Traditionally, AI and ML technologies have been used primarily to create forecasts, which in turn can serve as a starting point for planning and budgeting processes. In recent years, the development of generative artificial intelligence (Gen AI), particularly through large language models (LLMs), has accelerated rapidly. These ground-breaking technologies promise a wide range of use cases that go beyond mere forecasting.
However, with technological progress also comes significant pressure to meet expectations. These high expectations, if not met, can lead to considerable disappointment. In this blog post, we explore the potential applications of AI in planning processes. We do this by addressing the question: which applications are already feasible and what future opportunities lie ahead? Let’s take a look at what can already be used effectively and which innovations are still in their infancy.
Forecasts
Forecasting is a classic application area of artificial intelligence, but thanks to modern developments it’s becoming increasingly democratised. Traditionally, the generation of forecasts was the domain of specialists with extensive knowledge of static mathematical models. In practice, however, many companies lack the necessary tools and expertise to use these methods effectively.
By deploying AI, business users can create forecasts independently in order to gain an unbiased view of future developments, thus enabling scarce company resources to be allocated more accurately. AI-supported systems allow the automated selection, configuration and parameterisation of forecast models. This is a crucial step towards greater user autonomy. Business users no longer need in-depth statistical expertise to generate precise predictions. They can create forecasts quickly and easily on their own, with AI-based systems running multiple forecast models simultaneously in the background and selecting the model with the highest accuracy.
For example, with SAP Analytics Cloud users can create a time series forecast with just a few clicks and use it as a template for planning. Users can train multiple models using historical data in a user-friendly interface and select the model with the highest accuracy.
Forecasts based on these statistical-mathematical models offer an unbiased view of future developments, free from wishful thinking and bias. This is crucial for the accurate allocation of scarce company resources and enables decision-makers to develop sound and future-proof strategies. In a world characterised by uncertainty, this approach provides tremendous relief as it combines speed and accuracy in forecasting.
Self-service reporting
The introduction of Gen AI has fundamentally simplified the way business users work with data. These technologies allow complex systems to be operated intuitively and tasks to be carried out independently that previously required the expertise of specialists. In addition to creating forecasts, it also facilitates interaction with the data that serves as the basis for planning.
This allows business users to formulate their queries in natural language and doesn’t require knowledge of database languages like SQL. Gen AI translates the data into precise SQL queries, retrieves the relevant data and presents the results to the user in clear tables and charts.
For example, SAP Analytics Cloud (SAC) offers Just Ask – an integrated functionality for interacting with the data in SAC and in SAP Datasphere. The data models can be indexed and then used for queries in natural language.
Another example is the Auto Analysis AI solution developed by PwC, which combines internal SAP data with retrieval augmented generation (RAG) to enrich the answers with relevant information from the market environment. This improves the quality and accuracy of the generated answers.
This self-service approach strengthens the autonomy of business users and significantly accelerates decision-making processes. By being able to ‘speak’ directly to the data, companies can eliminate the delays that normally occur when involving the IT department. The required information is available directly and without detours.
Summaries of comments
As part of the planning process, planners create numerous comments to justify their assumptions. This commentary is valuable because it provides in-depth insights into the minds of those responsible for planning and contains crucial information and arguments. However, the sheer volume of individual comments is a challenge. Manually reviewing these texts to identify commonalities or key drivers is very time-consuming.
This is where Gen AI comes into play, as it can effortlessly summarise this wealth of information. All comments created during a planning cycle can be extracted and transferred to an LLM. The model distils the essential lines of argumentation and main drivers and summarises them clearly.
With the summaries created by AI, decision-makers can quickly get an overview of the prevailing opinions and arguments without having to read every single comment. This efficient analysis creates scope for a focused discussion about the key assumptions that influence the planning process.
Creation of comments and reports
The possibilities of AI go far beyond simply summarising comments. One impressive capability is the creation of detailed reports and presentations based on extensive data sets. In this process, the AI not only summarises the available information, but also provides its own comments on anomalies and key drivers of developments.
Besides commenting on the company’s own data, which is available to the company at any time, the use of RAG enables the data to be compared with that of industry peers. Thanks to this technology, it’s possible to view the company’s performance in the context of the industry and obtain additional insights.
This approach can also be complemented by automatic outlier detection, which can be achieved through the use of classic machine learning models. These models identify irregularities in the data and transmit their results to the language model (LLM), which comments on these outliers. Another area of application opens up in connection with ML-supported forecasts. In this way, the AI can provide an automatic assessment of planning quality by comparing the plans created by the department with automatically generated extrapolations.
In this context, AI can automate not only the creation of reports but also their distribution to relevant stakeholders. This accelerates the flow of information and ensures that those involved receive the analyses and recommendations they need for their decision-making processes in a timely manner.
In a data-driven environment, AI thus delivers significant added value by transforming large amounts of data into easily understandable reports that provide both detailed insights and recommended actions. For example, the Auto Analysis AI solution developed by PwC generates detailed financial reports based on a uniform data foundation. Another example is SAP’s AI ESG reporting, which generates automated drafts for regular internal or external reports based on the ESG data available in the SAP Sustainability Control Tower.
Strong Operational Performance: Key Ratios Demonstrate Financial Resilience and Strategic Growth in 2024
Combined Ratio
Our combined ratio performance showed positive momentum in 2024, improving to 93.8% from 94.0% in the previous year. While this represents progress in our operational efficiency, we remain 0.7 percentage points above the market average of 93.1%. The improvement was driven by successful implementation of strategic initiatives focused on expense management, despite challenging claims conditions in the natural catastrophe segment.
Claims Ratio
The claims ratio came under pressure in 2024, rising to 67.0% from 64.8% in 2023, placing us 1.2 percentage points above the market average of 65.8%. This increase was primarily driven by elevated natural catastrophe claims, highlighting the need for enhanced risk management strategies. The deviation from the market average underscores the importance of strengthening our catastrophe response capabilities.
Expense Ratio
We achieved substantial improvement in our expense ratio, which decreased to 26.8% from 29.1% in 2023, positioning us favourably 0.6 percentage points below the market average of 27.4%. This significant enhancement reflects the successful implementation of efficiency initiatives and demonstrates our commitment to cost optimisation while maintaining service excellence.
Key Drivers
Premium growth remained robust, with net earned premiums increasing by CHF 36.1 million to CHF 602.1 million. Claims costs rose by CHF 36.3 million to CHF 403.1 million due to natural catastrophes. Notably, we achieved a CHF 3.5 million reduction in net expenses to CHF 161.4 million. The acquisition ratio improved to 18.3% from 19.3%, while the administration expenses ratio decreased to 8.5% from 9.8%, reflecting enhanced operational efficiency across both distribution and administrative functions.
Interactive creation of dashboards
While the creation of simple queries using AI is already being implemented, the next step aims to develop entire dashboards and planning applications in the future using AI. This represents a significant development, particularly at a time when IT departments are under increasing pressure to meet decision-makers’ growing demand for detailed analytics.
The use of AI in this way offers potential relief for IT departments by shifting the majority of the development work to AI. A central aspect of this development is the automated creation of dashboards, which are intended to provide decision-makers with a user-friendly overview and analysis of company data.
With the introduction of Joule, SAP has already presented impressive demos that clearly highlight the potential of such technologies. Yet it remains to be seen how these innovations will prove themselves in practical use. Before we can provide a final assessment of their effectiveness and efficiency, we must wait for their actual implementation and use in practice. The future of AI reporting is promising, but there are still challenges to be overcome.
One of the main prerequisites for automated report generation using AI is a consistent data basis; this, however, is not always present in practice. Incomplete, outdated or incorrect data can lead to inaccurate results. In addition, data often needs to be integrated from different sources, which leads to further complexity.
What’s more, the AI model must be able to respond appropriately to ambiguities in the wording of queries and effectively handle language ambiguity. While manual aids such as synonyms that point to original terms offer some support, they can also create difficulties by leading the model down the wrong path.
Another problem is homonyms, where a term has different meanings. For example, ‘Switzerland’ can be interpreted as a sales region as well as a company code or a geographical location. These ambiguities require robust and intelligent processing and interpretation of the data by the AI model, which poses a significant challenge for the developers of these systems.
Conclusion
The introduction of Gen AI opens up a multitude of possibilities in the planning landscape, especially in the area of self-service. These technologies enable users to operate complex systems intuitively and independently complete tasks that previously could only be solved with special expertise. Applications such as the simplified creation of reports based on projected figures and the efficient summary of comments are particularly promising. These use cases offer significant potential for time and resource savings while promoting informed decision-making.
A glimpse into the future reveals further exciting potential ahead. Initial approaches to the automated creation of planning applications already exist, but these technologies aren’t yet fully developed and applicable in practice. The continuous development and refinement of these systems, however, promises to allow further democratisation and greater efficiency in corporate planning in the near future.
