built for accounting firms rather than for the finance teams that use Ramp cards and expense management.
The core idea is a skill. A skill is a codified SOP: how your firm reconciles a bank account, rolls a prepaid schedule forward, or books a recurring accrual. Agents read the task on a client’s close checklist, apply the matching skill, pull the data, and prepare the work. Stack ships with a starter set of skills for common tasks and expects firms to edit them and build new ones as their client base grows. Skills can be forked per client when a client’s setup calls for something different.
A few facts worth knowing before you compare:
- Stack posts to QuickBooks Online today. Other ledgers are on the roadmap.
- Nothing posts to the ledger without an accountant approving it first. Stack prepares the entry and waits for sign-off.
- Ramp reports a 65.8% success rate on its own benchmark of 200-plus accounting tasks graded by working accountants.
Why firms look for alternatives
The pattern we hear from firms evaluating close automation is that many don’t want to build skills from scratch.
Most firms don’t want to build their own automation. A skills-based tool asks the firm to do the implementation work before the automation delivers anything of value. That means documenting the close process for each client in enough detail that an agent can follow it, then testing the skill, then fixing it when it drifts.
The learning curve lands on the whole team. Skills only help once someone at the firm has learned to write them. Firms that have been through a heavy rollout with another platform describe it as a big lift, and they are cautious about repeating it.
Watching software work is not the same as reviewing completed work. Stack pauses before each posting action and waits for an accountant to approve it. The accountant ends up supervising the process piece by piece instead of doing higher-value work. Human review is a good control, and the better place for it is at the end: the system runs the full close, and the accountant reviews the completed books, the checks it ran, and the exceptions in one sitting.
Many firms have not documented their SOPs yet. A skills-based tool assumes you have a clean written process to codify. Plenty of firms are running a close that lives in a senior accountant’s head. Asking them to write it all down before any automation runs pushes value further out.
The alternatives
Meridian
Meridian is an AI accounting platform that runs the month-end close end to end and hands the finished books to your team for review. The close workflows are already built. Categorization, reconciliation, schedules, accruals, and checklist execution come with the platform, so there is no skills-building period before the first close runs.
Here is how it works in practice. During setup, the firm defines its standard close, chart conventions, cutoffs, materiality rules, and review thresholds. Client-level overrides handle the differences: one client on accrual, another with a different capitalization threshold, a third with departmental allocations. Meridian combines those instructions with the client’s historical books and prior decisions.
The Close then runs each client through four stages: Prepare, In Progress, Review, and Done. Your team enters at Review, with completed books, the checks that ran, a decision summary, exceptions, and the supporting evidence in one place.
If something is ambiguous or a material fact is missing, Meridian flags it rather than guessing. Deterministic tie-outs still have to hold regardless of materiality. Nothing reaches a client without the firm’s sign-off, and firms that want stricter pre-posting controls for sensitive clients can configure them.
Meridian works on top of the client’s existing ledger. QuickBooks Online and Xero are supported in production today, with NetSuite, Sage Intacct, and QuickBooks Desktop in development. The ledger stays the system of record, no migration required.
Onboarding does not require a build phase. Meridian is ready to run when you connect it. You plug in a client’s ledger, set your firm’s policies and any client-specific differences, and the system reads the existing books, chart of accounts, and prior decisions to adapt to how that client is already run. There is no skills library to write first and no migration.
The reason Meridian can ship a pre-built close is that these workflows have run a real firm’s books for more than a decade. Pilot, an accounting firm, built and refined this technology internally before making it available to other firms. The platform runs independently. Firms use it on their own clients, with their own policies and their own reviewers.
Best fit for: Firms that want the month-end close done rather than assigned, that do not want to build or maintain skills, that run QBO, Xero, or both. And for firms that would rather run on workflows proven across a decade of real closes than on tech that launched this year.
Basis
Basis is an AI agent platform aimed at the largest accounting firms. It has raised significant funding and has named customers among the top 200 firms. We cover it in more depth in our Basis alternatives guide.
Basis works by embedding a team with the firm to observe its close processes and build agents around them. That approach produces agents tuned to the firm, and it also means the implementation is the product. Timelines can run for several months, and in some cases longer, before reaching production.
Basis works best at standardized, QBO-primary firms with repeatable client workflows. At bespoke practices, where every client is different, there is less consistent process for the agents to learn from.
Fit for: Top-200 firms with standardized workflows, an appetite for a long implementation, and the internal resources to see it through.
Double
Double, formerly Keeper, is a practice management and client communication platform with newer AI features layered on. Double is described as a month-end checklist and client request tool: flagging uncategorized transactions, requesting W-9s, standardizing the review process. Categorization, reconciliation, and the actual bookkeeping still fall on the team. Our Double alternatives guide goes through the trade-offs in detail.
Double has a strong presence at smaller firms and a polished interface. Satisfaction drops as firm size and client complexity grow, because a checklist tool at 20 clients is a different thing than close automation at 200.
Fit for: Firms that want coordination and client communication rather than execution, especially at a smaller scale.
General-purpose AI (Claude, ChatGPT, Gemini)
Plenty of firms are already doing this. Connect Claude to QuickBooks through an MCP connector, describe the task, and let it draft entries or suggest adjustments. It is flexible and cheap to start.
The costs show up in review. General-purpose models sit outside the ledger, lose context between sessions, and produce probabilistic output. Firms running their close this way tell us the review burden ends up close to what they had before. It also requires someone at the firm to build and maintain the prompts, connectors, and workflows, which is the same skills problem in a different wrapper.
Fit for: Firms with technical staff who want maximum control and accept a high review load. Also a reasonable way to learn what you want from a purpose-built tool.
Side-by-side comparison
| Tool | Approach | What comes pre-built | Ledgers posting today | Review model | Rollout path |
|---|---|---|---|---|---|
| Ramp Stack | Agents run firm-built skills | Starter skills; firm builds and maintains the rest | QuickBooks Online | Approve each post before it lands | Build skills, then run |
| Meridian | Pre-built close runs end to end, team reviews | Full close workflows and accounting primitives | QuickBooks Online and Xero | Review completed books, checks, and exceptions | Subset of clients within the first weeks |
| Basis | Embedded team builds agents per firm | Built during implementation | QBO-primary | Varies by implementation | Months to production |
| Double | Checklist and client communication with AI features | Task templates and request flows | Does not post to ledger, but integrates with QBO, Xero, Sage | Team does the work, tool tracks it | Quick to start as a checklist |
| General-purpose AI | Firm builds prompts and connectors | Nothing accounting-specific | Depends on connector | Team reviews every output | Immediate, ongoing maintenance |
How to evaluate an AI close platform
The firms that end up happy with their choice tend to ask the same three questions.
- Does it do the work, or does it assign the work? A checklist that tells a person what to do is different from a system that reconciles the account, updates the schedule, and shows you the result.
- What is pre-built and what do I have to build? Get a specific answer. If the vendor’s demo spends half its time showing you how to build a skill, that is your implementation plan.
- Where does human review sit? Approving each entry mid-close and reviewing a completed close both keep the accountant in control. They are very different amounts of work.
Which one fits your firm
If you’re a firm that runs mostly on QBO, already uses Ramp expense across most clients, has documented SOPs, and someone on the team who wants to own a skills library, then Ramp Stack may be the best fit for you.
If you don’t want to build custom skills and don’t have time for a long onboarding, Meridian is the best fit. The close workflows are already built, so the setup work is defining your policies and client differences rather than authoring automation. Your team’s job starts at review. This is also the fit for firms that have watched a heavy rollout strain their team once and don’t want to do it twice.
If you’re a top-200 firm with standardized workflows and have a long window to implement, then consider Basis.
If you mainly need a checklist and a better way to chase clients for documents, then Double may be the right fit, especially at a smaller scale.