AI in Treasury and Cash Management: How Mittelstand CFOs Steer Liquidity in Real Time

A mid-sized German manufacturer with 180 million euros in annual revenue runs 23 bank accounts across four banks. Every morning, the treasury manager opens five different bank portals, copies balances into an Excel file, manually adjusts for expected payments and receipts, and produces a daily cash position. The process takes 90 minutes. The forecast horizon is four days. Beyond that, visibility drops sharply.

Meanwhile, the company has 4.2 million euros sitting in current accounts earning nothing. The ECB deposit rate is 2.5%. That is 105,000 euros per year in foregone interest - quietly haemorrhaging while the treasury manager is still copying and pasting.

This is not an unusual story. The 2025 EACT Treasury Survey found that cash-flow forecasting is the single top priority for European treasurers, yet only 29% of organisations can confidently model liquidity under multiple stress scenarios³. The tools exist to fix this. The adoption gap in the Mittelstand is still wide.

The Mittelstand Treasury Problem

Three structural forces have raised the cost of treasury underperformance in the German Mittelstand since 2022. Understanding them makes the case for modernisation without needing to appeal to AI hype.

The interest rate shock still has teeth

The 2022-2024 rate hike cycle changed the economics of working capital permanently. Interest expenses for German Mittelstand companies reached their highest level since 2014, according to the KfW Mittelstandspanel 2025¹. When short-term borrowing costs 3-4% and idle cash earns nothing, sloppy treasury is no longer a minor inefficiency - it is a direct margin hit.

• Idle cash cost - Every 1 million euros sitting in a zero-yield current account costs approximately 25,000-40,000 euros per year in opportunity cost at current rates.
• Overdraft exposure - Companies that cannot see their intraday cash position accurately tend to maintain larger precautionary buffers or pay unnecessary overdraft charges. German Mittelstand overdraft spreads typically run 250-450 basis points above ECB.
• Working capital cycle pressure - The J.P. Morgan Working Capital Index 2024 showed DSO and DPO both increased by 1-3 days across European markets²⁴, stretching cash cycles and increasing the value of accurate forecasting.
• Bank credit tightening - KfW research confirms stricter credit policies from banks, making liquidity self-management more critical. Companies that cannot demonstrate cash visibility to their Hausbank face higher financing costs.

The fragmented-systems trap

The Mittelstand treasury problem is not primarily a technology gap - it is a data aggregation problem. Most companies in the 50-500 million euro revenue range operate with:

• Multiple banks - 5-15 banking relationships is typical for a Mittelstand company with European operations. Each bank has its own portal, its own statement format, its own timing for intraday updates.
• Multiple ERP modules or systems - SAP S/4HANA Finance for group accounting, DATEV for subsidiary bookkeeping, Lexware for a smaller entity - all producing accounts payable and receivable data in different formats.
• Manual consolidation - The treasury manager aggregates this data manually into Excel. According to PwC’s 2025 Global Treasury Survey (350 respondents), poor data quality (76%), lack of effective tools (53%), and limited business unit contribution (46%) are the top three obstacles to better forecasting⁴.
• Knowledge concentration risk - As Boomer treasurers retire, the Excel models, formula logic, and banking relationships they maintained are poorly documented. Succession planning is a real risk in companies where one person holds all the treasury knowledge.
• No real-time position - Without automated bank statement aggregation, the daily cash position is always a few hours stale by the time it is ready. For a company with significant daily payment volumes, this is material.

The talent and succession gap

Treasury knowledge in the Mittelstand is highly concentrated. The demographic exit of experienced Boomer treasury professionals - who built and maintained the company’s Excel treasury infrastructure over 20-30 years - creates a structural risk that AI tooling directly addresses. Documented, automated processes that run on structured bank data are more resilient than undocumented spreadsheet models that only one person understands.

What AI in Treasury Actually Does

The word “AI” is used loosely in treasury vendor marketing. Here is a precise breakdown of what current AI technology actually delivers - distinguishing proven capabilities from roadmap promises.

What is proven and available today

• Multi-bank cash consolidation - AI systems connect to bank accounts via EBICS 3.0 or bank APIs, pull intraday statements automatically, and produce a consolidated cash position in real time. This replaces manual portal-hopping with a single dashboard. No more 90-minute morning routine.
• ML-based cash flow forecasting - Machine learning models trained on 12-24 months of historical transaction data produce rolling forecasts at day, week, and month horizons. They learn seasonal patterns, customer payment behaviour, and supplier payment cycles. Research shows a 37.2% reduction in forecast error versus traditional methods⁹. Vendor benchmarks (HighRadius, Kyriba) cite 90-95% accuracy at the 13-week horizon^8,28.
• Anomaly detection in payment runs - AI models trained on historical payment patterns flag deviations: unusual beneficiary accounts, amounts outside normal ranges, new payment destinations, timing anomalies. This is the primary technical defence against BEC and CEO Fraud.
• Automated bank statement reconciliation - AI matches incoming payments against open receivables, flags exceptions for human review, and reduces unmatched items that currently absorb accounting time.
• FX exposure calculation - AI aggregates FX-denominated receivables, payables, and cash positions from ERP data and produces a consolidated net FX exposure by currency. This replaces manual currency-by-currency spreadsheet work.
• Bank fee analysis - AI parses bank fee statements (often in PDF or MT940 format) and benchmarks fees against market rates, identifying overbilling. Industry experience suggests 10-30% of bank fees are overcharged or miscategorised.

What is emerging but not yet fully proven

• Autonomous hedging recommendations - Some platforms (SAP TRM + Joule, Kyriba) generate natural-language FX hedge recommendations, but execution still requires human approval. Full autonomous hedging is not appropriate for most Mittelstand companies.
• Counterparty risk monitoring - Automated monitoring of bank credit ratings and ESMA data for counterparty alerts is available in enterprise TMS platforms but requires data infrastructure not commonly present in mid-market companies.
• Natural language treasury reporting - Joule in SAP S/4HANA and FIS Treasury GPT can generate narrative treasury reports from structured data. The quality is improving rapidly but still requires human review before board distribution.

8 Use Cases With Real Numbers

Each use case below has at least one concrete data point. Where vendor-specific claims are cited, they are labelled as such. Where independent research exists, it is preferred.

1. Daily Liquidity Position Consolidation

This is the highest-frequency, most operational treasury task. For a company with 10-25 bank accounts, manual consolidation takes 60-120 minutes every morning. AI replaces this with automated intraday data pulls via EBICS 3.0 or bank APIs.

• Before AI - Treasury manager opens 5-15 bank portals, copies balances into Excel, adjusts for known cash flows, produces a daily position by mid-morning.
• After AI - Dashboard shows consolidated position across all accounts in real time, updated every 15-30 minutes. Morning routine drops from 90 minutes to a 5-minute review.
• Value driver - Freed time plus more accurate intraday positioning reduces idle cash and overdraft costs. A company with 50M EUR average daily cash and a 1% improvement in overnight placement earns 500,000 EUR per year.
• SAP note - SAP S/4HANA Cash Management with Joule can generate the daily treasury dashboard automatically at a configured time and distribute it to relevant stakeholders, with no manual steps¹⁷.

2. 13-Week Rolling Cash Flow Forecast

The 13-week rolling forecast is the standard liquidity visibility tool for CFOs and heads of treasury. It covers one financial quarter and is used for investment decisions, credit line management, and board reporting. Manual Excel-based 13-week forecasts in the Mittelstand typically achieve 60-75% accuracy at four weeks.

• AI improvement - Machine learning models trained on 12-24 months of historical bank statement data, accounts payable/receivable aging, payroll cycles, and seasonal patterns achieve 90-95% forecast accuracy at 13 weeks^8,9,28.
• Research basis - Tobelem and Reinhart (2024) demonstrated a 37.2% mean reduction in cash flow prediction error using deep learning (LSTM/Transformer networks) versus traditional statistical methods⁹.
• Practical requirement - Models need at least 12 months of clean historical bank statement data plus structured AP/AR data. Most Mittelstand companies already have this; the bottleneck is aggregation, not data existence.
• Time to value - Most teams see a validated 13-week AI forecast running within 60-90 days of connecting bank data²⁷.

3. FX Exposure Detection and Hedging Recommendation

German exporters face persistent FX risk. EUR/USD volatility in 2024-2025 was significant, with the pair moving between 1.02 and 1.12 - a 10% range that directly affects the EUR value of USD-denominated receivables. For a company with 30M EUR in annual USD exports, a 5% unhedged FX move is a 1.5M EUR P&L impact.

• Current practice - Many Mittelstand companies hedge FX exposures quarterly or annually based on a rough review of open invoices. Intra-quarter exposures are often unmonitored.
• AI approach - AI aggregates all FX-denominated positions from ERP (open receivables, payables, purchase orders, confirmed sales orders) and calculates a real-time net exposure by currency pair. It generates recommended hedge amounts and tenors based on exposure profile and company hedging policy.
• Execution - The treasury manager reviews the recommendation and executes the hedge with the Hausbank or via an FX platform. The AI does not execute autonomously.
• Vendor example - Kyriba’s FX module consolidates ERP data from SAP, Oracle, and other sources to produce a natural-language FX exposure summary⁶. Coupa Treasury (Bellin TM5) has similar functionality with particular strength in DACH multi-entity structures.

4. Working Capital Optimisation (DSO/DPO/DIO)

Working capital is the single largest lever on cash generation for most Mittelstand companies. A 5-day improvement in DSO (Days Sales Outstanding) for a company with 100M EUR in annual revenue releases approximately 1.4M EUR in cash.

• AI on DSO - Machine learning models analyse customer payment behaviour and predict which invoices will be paid late, by how many days, and with what probability. Collections teams can prioritise follow-up on high-risk invoices before they become overdue.
• AI on DPO - AI analyses supplier payment terms across all purchase orders and identifies opportunities to take early payment discounts (Skonto) or extend payment terms, with full cash flow impact modelling.
• AI on DIO - Integration with inventory data allows AI to flag working capital tied up in slow-moving stock and model the cash impact of inventory optimisation decisions.
• Benchmark - HighRadius reports that AI-driven collections prioritisation reduces DSO by 3-8 days for mid-market companies⁸.
• E-Rechnung impact - Germany’s mandatory B2B e-invoicing (structured data reception required since January 2025) gives AI working capital tools machine-readable invoice data automatically - improving both the speed and accuracy of AP cash flow forecasts¹³.

5. In-House Banking and Konzern-Cash-Pooling

Mittelstand groups with 3-30 legal entities face a specific challenge: each subsidiary manages its own liquidity, often holding idle cash while the parent or a sister entity is paying overdraft charges. Cash pooling (Konzern-Cash-Pooling) solves this but requires daily or intraday visibility across all entities.

• Manual status - Treasury manager receives morning balance reports from subsidiary CFOs by email or shared Excel. Intercompany sweeps are executed manually, often once per day. Missed sweeps are common.
• AI status - AI monitors all entity accounts in real time, calculates optimal sweep amounts, generates intercompany loan documentation, and either executes or presents the sweep for one-click approval. Timing and amount are optimised daily.
• Vendor strength - Coupa Treasury (Bellin TM5) has particularly strong in-house banking and netting features, reflecting its DACH Mittelstand group heritage⁵. SAP TRM handles this natively for SAP-native groups.
• Value - Reducing idle cash in entity current accounts by 20% for a group with 20 entities and 2M EUR average per entity is a 4M EUR improvement in group cash efficiency, worth 100,000-160,000 EUR per year in interest at current rates.

6. Payment Fraud Detection (BEC and CEO Fraud)

Business Email Compromise (BEC) and CEO Fraud are the dominant payment fraud threats for German Mittelstand companies. Bitkom reported that cyber attacks caused 178.6 billion euros in damage to German companies in 2024, an increase of 30.4 billion euros year-on-year¹⁴. BEC attacks specifically target payment authorisation workflows - the attacker impersonates a CEO or supplier and requests urgent payment to a new account.

• Scale of the problem - The Association of Corporate Treasurers reports CEO fraud targeting at least 400 firms per day globally²⁰. FBI IC3 data shows BEC losses globally reached 2.8 billion USD in 2024 alone²¹.
• How AI detects fraud - Trained on historical payment patterns, AI flags: new beneficiary accounts not seen before, amounts deviating significantly from supplier norms, urgent same-day payment requests outside normal scheduling, and BIC/IBAN combinations that do not match vendor master data.
• Practical deployment - FIS Neural Treasury’s fraud detection module monitors transaction patterns continuously and adapts detection capability over time⁷. SAP S/4HANA has built-in payment fraud detection rules that Joule can explain and escalate.
• Human-in-the-loop requirement - AI fraud flags always require human review before payment blocking. No AI system should autonomously block payments without human confirmation - the risk of false positives (blocking legitimate supplier payments) is too high.

7. Bank Fee Analysis and Reconciliation

Bank fee reconciliation is one of the most consistently underperformed treasury tasks in the Mittelstand. Banks bill for dozens of service types (SEPA credits, standing orders, EBICS connections, account maintenance, SWIFT messages) in formats that are difficult to parse automatically. Many Mittelstand companies conduct a thorough fee review only when renewing their banking mandate - often every 3-5 years.

• The opportunity - Industry experience from treasury consultants suggests 10-30% of bank fees contain errors or charges for services not actually consumed. For a company paying 50,000 EUR per year in bank fees, this is 5,000-15,000 EUR recoverable per review cycle.
• AI approach - Tools like Nomentia and Serrala FS2 parse bank fee statements in MT940, CAMT.086, and PDF formats, categorise each charge, and benchmark it against contracted rates or market benchmarks. Discrepancies are highlighted for dispute.
• ISO 20022 improvement - The switch to ISO 20022 MX formats (mandatory for SWIFT cross-border payments since November 2025) brings richer, more structured fee data, making AI fee reconciliation more accurate¹¹.

8. Overnight Investment and Surplus Cash Optimisation

With ECB rates at 2.5%, holding surplus cash in current accounts is an explicit cost decision. A treasury team that can identify available overnight liquidity accurately and place it in money market funds, repos, or call accounts earns meaningful income on surplus cash.

• The problem - Without accurate same-day cash positioning, treasurers hold precautionary buffers. The average Mittelstand company, in our experience, holds 15-25% more cash in current accounts than is operationally necessary due to forecast uncertainty.
• AI approach - AI provides a same-day net cash position, subtracts a configurable safety buffer, and presents the investable surplus with suggested maturities based on the 13-week forecast. Investment execution remains with the treasury manager.
• Regulatory note - SEPA Instant Payments (fully mandatory since October 2025) means intraday liquidity requirements are higher - funds can arrive and need to leave in under 10 seconds¹⁰. This makes real-time position visibility even more important, not less.

TMS Vendor Landscape 2026: Who Delivers What

The German Mittelstand has a distinct vendor market shaped by EBICS connectivity, SAP integration, German-language support, and DSGVO data residency requirements. Below is an honest assessment of what each major player actually delivers in 2026 - not their marketing claims.

SAP S/4HANA Treasury Management + Joule

• Who it is for - Companies already running SAP S/4HANA Finance. The treasury module is a natural extension with no additional integration cost.
• What Joule delivers in 2026 - A Cash Management Agent that reasons over daily bank statements and automates reconciliations, potentially saving up to 70% of the time finance teams spend on manual cash positioning¹⁷. Joule can auto-generate a daily treasury dashboard at a scheduled time and send it to the CFO. Natural-language FX exposure queries are available in S/4HANA Private Cloud.
• Limitations - Joule’s treasury capabilities are stronger in S/4HANA Cloud than on-premise. Custom Joule Skills (available via Joule Studio since December 2025¹⁸) require developer resource to build. Not appropriate as a standalone treasury tool for companies without SAP.
• DACH fit - Excellent. Native EBICS 3.0, SEPA, and German banking connectivity. Strong HGB/IFRS accounting integration.

Kyriba (CashAI)

• Who it is for - Larger Mittelstand and lower mid-market companies (200M EUR+ revenue) seeking a dedicated cloud TMS with advanced AI.
• What it delivers in 2026 - Kyriba launched its AI cash forecasting suite in April 2024, adding ML-based cash flow forecasting that learns seasonality and payment behaviour patterns⁶. The platform cites 90% forecast accuracy for customers using the AI feature²⁸. Bank Connectivity-as-a-Service supports EBICS and SWIFT connectivity. GenAI is used to update payment formats including ISO 20022 XML.
• Limitations - Kyriba is priced for enterprise. Implementation and licensing for a mid-Mittelstand company is a significant investment. Implementation timelines of 6-12 months are common.
• DACH fit - Good. European data centre options available. EBICS connectivity supported but requires configuration.

FIS Neural Treasury (formerly Quantum)

• Who it is for - Larger corporations and financial institutions needing a full-spectrum TMS covering cash, risk, debt, and investments.
• What it delivers in 2026 - FIS launched the Neural Treasury suite in 2025, incorporating Treasury GPT - the first large language model specifically designed for the treasury industry⁷. Neural Treasury uses AI to predict cash flows, detect fraud, and support liquidity management. The Quantum Cloud Edition (launched mid-2025²⁵) brings cloud-native architecture. FIS won the 2025 Treasury Management International Award for Innovation.
• Limitations - FIS targets large corporates and banks. Implementation cost and complexity is high. Not a natural fit for Mittelstand below 500M EUR revenue unless the company has complex treasury operations.
• DACH fit - Moderate. Strong global bank connectivity but less DACH-specific heritage than SAP or Serrala.

Nomentia (acquired TIPCO in 2021)

• Who it is for - European Mittelstand companies, with particular strength in the DACH, Benelux, and Nordic markets. Following the 2021 acquisition of TIPCO (Vienna, Austria), Nomentia has the strongest DACH mid-market footprint among non-SAP TMS providers.
• What it delivers in 2026 - Modular cash and treasury management: payments, liquidity, risk, and reconciliation modules can be deployed independently. Strong EBICS 3.0 connectivity. The Nomentia 2025.11 product update¹⁵ expanded AI-assisted cash forecasting and improved bank statement matching. Multi-entity cash pooling for Mittelstand groups.
• Limitations - Less AI investment than US-based competitors. The TIPCO heritage product (TIP) is being migrated to the Nomentia platform, creating some transition complexity for existing TIPCO customers.
• DACH fit - Excellent. European data residency, German language support, EBICS 3.0 native, strong Hausbank connectivity.

Agicap

• Who it is for - Smaller Mittelstand companies and growing SMEs (5-100M EUR revenue) that need cash flow visibility without the cost and complexity of a full TMS. Fast-growing in the DACH market - Agicap has over 8,000 clients across Europe following its 45M EUR Series C in 2024¹⁶.
• What it delivers in 2026 - 13-week and rolling cash forecasts, scenario modelling, daily cash flow dashboards with ERP and bank connectivity. AR automation (CashCollect), AP automation, and expense management modules. The platform is positioned as accessible for companies without a dedicated treasury function.
• Limitations - Less sophisticated for complex multi-currency or multi-entity treasury compared to full TMS platforms. Not suitable for companies with significant capital markets activity (bonds, derivatives).
• DACH fit - Good. German UI, DATEV connectivity, EBICS support, and active DACH sales team.

Serrala FS2 (formerly Hanse Orga)

• Who it is for - German Mittelstand companies running SAP, seeking a German-native treasury and payments solution. Serrala (formerly Hanse Orga Group, Hamburg) is one of the few German-owned treasury software providers.
• What it delivers in 2026 - FS2 Cash covers cash positioning, bank statement processing, and electronic bank communications (EBICS/SWIFT). Deep SAP integration - FS2 is built as an SAP-adjacent layer, sharing SAP master data. Strong in accounts payable automation and payment factory. Won TMI award as Best ERP-Based Treasury Solution.
• Limitations - Less ML-based forecasting capability than US-based AI-first vendors. Stronger in payments and reconciliation than in predictive analytics.
• DACH fit - Excellent. German-native, deep EBICS integration, German banking partner ecosystem, headquartered in Hamburg.

Coupa Treasury (formerly Bellin TM5)

• Who it is for - Mittelstand groups and mid-market companies needing strong multi-entity treasury, netting, and in-house banking. Bellin was founded in Ettenheim, Baden-Wurttemberg, and has deep DACH Mittelstand roots.
• What it delivers in 2026 - TM5 covers cash management, payments, netting, FX, debt, investments, and intercompany accounting. Particularly strong for Konzern structures with multiple subsidiaries. The platform integrates with SAP via direct APIs and supports EBICS connectivity.
• Limitations - Coupa’s acquisition focus has been on spend management; treasury investment under Coupa ownership has been more modest than under Bellin’s independent management. Some customers report slower feature velocity post-acquisition.
• DACH fit - Excellent. Bellin heritage, German language, DACH banking connectivity, strong SAP integration.

ION Treasury (Wallstreet Suite, IT2, Reval)

• Who it is for - Larger enterprises and financial institutions with complex capital markets and risk management needs. ION Treasury portfolio spans Wallstreet Suite, Reval, IT2, and Treasura - each targeting different maturity levels.
• What it delivers in 2026 - ION is transforming its platforms into adaptive, agentic ecosystems with domain-specific LLMs for cash forecasting, debt decisions, and liquidity stress scenarios¹⁹. Focus on human-in-the-loop governance and real-time decisioning agents.
• Limitations - The breadth of the ION portfolio (multiple acquired products) creates complexity for customers. Not typically appropriate for Mittelstand below 250M EUR revenue.
• DACH fit - Moderate to good. European operations but less DACH-specific heritage than SAP, Serrala, or Coupa/Bellin.

Regulatory Tailwinds: Three Changes That Help AI Treasury

Three regulatory changes that became effective between 2024 and 2025 create the data infrastructure and payment velocity changes that make AI treasury more valuable and more practical for the Mittelstand.

SEPA Instant Payments - Mandatory Since October 2025

The EU Instant Payment Regulation (EU 2024/886) set a two-stage mandatory deadline¹⁰:

• 9 January 2025 - All EU payment service providers must be able to receive SEPA Instant Payments (transfers processed in under 10 seconds, 24/7/365).
• 9 October 2025 - All EU payment service providers must be able to send SEPA Instant Payments.

SWIFT ISO 20022 Migration - Completed November 2025

As of 22 November 2025, SWIFT retired legacy MT message formats for cross-border payments. All cross-border payments now use ISO 20022 MX format¹¹. For Mittelstand treasury teams:

• Richer data - ISO 20022 MX messages carry structured remittance information, purpose codes, and extended address fields that MT messages did not. AI cash reconciliation tools use this structured data to automatically match incoming payments against open invoices.
• Bank system updates - Your banking partners updated their systems by November 2025. Your TMS or treasury integration layer needs to handle MX format - check with your vendor that this is done if you send or receive SWIFT cross-border payments.
• Corporate BIC note - Companies with a regular corporate BIC (Corp Category) were not in the mandatory migration scope, but banking partners may be sending you MX-format statements. Confirm with your bank.

Germany’s E-Rechnung Mandate - Phased 2025-2028

Germany’s B2B e-invoicing mandate requires structured electronic invoices for all domestic B2B transactions¹³:

• 1 January 2025 - All businesses must be able to receive structured e-invoices (XRechnung, ZUGFeRD). Sending remains optional for most during transition.
• 1 January 2027 - Companies with annual turnover above 800,000 EUR must issue structured e-invoices for domestic B2B transactions.
• 1 January 2028 - All businesses regardless of size must issue e-invoices for domestic B2B transactions.

For treasury, structured e-invoices are machine-readable payment obligations. AI cash flow forecasting tools can automatically extract due dates, amounts, and payment terms from incoming XRechnung or ZUGFeRD files, creating a more accurate accounts-payable forecast than was possible from PDF invoices or paper. This is a genuine infrastructure improvement for 13-week AP forecasting.

EBICS 3.0 - The Foundation of Multi-Bank Connectivity

EBICS 3.0 has been mandatory in Germany since November 2021 and is the standard channel for secure bank-to-corporate data exchange across Germany, Switzerland, and Austria¹². All major German banks (Deutsche Bank, Commerzbank, Helaba, BayernLB, LBBW, DZ Bank, Sparkassen) support EBICS 3.0. It replaced the previous national dialect variants and now provides a single harmonised standard across the GSA region.

• What it enables for AI - EBICS 3.0 is the data feed that every AI cash positioning tool runs on. It provides intraday and end-of-day bank statement data in CAMT.052/053/054 format automatically. Without EBICS connectivity, AI cash positioning cannot function.
• Practical setup - Most Mittelstand companies already have EBICS connections for their ERP payment runs. Extending the same EBICS connection to a treasury tool requires a banking mandate amendment - a few weeks of administrative work, not months of technical complexity.

Implementation: Cost, Timeline, and Where to Start

The practical question for a Mittelstand CFO is not whether AI treasury makes sense - it clearly does at today’s interest rates and fraud risk levels. The question is: what does it cost, how long does it take, and where do you start?

Cost ranges by company size

The 60-day starting path

1. Week 1-2: Baseline audit - Document current daily cash process: how many bank accounts, how many portals, how long it takes, what the forecast horizon and accuracy is. Quantify the cost of idle cash (daily average balance x ECB rate x 0.15 to represent typical inefficiency).
2. Week 3-4: Bank connectivity check - Confirm which banks support EBICS 3.0. Request EBICS access from any bank where the company only has online banking portal access. This is the technical foundation - without it, nothing else is possible.
3. Week 5-6: Tool selection and vendor demos - Based on revenue and existing ERP, narrow to 2-3 vendors. Run demos with live data if possible. Ask specifically: How long to connect our first bank account via EBICS? What does the 13-week forecast require from us in terms of data preparation?
4. Week 7-8: Pilot scope definition - Define one specific pilot: either daily cash positioning across all bank accounts, or 13-week cash flow forecast for one entity. Do not try to solve everything at once. Define the baseline KPI (current time spent, current forecast accuracy) before starting.
5. Week 9-12: Pilot execution - Connect the first bank, pull historical data, generate the first AI forecast. Compare to existing Excel forecast for the first month. Measure the difference. Adjust the model based on what the first month reveals.
6. Week 13-16: Prove and expand - Present the pilot results to the CFO with the measured ROI. Expand to additional bank accounts, entities, or use cases. The value case is now based on real data, not vendor promises.

How Superkind Fits Into Treasury Modernisation

Superkind builds custom AI agents that connect to your existing infrastructure without replacing it. In treasury, this typically means one of three deployment patterns:

Pattern 1: AI cash positioning layer on existing SAP or DATEV

For companies running SAP S/4HANA or DATEV that do not want a full TMS implementation, a Superkind agent connects to your existing EBICS bank connections and ERP data, aggregates the daily position automatically, and produces a 13-week forecast. The agent surfaces insights in a dashboard or pushes a morning summary to the CFO by 8 AM. Implementation is 6-10 weeks.

• Suitable for - Companies with 20-300M EUR revenue, 3-20 bank accounts, and an existing SAP or DATEV infrastructure who want AI cash visibility without a full TMS replacement.
• What the agent does - Daily position consolidation from EBICS data, 13-week rolling forecast, anomaly detection in daily payment runs, idle cash alert when investable surplus exceeds a defined threshold.
• What it does not replace - Your existing SAP or DATEV installation. The agent sits alongside it, reading data and generating outputs, without modifying your ERP configuration.

Pattern 2: In-house banking automation for Mittelstand groups

For groups with 3-15 legal entities, a Superkind agent monitors all entity cash positions in real time, calculates optimal intercompany sweep amounts, generates intercompany loan documentation, and presents sweeps for one-click approval. This eliminates the daily manual email process between the group treasury and subsidiary CFOs.

• Suitable for - Mittelstand groups with multiple entities, existing SAP or Coupa Treasury infrastructure, and a group treasury function that currently spends 2-4 hours per day on intercompany cash management.
• Value driver - Reducing idle cash across entities by 15-20%, plus eliminating 10-15 hours per week of manual treasury work across the group.

Pattern 3: Payment fraud detection layer

A Superkind fraud detection agent analyses all outgoing payment runs before execution, comparing beneficiary accounts, amounts, and timing against historical patterns. Anomalies are flagged with a risk score and require explicit human approval before the payment is released.

• Suitable for - Any company running SAP payment runs or manual bank transfer approval workflows, particularly those without a formal 4-eyes control for all payments.
• Value driver - Prevention of a single successful BEC attack (average cost: 50,000-500,000 EUR) more than pays for 3-5 years of agent operation.

The right choice depends on your company’s current treasury maturity, existing systems, and the complexity of your cash management needs. For most Mittelstand companies between 20M and 200M EUR revenue that are not yet running a dedicated TMS, a focused AI agent on top of existing infrastructure is the fastest path to real-time liquidity visibility with a clear payback within 12 months.

Henri Jung

Co-founder of Superkind, where he helps SMEs and enterprises deploy custom AI agents that actually fit how their teams work. Henri is passionate about closing the gap between what AI can do and the value it creates in real companies. He believes the Mittelstand has everything it needs to lead in AI - it just needs the right approach.