EP555: Jessica McKellar - Spotlight: Iron Man Suit For Accountants: Automating The Month-End Close With AI Tools

See what the team at The Successful Bookkeeper has on right now →

Jessica McKellar has spent ten years proving that bookkeeping can be automated to a high degree, first by building an accounting firm as a test lab and now by packaging that technology for other firms as Meridian. In this Spotlight episode she walks Louie Prosperi through what an autonomous month-end close looks like, where AI belongs in a firm's workflows, and what it does to review time, staff morale, and security.

Chapters

• [01:30] Jessica's background and Pilot's origin
• [04:00] The Iron Man suit and LLMs
• [07:00] How Meridian works with your ledger
• [11:40] What firms should delegate to AI
• [14:10] Hallucinations and deterministic structure
• [18:20] Cleanup work and messy onboarding
• [22:10] Broken feeds and sync monitoring
• [25:00] Reviewing work and firm visibility
• [28:00] Staff culture and $1M portfolios
• [31:00] Data security and privacy

From test lab to Meridian

Jessica and her co-founder began by writing software to help with their own month-end close, then spent a decade building Pilot as a place to stress-test automation against nearly 4,000 client businesses a month. Her shorthand for the first version is "the Iron Man suit for an accountant", with a human operating it. About 18 months ago, newer LLMs let them take the accountant out of the suit and plug in AI for an end-to-end close. Her motive for releasing it to other firms is simple: "I can't reach every small business owner myself."

What Meridian actually does

Meridian has four pieces: recurring work written as a checklist and executed in software, delegation of ad hoc tasks, an optional client portal, and team and firm management. It sits on top of your existing QuickBooks or Xero with no migration, posts directly to the GL, and inherits and rolls forward your supporting schedules. You can run the default close checklist out of the box, or customize a firm-level checklist down to the task level per client.

Keeping AI honest

Jessica would not "directly plug Claude into a client's GL and say have at it on the month-end close". That is too open-ended, with too few checks and too little visibility. In Meridian, work like reconciliations and schedule roll-forwards is deterministic software doing the math, and AI isn't asked to invent an answer. She advises firms building their own setup to decide what must happen the same way every time and what can be left to a model's judgment. Structure also surfaces the issues worth a client conversation, such as a payroll liability that isn't clearing or an invoice payment posted with no record behind it.

Cleanup work and broken feeds

Jessica describes onboarding a client with no business bank accounts, 800 receipts in multiple currencies, and incorrect Xero financials. She delegated the catch-up books as an ad hoc task. Meridian flagged duplicates, asked how to handle currency conversion, and called out receipts with no amounts. It finished in about two hours while she ate dinner with her kids. On data feeds, Meridian supports native GL feeds, Plaid-style third-party syncs, and statement uploads, and it can reveal a broken sync through balance sheet checks, like cash going out with no offsetting liability.

Review, capacity, and the team experience

If the operational work is done by software, the human time shifts to review, and clean books make for a faster review. Meridian leaves notes on why items aren't clearing and records client preferences in the checklist, which Louie, a reviewer who always asks staff to "show your work", appreciated. Pilot's internal target is $1 million in client revenue per production accountant. Jessica says the team feedback is "oh my gosh, this is so much better", and that after eight years of some clients always getting the short straw on deadlines, late closes no longer happen. Less context walks out the door when staff change, and PTO and turnover are easier to absorb.

Security and privacy

Meridian works on a principle of least access, and you can use it without storing data in it at all. Client data is not fed back into foundation models as training data. Jessica's larger point is that most security risk comes from a firm's own habits, such as password storage, credential rotation when employees leave, and who can download personal information, and less from the close automation tool itself.

Links mentioned

• Meridian: meridian.pilot.com
• Pilot: pilot.com
• The Successful Bookkeeper: thesuccessfulbookkeeper.com

About the guest

Jessica McKellar is the co-founder and CEO of Pilot, where she helps shape how AI and technology are changing bookkeeping and accounting. She has a background in technology and engineering and has worked with her co-founder for nearly 20 years, since they met in the computer club at school. She still maintains a $500,000 client portfolio herself to stay close to the technology.

About the host

Michael Palmer

Michael Palmer is the host of The Successful Bookkeeper podcast — almost 600 episodes and counting — and founder of The Successful Bookkeeper Global. He's spent over 15 years in the bookkeeping industry, working one-on-one with thousands of bookkeepers across the globe. He started this work because of his father — a brilliant electrical contractor who worked twice as hard as he should have had to, because nobody on the financial side was in his corner. That gap is what The Successful Bookkeeper exists to close. His view: bookkeepers are the most undervalued force in small business — and every bookkeeper who builds a real business changes two families: theirs, and their clients'.