Match+
What is Match+?
Section titled “What is Match+?”Match+ adds an AI agent — Mrs. Ledgerwick — to the Unplaced Payments workflow in Quiddly Operations. When a bank-file payment arrives that the deterministic matching engine cannot place, she reads the payment’s information, compares it against your open records, and proposes where it belongs — with her reasoning and a confidence score attached.
She never places anything herself. Every placement decision is made by an operator. Mrs. Ledgerwick is an AI system: her suggestions are AI-generated and can be wrong, so verify before applying. Conversations with her are logged for audit.
How it works
Section titled “How it works”- The matching engine goes first. Payments it can match are handled as of today — Mrs. Ledgerwick marks these Skipped and leaves them alone.
- She analyses what’s left. For each remaining payment, she reads its payment information (references, amounts, sender details) and searches your open records: invoices, part-payment plans and accounts, client accounts, collection errands, enforcements, and business loans.
- She proposes a placement. The proposal shows which record(s) to place the payment on, the amount per record, her evidence (for example, a reference match or an exact amount match), and a confidence score.
- You decide. Review the proposal in the list or open her drawer to challenge it. Apply it, adjust the amounts first, or dismiss it.
Every proposal also passes automated safety checks before you see it: amounts are capped to what each record can actually absorb (mirroring the manual placement window), proposed records must have been found by her own record searches, and suggestions that fail these checks are withheld with the reason shown.
What you see in the Unplaced Payments list
Section titled “What you see in the Unplaced Payments list”A result chip on each payment row:
| Chip | Meaning |
|---|---|
| Confident match | Her top suggestion scored 80% or higher |
| Medium | Scored between 70% and 80% |
| Partial | Part of the balance is placed confidently; a residual is left to decide |
| Low confidence | A proposal exists, but it scores below the threshold |
| No confident match | She analysed the payment and ruled everything out |
| Pending | Not analysed yet |
| Analyzing | She is working on this payment right now |
| Skipped | Already matched by the matching engine |
Clicking a chip opens her drawer for that payment.
The Mrs. Ledgerwick card above the list is both a filter control and a scoreboard. The chips filter the list by result. Expanded, it also shows the queue snapshot, throughput for the last 7 or 30 days (payments analysed, verdicts produced, payments resolved, and how many were placed from her suggestions), and a resolved-per-day trend.
The drawer: review, challenge, apply
Section titled “The drawer: review, challenge, apply”Opening a payment chip starts a conversation. The drawer shows the payment amount, sender, and date, her step-by-step work (“Reading payment details”, “Listing open invoices for debtor”…), and her proposal as a card with the evidence spelled out. From there, you can:
- Ask her anything about the payment — why this invoice, what else it could be, why the confidence isn’t higher. Suggested follow-up questions adapt to each answer.
- Adjust amounts directly on the proposal before applying.
- Apply the placement — the same action, validations, and bookkeeping as placing manually.
- Save the conversation to the payment’s notes.
Memory cards: she learns how your debtors pay
Section titled “Memory cards: she learns how your debtors pay”A memory card is one short fact — “Bergkvist Bygg pays several invoices with one bank transfer; split by oldest reference first.” — scoped to a debtor, a client, or your whole organisation. Mrs. Ledgerwick reads the relevant cards every time she analyses a payment.
Cards come from two sources: operators write them (from the drawer mid-conversation or on the Memory Cards page), and Mrs. Ledgerwick drafts her own from patterns she observes — always visible, editable, and deletable by you. Each card carries a confidence weight and a last-confirmed date, so stale knowledge is easy to spot and refresh.
Data handling and compliance
Section titled “Data handling and compliance”- Sanitised payment fields, debtor metadata, and record-search results are sent to Anthropic via Google Vertex AI to generate suggestions. No raw bank-file payload leaves your tenant.
- Every conversation — her reasoning, your messages, and the accept/reject outcome — is stored verbatim and surfaced in the changelog for audit.
- The product discloses its AI nature in line with the EU AI Act’s transparency requirements: operators are informed they are interacting with an AI system, outputs are marked as AI-generated, and a human makes every placement decision.
- Mrs. Ledgerwick answers questions about placing the payment in front of her. She does not give legal, tax, or compliance advice.
Getting started
Section titled “Getting started”Match+ is an add-on module. To activate it for your organisation, contact Quiddly. Once activated, the Mrs. Ledgerwick card appears on the Unplaced Payments page, and each operator sees a short introduction on first use.
Frequently asked questions
Section titled “Frequently asked questions”Why does a suggestion sometimes place less than the payment amount? Placements are capped at what each record can absorb under your organisation’s placement rules. When a remitter settles an invoice’s full balance, including interest and fees, the capital is placed, and the remainder stays on the payment as an honest residual — exactly as a manual placement would.
Why does a payment say “Skipped”? The deterministic matching engine already matched it. Mrs. Ledgerwick only works on payments the engine could not place.
Can she place payments automatically? No. By design, every placement is applied by an operator.
What does the confidence score mean? It is the model’s own estimate of how likely the placement is correct, based on the evidence it found. Treat it as a starting point, not a verdict — the evidence list shows what it’s based on.