CodelessOps · Karim Lameer
AI and automation for finance and FP&A, built by an accountant.
CIMA-qualified · Certified Master Anaplanner · 15 years in finance operations
I automate the work finance teams do by hand: the close, reconciliations, variance commentary, forecasts. I did that work myself for fifteen years, so I know what to automate, what to fix first and what to leave with a person. I build with Claude Code, and I'm packaging the month-end close as skills and plugins a finance team can run.
- Month-end close
- 12 days to 8 hours
- Invoice processing, a month
- 60 hours to 2
- My platform on a public finance AI benchmark
- 81 of 81
Available for contract roles, outside IR35.
Month-end close · April run
- 1 · Bank reconciliation Signed
- 2 · Variance report Signed
- 3 · Management packAwaiting reviewer
- 4 · Audit binderAt close-out
Flagged for a personOne stale cheque on the bank rec. The run stops until someone decides.
01 · What I automate
Month-end close and reconciliations
Bank rec, accruals, and the data loads that hold everything up. Every run leaves a trail a reviewer can sign.
Variance and board commentary
Figures computed by script, commentary drafted from the evidence, and a named person signs it.
Planning and forecasting
Driver-based models, workforce and project planning, actuals that load themselves. Anaplan where it fits, Python where it doesn't.
Invoices and contracts
AI extraction with a person validating every output. Hundreds of contracts per run at Fox, and invoice processing down from 60 hours a month to about 2.
02 · Skills and plugins for finance
The close, packaged as skills.
I'm building my month-end process as skills and plugins for Claude, so a finance team can run it inside the licence they already pay for. Four work today on a demo company. A reviewer gate sits between each step, and the run ends with an audit binder.
- Close checklist
- Bank reconciliation
- Variance report
- Management pack
Next I'm joining them up with Anthropic's finance plugins and packaging them as one install.
03 · What I've built
Built, shipped and written up.
Grounded
A platform for asking questions of finance documents. Every number links to the exact cell or page, and it refuses when the documents can't support an answer. It started as a fork of a course codebase; the verification, cell-level spreadsheet handling, evals, tenancy and audit are mine. It's the most technical thing I've built, and it shows what I can build for you.
The Board Pack Test
A public exam for finance AI: 34 documents, 25 questions, a grader anyone can re-run. Grounded scored 81 of 81. Raw Claude and ChatGPT scored 66 and 71.
Anaplan Estate
A free tool that runs 42 tests on an Anaplan model's exports and shows what depends on what.
Anaplan ArchitectLooking for pilot estates
Inheriting an Anaplan estate should not take three months. A knowledge base from the exports; every answer shows its source.
04 · How I build
Controls first.
I spent fifteen years inside finance teams doing the work by hand first: ledgers, reconciliations, month end, management accounts, forecasts. I was automating it with SQL and VBA long before AI turned up, then with Anaplan and Python.
So what I build comes with approval gates, audit trails, tests, and an eval gate before each release. Where arithmetic beats a language model, I use arithmetic. Everything is documented, and it's yours when I leave.
Where I've worked
- ExscientiaThe full planning platform, and the automated loads that took month-end close from 12 days to under 8 hours.
- FoxAI agents drafting board commentary and variance analysis, and contract extraction with a person validating every output.
- AstellasAnaplan planning builds inside the finance function.
- Goldman SachsAnaplan model builds on contract.
- NHSAnaplan planning work.
05 · Writing
I write down how it works.
Automating the close
Building with AI
06 · Common questions
What finance leaders ask me first.
Doesn't AI get numbers wrong?
It does, and that's the premise of everything I build. Nothing I set up posts a number without showing its working, and a named reviewer signs off before anything reaches the ledger. I published a public benchmark on it: raw frontier models scored 66 and 71 of the 81 machine-checkable points on a realistic board pack; a verified pipeline scored 81 of 81.
What will our auditors say?
They'll ask about AI in your close this year. This setup is the answer: an audit pack every close, each AI-assisted output with its sources and the reviewer's initials.
What happens when you leave?
Your team takes over. The run-book marks which changes are safe to self-serve, and your named admin makes a few of them, live, before I sign off the handover. It's documented well enough that any competent engineer could pick it up.
07 · Hiring?
Finance automation or FP&A systems role?
Tell me what the team does by hand today. I'll tell you what I'd automate first, and what I'd leave alone.
- Contract, outside IR35
- Available now
Featured
RSS FeedCitations aren't verification
What is verifiable AI? The difference between a citation and a verification, why AI arithmetic needs a calculator, and the four controls a finance team should demand.
The Board Pack Test: I gave four AIs the same board pack, then graded them with a script
I published an exam for finance AI: 34 realistic documents, 25 questions, a deterministic grader, answer key included. The frontier models got most answers right. Proving them right was a different story.
The 4:45-on-a-Friday question
Every finance team knows the out-of-nowhere question that lands at the worst moment. It isn't an interruption. It's a tax. Here's what it actually costs, and why you'll never get headcount to fix it.
Recent Posts
Continuous close for FP&A
Updated:A three-part series: what a continuous (zero-day) close is and why to aim at one, one FP&A month worked old way and new, and how to build it with mainstream AI tools.
What a continuous close is, and why to aim at zero days
Zero-day close as a direction, what a continuous close actually is, who is talking about it, why the forecast is the point, and why a finance team can now build one: writing the code got cheap, owning it did not.
The use case: a contract signed on the 6th, on the forecast by the 8th
One FP&A month, old way and new: a contract that reaches the forecast in week six or on the morning of the 8th, which numbers can close early, which must wait, and the chain from purchase order to accrual that decides it.
How to build it: small processes, one log, approvals where people already are
The architecture of a continuous close: small processes orchestrated centrally, one append-only log, people at the gates, approvals in Slack or email, and a console that is only a view over the log.