The Reconciliation Layer

The Reconciliation Layer between your systems of record and the finance team.

Modular RECONs across enterprise Finance. Customized to your data, systems and cases. Every line item, reconciled in real time.

at a glance

The Reconciliation Layer

Flexibly adapts to your systems of record and ingests your data into one Reconciliation engine.

Systems of Record

ERP · SAP · Oracle · D365

Bank · MT940 · CAMT.053 · MT942

EDI · 810 · 820 · 856

PDF · Invoices · Remittance

T&E · Corporate Card · PSP

The Reconciliation Layer Automates

Ingestion and normalization

Matching Engine

Real-time detection and classification

Exception routing

Full audit trail

The Reconciliation Layer Automates

Ingestion and normalization

Matching Engine

Real-time detection and classification

Exception routing

Full audit trail

Finance Teams

Postings ready for ERP

Exception workflows

Real-time dashboards

Auditor evidence on demand

ISO 27001 · EU Hosted · Frankfurt

Zero Training Policy

Demo to Live in 4–6 weeks

The architecture behind the Layer

The integrated Layer solutions establish automated confidence.

your data And Systems

Automated data source ingestion

Every system speaks a different dialect. The Layer ingests all of your data sources, structured or unstructured. ERP, PDF, EDI, bank, expense, TMS.

Native ERP connectors

MT940 · CAMT.053 · MT942

EDI 810 · 820 · 856

T&E · corporate card feeds

PDF OCR + extraction

Flexible

Customized matching

Customized deterministic + AI-authored matching rules at line-item level. Confidence score on every match.

Exact · tolerance · multi-leg

AI-authored rules

Team-approved

Confidence score per match

customizable

Real-time dashboards

Every transaction reconciled on the day it posts. Customizable dashboards reporting in real-time T+0.

Streaming ingestion

Same-day match

Same-day exceptions

Dispute window preserved

Audit ready

Automatic audit trail

Every decision logged immutably. Rule, model, timestamp, user, confidence, rationale.

Immutable provenance

Point-in-time replay

GoBD · IFRS · SOX · BEPS

Regulator-ready exports

The Layer, delivered through RECONs

The RECON library.

Each RECON covers one of your reconciliation use cases end-to-end and strengthens the Layer's coverage without rebuilding anything underneath.

Start with the RECON you need. Build your Layer over time.

Accounts Receivable

Cash Application RECON

Deduction Handling RECON

Short-Pay RECON

Chargeback RECON

Customer Rebate RECON

Statement Matching RECON

Accounts Payable

Three-Way Match RECON

GR/IR Clearing RECON

Vendor Rebate RECON

Duplicate Invoice RECON

Vendor Dispute RECON

Statement Matching RECON

Bank

Bank Statement RECON

Intraday Position RECON

Bank Fee RECON

Payment Execution RECON

Intercompany

IC Matching RECON

IC Elimination RECON

IC Netting RECON

Transfer Pricing RECON

configured per RECon

The Matching Engine.

AI customizes. The matching runs on rails.

The Matching MiX

customized by AI

Deterministic rules

80%

AI-drafted rules

15%

Team review

5%

Auditable at every step

T+0

The Matching Engine. Configured per RECON.

Your rules. Customized by AI.

AI writes code, not data. The Matching Engine customizes deterministic rules to your specific patterns. When it encounters a non-obvious pattern in your data, it drafts a rule. Your team reviews, adjusts, and approves the rules. The output is a customized matching that runs on rails. Transparent, auditable, yours to extend.

Auditable by design

Auditors read rules, not model outputs. Every rule is inspectable and approved by the team. Every override is logged.

Tolerance matching

Configurable per RECON and per counterparty. Cents, percent, days.

Multi-leg matching

One payment against many invoices. Many payments against one. Partial, aggregate, split.

Zero Training Policy

Customer data is never used to train models. Tenant-isolated inference. Contractual in DPA.

AI Framework

AI that finance can audit.

The Reconciliation Layer uses AI where it earns its place: to read data and write customized rules against it. Never as a replacement for deterministic logic, never as a black box. Four principles, applied across every RECON.

01

Principle 1

Human-on-the-Loop, by design

Every AI-drafted rule is reviewable, adjustable, overridable. The team approves the rules. AI accelerates the work, the team owns the decision.

AI writes code, not data.

03

Principle 3

RECONs, not Agents

Scoped reconciliation workflows with deterministic inputs, evidence-traceable outputs, and explicit guardrails. Not autonomous agents pretending to "run finance." That distinction is in the product name.

Named RECONs, not Agents.

02

Principle 2

Zero Training Policy

Your data is never used to train foundation models. Tenant-scoped inference with no cross-tenant learning. Architecturally enforced and contractual in your DPA. Public URL, shareable with your Security team.

Your data trains nothing.

04

Principle 4

Confidence score on every match

Every match carries a confidence score and a rationale. Rule version, model revision, timestamp, user, evidence chain. Confidence stops being a feeling and becomes a metric the team can defend and the CFO can put in front of the audit committee.

Confidence is measurable, not assumed.

01

Principle 1

Human-on-the-Loop, by design

Every AI-drafted rule is reviewable, adjustable, overridable. The team approves the rules. AI accelerates the work, the team owns the decision.

AI writes code, not data.

02

Principle 2

Zero Training Policy

Your data is never used to train foundation models. Tenant-scoped inference with no cross-tenant learning. Architecturally enforced and contractual in your DPA. Public URL, shareable with your Security team.

Your data trains nothing.

03

Principle 3

RECONs, not Agents

Scoped reconciliation workflows with deterministic inputs, evidence-traceable outputs, and explicit guardrails. Not autonomous agents pretending to "run finance." That distinction is in the product name.

Named RECONs, not Agents.

04

Principle 4

Confidence score on every match

Every match carries a confidence score and a rationale. Rule version, model revision, timestamp, user, evidence chain. Confidence stops being a feeling and becomes a metric the team can defend and the CFO can put in front of the audit committee.

Confidence is measurable, not assumed.

Reconciliation is inherently customer-specific

Configured to your data, systems and cases.

No two businesses have the same data sources, the same matching rules, or the same exception patterns. Every RECON is configured for the customer-specific cases. A solution that fits your ecosystem.

Read from your systems of record

ERP, bank, EDI, PDF, TMS, corporate card, T&E feeds. Native connectors, one data model. No middleware, no schema forcing. Fully automated.

SAP

Oracle

D365

MT940

CAMT

pain.002

OCR

API

The engine writes rules against your data

The engine customizes deterministic rules to your specific patterns. When it encounters a non-obvious pattern, it drafts a rule. Your team reviews, adjusts, and approves the rules.

Customized rules

Reason codes

Tolerance bands

Routing

Per RECON

Write back, additive and removable

Postings into your GL and sub-ledgers the way your auditor expects. Reviewed and approved by your team. Read-from / write-to.

Closed-loop integration

File or API

Removable in one flag

Connectors & Integrations

Native to the systems your finance team already runs.

Pre-built, certified integrations across ERP, bank, EDI, PSP, T&E and corporate card systems, plus an open ingestion API for anything custom. No middleware, no parallel operations, no ETL project disguised as a rollout.

MT940 · CAMT.053

Bank · End-of-day

MT942 · CAMT.052

Bank · Intraday

pain.002

Bank · Payment exec

Ingestion API

Custom

MT940 · CAMT.053

Bank · End-of-day

MT942 · CAMT.052

Bank · Intraday

pain.002

Bank · Payment exec

Ingestion API

Custom

Don't see your system? Ingestion API →

Audit Trail

Provenance log - Run 04.23

TIME (UTC)

Event

Amount

09:14:02

RULE

INV-29481 ↔ REF 48210 · exact match · tol €0.00

€48,210

09:14:02

RULE

INV-29482 ↔ REF 12980 · exact match · tol €0.00

€12,980

09:14:02

AI

INV-29483 · multi-pay rule drafted · confidence 0.94

€7,423

09:14:02

RULE

INV-29484 ↔ REF 03810 · exact match

€3,810

09:14:02

AI

INV-29485 · short-pay · reason EL-12 (pricing)

€23,240

09:14:02

CTRL

INV-29485 · deduction approved by A. Meier

€860

09:14:02

MATCH

Batch 04.23.01 · 412 of 418 auto-resolved

98.6%

09:14:02

MATCH

Batch 04.23.01 · 412 of 418 auto-resolved

98.6%

Audit Trail · Architecture

An audit trail generated as a byproduct of the process.

Audit-readiness is a byproduct, not an extra task. Every matching decision is logged immutably — rule, model, timestamp, user, confidence, rationale. Sampling is structurally replaced with 100% line-item coverage. GoBD, IFRS, and SOX stop being a project and start being an operating condition.

Immutable provenance

Append-only log. Auditor, regulator, and forensic accountant see the same record — down to the line item.

Point-in-time replay

Reconstruct any match from any month — without re-running the source systems. Continuous monitoring replaces year-end sampling.

Regulator-ready exports

GoBD · IFRS 10 · HGB §297 · SOX Sec. 404 · BEPS Action 13. Exportable evidence bundles, not screenshots.

Deployment & Engagement Model

Live in 4–6 weeks.

No middleware. No parallel operations. No platform transformation. We start with the RECON that addresses your sharpest pain, prove ROI on your own data, then add more over time. Each new RECON strengthens the Layer without rebuilding anything underneath.

Week 1–2 · Phase 1

Scan

30-day Leakage Scan against a representative sample of your AR, AP, expense, bank, or intercompany data. Output: a board-grade number, source-traceable, on your own production data.

Outcome

Quantified leakage · ROI before signature

Week 3–4 · Phase 2

Pilot

One RECON, one subsidiary, one month of live matching. Integrated with your ERP in read-from / write-to mode. Deterministic rules calibrated to your data; AI-authored rules approved by your team.

Outcome

Live RECON

Week 5–6 · Phase 3

Rollout

Expand the RECON across the perimeter — additional entities, currencies, reason-code taxonomies. Plan the next RECON. Full audit trail from day one.

Outcome

RECON adopted across perimeter · measurable delta · next RECON ready

Validate the business case on your data

Quantify on your data.

Real data of one month of your choice. We configure and implement the first RECON, review findings, measure match-rate, money-at-risk, close acceleration potential, and audit exposure on real transactions.

Donnerstag.ai

English

ISO 27001

EU HOSTED

ZERO TRAINING

©

2026

Donnerstag.ai Technologies GmbH

Donnerstag.ai

English

ISO 27001

EU HOSTED

ZERO TRAINING

©

2026

Donnerstag.ai Technologies GmbH

ISO 27001

EU HOSTED

ZERO TRAINING

©

2026

Donnerstag.ai Technologies GmbH