User stories · 19 Aug 2026 · 4 min read

From accounting firm to AI accounting platform: Why we built Meridian

Pilot co-founder Jessica McKellar on the decade-long bet to automate the month-end close, and what changes for accountants when the work starts at review.

Jessica McKellarThe Ledger

The original hypothesis for Pilot, when we started 10 years ago, was: is it possible to automate the month-end close?

This was an exciting problem because, if you can do it, you have a foundational capability that can transform not just CAS but advisory, tax, and other services built on top of the books and derived financials.

This automation is now available to other firms as Meridian.

What it means to automate the month-end close

A month-end close encompasses collecting, importing, and reconciling the right source data, executing a close workflow in accordance with firm-wide and client-specific policies and preferences, rolling forward supporting schedules, producing a reporting package, and validating the results. There is both a lot of repetitive/mechanical work, and a lot of judgement.

Effective bookkeeping automation therefore cannot be a black box that produces a set of numbers and asks the accountant to trust them. The work needs to be decomposed into well-defined activities with checks throughout the process. Each step should produce evidence that it was completed correctly; the system should be able to check its own work; and the final review should tell the accountant what happened, which decisions were made, and what needs attention.

It also needs to be easy to direct the system to make changes.

This is one reason accounting is such a compelling application for AI. A close contains a large amount of repeatable, verifiable work, but it also has endless variation across clients. Modern AI gives us a way to combine flexible reasoning with deterministic processes, controls, and review.

Why we built the system inside an accounting firm

To prove that bookkeeping automation works, we had to use it to close the books for tons of real businesses, month after month.

That is why Pilot has historically operated as both a technology company and an accounting firm: so our R&D teams could work alongside accountants with diverse close portfolios, the relentless deadlines of a monthly close, the crunch of Busy Season, the stress of urgent client requests, and everything else that goes into the daily life of an accountant.

The Iron Man Suit

For years, we described the software we built as an Iron Man suit for our accounting team. The system gave accountants leverage, but the accountant still performed much of the work through the tools.

Advances in large language models in the last 2 years meant that we could put an AI agent inside the Iron Man Suit instead, and have it execute the close from beginning to end: gathering and interpreting the available data, completing the close, checking the work, and presenting the result for review. Given the right source data, the close could move from a set of manual production tasks to a largely autonomous process with an accountant responsible for review and oversight.

The difference is not “the same close, only faster.” It is a different operating model, where the work of the team starts at review.

The Meridian accountant experience

In the traditional model, an accountant often has to spend much of their time on important but fundamentally mechanical production work, to get to the parts that most benefit from their experience, judgement, and relationship with the client.

With Meridian, the starting point is different, because the starting point is the review. The accounts have been imported, reconciled, and categorized. Revenue has been booked and the supporting schedules have been rolled forward. The reporting package has been generated. The system has checked the work and summarized the important decisions and noteworthy changes.

If the accountant has changes or refinements they'd like to make, they simply tell the system to make the changes, and the Meridian agent makes them and reports back with what it did.

The same system can also be delegated complex ad hoc tasks.

As an example: in one onboarding, a business provided a zip file with a bunch of folders with nearly 800 receipts covering a year of expenses paid from several owners' personal accounts. Manually, the job would have meant extracting data into spreadsheets, finding duplicates, converting foreign-currency transactions, and then booking the reimbursements correctly.

I instead just gave the folders to Meridian with the instruction to triage what was in them and book the liabilities appropriately. A few hours later, the system had extracted the receipts, identified a bunch of duplicates, converted the foreign-currency activity, and booked the expenses through liability accounts for the appropriate owners. The books were ready to review.

What Meridian enables for firms

A predictable, autonomous month-end close unlocks many opportunities for firms, including:

  • Higher bookkeeping margins, that can be reinvested in the team and the client experience
  • More predictable capacity needs, which reduces stress on the team
  • Improved CFO/advisory and tax offerings, which benefit from clean, timely books

Explore with AI

See the close run itself

30-minute walkthrough of the platform. See how Meridian runs a month-end close.