HHISAB 360 Get HISAB

Reconcile a bank statement in Excel with AI — a 5-minute walkthrough

Guide · ~5 min read

If you've ever spent an afternoon eyeballing a bank statement against a GL extract in Excel — copy-pasting, colour-coding, building yet another INDEX/MATCH column — you already know the shape of the problem. The numbers usually agree, the labels never do. "STRIPE PAYOUT 04/12" on the statement is "Stripe — April settlement" in the books, and a 14-cent FX rounding lives in neither.

This is exactly the kind of work HISAB 360 was built to take off your plate. You don't leave Excel, you don't paste data into a chatbot in the cloud, and the AI has to ask for your nod before it writes anything back into your sheet.

What you need

That's it. No new SaaS, no upload, no separate "rec tool."

The 5-minute walkthrough

1. Drop both datasets into one workbook

Put the bank lines on Sheet: Bank and the GL lines on Sheet: GL. Don't clean them. HISAB is fine with messy column names, mixed date formats, and trailing blank rows — it will tell you what it sees before it does anything.

2. Open the HISAB chat pane and set the mode to Plan

Plan mode means the AI proposes the steps and shows you the exact cell ranges it will read or write, but does nothing until you click Approve. For a first reconciliation, this is the mode you want. See Approvals & safety.

3. Ask in plain English

A prompt that works well:

Reconcile Bank!A:E against GL!A:F on amount and date (±2 days). Match descriptions fuzzily. Write the matched pairs to a new sheet Matched and the unmatched bank/GL lines to Unmatched_Bank and Unmatched_GL. Show me your plan first.

HISAB will respond with a plan: the columns it intends to treat as date, amount, and description on each side; the tolerance it will use; and the three sheets it will create. Read it. If a column is wrong, just say so — "amount on GL is column E, not F" — and it adjusts.

4. Approve, then review

Click Approve. Under the hood, HISAB runs a sandboxed Python step (pandas/openpyxl) to do the matching and writes the results back into the workbook. The whole thing is journaled in the audit log: who ran it, when, which ranges were touched, the before/after values, and the AI turn that triggered it. See Audit log format.

You'll typically end up with three sheets:

5. Iterate on the leftovers

The 80% case is solved. The remaining lines are where judgement actually lives, and this is where chatting beats a static rec tool. Try:

For each row in Unmatched_Bank, suggest the most plausible match in Unmatched_GL and the reason. Don't write anything yet.

HISAB returns a ranked suggestion list inline in chat. You eyeball, you say "yes to 1–3, no to 4," and only then does anything land in the sheet.

Why do it in Excel instead of a SaaS rec tool?

Three reasons you'll feel within a week:

  1. Data doesn't leave your laptop. The workbook stays local. Only the specific context you reference in chat — ranges, column names, sample rows — is sent to your AI provider. (See Security model for exactly what crosses the wire.)
  2. Auditors get a real trail. Every write is journaled with before/after values. You can hand over an audit-ready log without screenshots.
  3. You keep your formulas. Your existing rec template, pivot, or variance analysis still works — HISAB writes into your workbook, it doesn't replace it.

Going further

Once the bank-reconciliation pattern is working, the same shape covers:

If your data is already in Odoo, Xero, QuickBooks Online, Zoho Books, or any OData/REST endpoint, skip the export-CSV step entirely — HISAB can pull the GL extract live and, on the Professional plan, post the reclass journal back when you approve it.

Try it on a real statement

The 14-day trial is the full product, no feature gates. If you're sitting on this month's bank rec right now, that's the file to try it on.

Start free → See pricing