The traditional going concern assessment reaches its limits in volatile markets. Learn how machine learning is revolutionising cash flow projections, scenario analyses and early warning systems -- and which legal requirements under IDW S11, § 252 HGB and § 19 InsO apply.
Going Concern 2.0: How AI Calculates Viability
A medium-sized mechanical engineering company with 200 employees loses two major customers within three months. Management must prepare a going concern assessment under time pressure -- yet the traditional methodology with spreadsheets and linear trend extrapolations barely captures the dynamics of the crisis. This is precisely where a new approach comes in: Artificial Intelligence as a tool for calculating corporate viability.
The Legal Framework: When Is a Going Concern Assessment Required?
§ 252 HGB -- The Going Concern Principle
The German Commercial Code enshrines the going concern principle in § 252 (1) No. 2 HGB: valuation must be based on the assumption of continued business operations, unless actual or legal circumstances indicate otherwise. This principle forms the foundation of all commercial law accounting and substantially determines the valuation of assets and liabilities. If the going concern assumption ceases to apply, assets must be recognised at liquidation values -- often with dramatic effects on the balance sheet.
§ 19 InsO -- Over-indebtedness and Continuation Prognosis
Under § 19 InsO, over-indebtedness exists when the debtor's assets no longer cover existing liabilities, unless the continuation of the business over the next twelve months is more likely than not given the circumstances. The positive continuation prognosis is thus the decisive lifeline: if achieved, computational over-indebtedness becomes irrelevant. If it fails, there is an obligation to file for insolvency within six weeks pursuant to § 15a InsO.
IDW S11 -- The Audit Standard
The IDW S11 (2024 revision) of the Institute of Public Auditors in Germany defines the requirements for assessing insolvency opening grounds. The standard requires an integrated financial plan comprising income, balance sheet and liquidity planning for a period of at least twelve months. The prognosis must be based on traceable assumptions and demonstrate the company's solvency throughout the forecast period.
The Limits of Traditional Going Concern Assessments
Conventional preparation of going concern assessments has significant weaknesses that regularly lead to sub-optimal results in practice:
Cognitive Biases
Humans are prone to optimism bias: directors and advisors systematically overestimate the situation compared to reality. Studies show that management forecasts in crisis situations overestimate actual developments by an average of 20 to 30 per cent. Anchoring effects compound this -- existing plans are extrapolated rather than fundamentally questioned.
Linear Extrapolation in Non-linear Markets
Classical models extrapolate historical trends, implicitly assuming linear relationships. In a world where supply chains can collapse within weeks, energy prices triple or entire business models become obsolete through technological shifts, this methodology is inadequate. Industry-specific disruption risks are systematically underestimated.
Limited Data Processing
A manual financial plan can realistically calculate five to ten scenarios. The actual range of possible developments, however, is orders of magnitude more complex: every combination of revenue trajectories, cost structures, customer payment behaviour, interest rate developments and supply chain risks generates its own scenario.
How AI Is Revolutionising Going Concern Assessments
Machine Learning for Cash Flow Projections
Modern ML models -- particularly gradient boosting methods (XGBoost, LightGBM) and recurrent neural networks (LSTM) -- can identify patterns in historical financial data that remain hidden to human analysts. According to a study by TU Graz, AI models achieve a hit rate of over 85 per cent in insolvency prediction -- significantly higher than classical statistical methods. The models process not only financial ratios but also qualitative factors such as industry affiliation, company age and management changes.
Real-time Scenario Analysis
Instead of five manual scenarios, AI calculates thousands of Monte Carlo simulations in minutes. Each simulation combines different assumptions about revenue, costs, payment defaults and market developments and generates a probability distribution for future liquidity. The result is not a single number but a confidence interval: continuation is secured with a certain probability -- or it is not.
Integration of External Data Sources
A decisive advantage of AI-supported forecasts is the ability to incorporate external signals into the analysis. These include:
- Market data: Industry indices, commodity prices, exchange rates and interest rate developments
- Supply chain data: Supplier creditworthiness, transport costs, inventories in the value chain
- ESG factors: Regulatory risks, carbon pricing, reputational risks
- Media and sentiment analyses: News flow on customers, competitors and industries
- Macroeconomic indicators: GDP forecasts, labour market data, credit default rates
As Rödl & Partner describes under the heading Predictive Insolvency, data-driven models enable early crisis detection with a lead time of six to twelve months before actual insolvency materialises.
The International Standard: ISA 570 (Revised 2024)
The international audit standard ISA 570 (Revised 2024) by IAASB places heightened demands on the audit of financial statements with respect to the going concern assumption. Key changes include:
- More robust risk assessment: Auditors must evaluate more promptly and thoroughly whether events or conditions exist that cast significant doubt on the entity's ability to continue as a going concern
- Extended assessment period: The assessment period now extends to at least twelve months from the date of approval of the financial statements
- Enhanced transparency: Strengthened reporting obligations towards governance bodies and the public
This tightening underscores that the quality of going concern assessments is gaining importance internationally -- and that technological support is not merely desirable but increasingly necessary.
Practical Pilot Applications: Where AI Is Already Being Used
Early Crisis Detection at Credit Institutions
Banks use AI-supported early warning systems to assess the default risk of their borrowers. The models continuously analyse account movements, payment behaviour and external market data, generating risk scores that enable proactive credit management. As documented by Finbridge, ML models significantly outperform classical scoring methods.
Due Diligence Support
In restructuring situations, AI accelerates the analysis of financial data. Automated systems can evaluate annual reports, management accounts and cash flow statements in minutes, identify inconsistencies and flag risk indicators that would be overlooked in manual reviews.
Ongoing Monitoring in Self-administration
Within self-administration proceedings under §§ 270 ff. InsO, AI systems can take over ongoing liquidity monitoring. They compare actual values in real time against the approved financial plan and automatically alert management and the custodian when deviations occur.
Limitations and Risks of AI Deployment
Data Quality as the Achilles Heel
AI models are only as good as their training data. In practice, the financial data of companies at risk of insolvency is often incomplete, inconsistent or outdated. A going concern assessment based on flawed data inevitably produces flawed results -- the garbage-in-garbage-out problem applies without reservation.
Explainability and Traceability
IDW S11 demands traceable assumptions. Black-box models whose decision logic is not transparent do not meet this requirement. For use in insolvency law practice, Explainable AI approaches (XAI) are therefore required that can justify their forecasts -- for instance through SHAP values or LIME explanations.
Human Responsibility Remains
As the research overview by Nordantech emphasises: AI can provide indications, calculate scenarios and quantify risks -- but the decision on restructuring, reorganisation or insolvency filing remains with humans. The going concern assessment is not a purely mathematical exercise; it requires entrepreneurial judgement and legal evaluation.
Recommendations for Practice
Build data infrastructure: Ensure your financial data is current, complete and structured. Invest in ERP systems and automated interfaces.
Understand AI as a supplement, not a substitute: Use ML models to support your financial planning, but do not rely blindly on algorithmic forecasts. Human plausibility checks remain indispensable.
Systematically expand scenarios: Supplement your traditional best-case/worst-case planning with Monte Carlo simulations and stress tests. Identify the assumptions that have the greatest impact on going concern viability.
Incorporate external data: Integrate market, industry and supply chain data into your forecast models. Isolated analyses of your own financial data are insufficient in volatile markets.
Ensure explainability: Ensure that AI-supported forecasts can be presented in a traceable manner to creditors, courts and supervisory authorities.
Monitor regulatory developments: Follow the implementation of ISA 570 (Revised 2024) and the further development of IDW S11. The requirements for going concern assessments will continue to increase.
Conclusion: The Going Concern Assessment of the Future Is Hybrid
The going concern assessment is at a paradigm shift. Artificial intelligence does not replace the experienced restructuring advisor or auditor -- but it considerably extends their toolbox. The combination of legal expertise, entrepreneurial judgement and data-driven analysis defines the going concern assessment 2.0.
Companies in crisis that adopt AI-supported forecast models early gain valuable time -- which in restructuring can decide between success and failure. At the same time, regulatory requirements for the quality and traceability of forecasts are increasing. Those who meet these requirements with outdated methods risk not only inadequate results but also personal liability.
At compleneo, we support you in preparing going concern assessments in accordance with IDW S11 and in integrating modern analytical methods into your early crisis detection. Get in touch with us.
