# Demo Talk Track — AR-Reconciliation-Tool

**Setup before the meeting:** `Frontend.accdb` open, sample Excel Input files ready on
Desktop, `generated-design-doc-sample.md` open in a second window/tab.

## Part 1 — The legacy tool (5 min)

1. Show `frmMain`. Say: "This is a small sample we built ourselves — same architecture as
   your 8 tools: Excel in, Access does the logic, Excel out."
2. Click **Import**, browse to `BankStatement_202606.xlsx`, click **Start Import**.
   Point out the status label reporting rows imported / errors.
3. Click **Import** again, browse to `Transactions_HN_202606.xlsx`, click **Start Import**.
   **Import order matters:** the bank statement must go first and the transactions file
   last, so the tool's "last imported batch" pointer ends up on the transactions batch —
   that's what Reconcile/Export use by default. Point out the status label again.
4. Click **Reconcile** on the main menu. Show the result list — point out different
   `MatchStatus` values: exact match, fuzzy match (date-shifted), duplicate, unmatched.
5. Click **Export Report**. Open the generated `.xlsx` — point out the red-highlighted
   variance row.
6. Optional resilience beat: open a sample Excel file in a second Excel window first, then
   try Import on it — show the retry-on-lock behavior instead of a crash.

## Part 2 — What AI can do with this (5-10 min)

1. Say: "Everything you just saw — the VBA code and the Access tables — is exactly the
   kind of input we can feed into AI, instead of a person reading 15K lines of code by
   hand."
2. Show `generated-design-doc-sample.md` — walk through 2-3 sections that correctly
   describe the reconciliation logic (exact match, fuzzy fallback, duplicate handling).
3. If the live TextIQ run was confirmed good in rehearsal (design spec Section 5
   checkpoint): run `textiq-analyze-business` live on `modReconcile.bas` and ask "explain
   the matching logic" — let the audience see it work in real time.
   If not confirmed good: skip live execution, stay with the pre-run doc, and say "we ran
   this ahead of time to keep today's demo focused."
4. Close with the scale argument: "This tool's logic took about X minutes to explain. The
   customer's real assessment of 15K LOC took ~7 person-months by hand. This is the same
   kind of reading, done by AI, and it scales to the full 250K LOC."