Fully Autonomous Finance? Is that our Goal? (A Case for Human-in-the-Middle Scenarios)

[KatharinaR]

In the rapidly evolving world of finance, AI is transforming how businesses operate. From automating routine tasks like transaction matching to summarizing and analyzing complex data sets including structured and non-structured information, AI offers remarkable efficiencies.

However, the human element remains crucial in the AI process, ensuring that technology serves its intended purpose effectively and ethically, while meeting specific compliance requirements.

Firstly, human oversight is essential for maintaining ethical standards. AI systems can process vast amounts of data, but they lack the ability to understand the nuances of ethical decision-making. Humans provide the moral compass, ensuring that AI applications align with societal values and regulatory requirements. This oversight helps prevent biases and ensures fairness.

SAP’s approach to Business AI addresses the requirements of ethical responsibility in AI – as well as relevance and reliability in AI. SAP has an AI ethics steering committee and an AI ethics advisory panel with industry leaders to ensure the ethical use of AI in the business context. These teams work with our customers for a more ethical use of AI that aligns with our core values.

Secondly, human expertise adds a layer of critical thinking that AI cannot replicate. While AI excels at identifying patterns and trends, it doesn't possess the intuition and judgment that come from years of experience. Financial professionals can interpret AI-generated insights, considering broader economic contexts and existing relationships with business partners and networks. This leads to more informed decisionmaking considering factors that machines alone might miss. It’s then the combination of the human and AI that delivers more value that leads to better decisionmaking.

Example:  Finance experts provide detailed commentary on financial results. AI can generate a summary that combines the commentary with financial analysis, making a well-explained, comprehensive view of business performance available to business managers that may not be financial experts.

In the world of finance, there could be a legal burden of proof when decisions lead to results that destroy shareholder value. For example, in Germany, this applies to both the management boards of publicly traded stock corporations (AG) as well as limited liability companies (GmbH), regardless of their size. In these cases, company officers might be required to show the data they used as a basis for their decisions[1]. This often includes demonstrating how the risks and opportunities were weighed, and that the decisions that were made can be reasonably considered to be “in the best interest of the organization”. If AI is making decisions autonomously, it might be impossible to re-create the data foundation of the decision. This could leave company officers liable to legal consequences.

Moreover, humans play a vital role in balancing risk and rewards, based on professional experience. AI systems, though powerful, are not infallible. They can make errors or be vulnerable to cyber threats. Human involvement ensures continuous monitoring and quick intervention when anomalies occur, safeguarding financial systems from potential disruptions.

AI at SAP: Human-in-the-Middle

At SAP, as AI is embedded into finance business processes, the processes are designed to keep experts involved. In the early days of AI, machine learning provided recommendations for expert intervention based on previous experience – for example suggesting a write-off of a particilar goods-and-invoice-receipt reconciliation issue. The system proposed the “(most-likely) next step” option for moving the process forward. The accountant would then restart the process with a single click.

With the evolution of generative AI, SAP Business AI can propose the mapping of one chart of accounts to a standard chart, suggest configuration settings for confirmation, and even explain the accounting details behind the evolution of asset values over times.

The current focus of many teams currently are AI agents. AI agents act as co-workers for accountants, assisting in real time. The accountant has the experience of having their own dedicated, super-personalized assistant that knows all your data, speeds up tasks, and enhances decision-making.

AI agents handle the most simple processing tasks and summarize decision options for the accountant. In these cases, the accountant is making the key decisions, and the follow-on actions are triggered via workflows that are performed by a combination of real and virtual agents.

In these examples you can see that AI supports and assists the accountant for more effective outcomes. While AI prepares analyses for decisionmaking, it’s the human that makes the decision.

Example: AI can use your company’s specific accounting policy and your sales forecast data to derive accrual postings related to sales commissions. However, rather than post these accruals directly, the system generates a posting proposal which is confirmed by the local accountant and then approved by the accounting manager. The journal posting is triggered only after this principle of dual control is followed. This reduces compliance risk for the CFO who remains personally responsible for the completeness and accuracy of corporate financial reports in many jurisdications. Read this blog for more details.

In conclusion, while AI revolutionizes finance, the human touch remains indispensable. By integrating human insight and control, businesses can harness AI's full potential, ensuring a future where technology and accountants work hand in hand. Accountants are not going to be replaced by AI. Accountants who use AI effectively will replace accountants who don’t.

Learn more about SAP Business AI in Finance here: https://www.sap.com/products/artificial-intelligence/finance.html

(And thanks to my colleague Daniel Welzbacher for his input!)

[1] Business Judgement Rule (§93(1)(2) Germany’s Stock Corporation Act (AktG)).