Technology Blogs by SAP
Learn how to extend and personalize SAP applications. Follow the SAP technology blog for insights into SAP BTP, ABAP, SAP Analytics Cloud, SAP HANA, and more.
Showing results for 
Search instead for 
Did you mean: 
By Dr Kim Oosthuizen

2023 has been an eventful year for Artificial Intelligent (AI), especially generative AI and Large Language Models (LLM's). Since January, users have been posting vigorously on social media showing off their experiences with the new generative AI tools. The popularity of the tools is driven by the fact that people can experience AI capabilities for the first time, and it has generated significant attention from news organisations, academics, businesses, and governments.

With so many new generative AI tools hitting the market, all helping to change how we do our daily tasks, let me first explain what generative AI is. Generative AI is a subset that focuses on creating and generating new content. AI-generated content is trained on text, images, audio and other data forms. The data, however, is limited by the quality and quantity of the data used for training, and it is crucial to verify the information the tools provide. There is much hype about generative AI and its possibilities, creating many conversations about AI, ethics concerns and the best way to use the tools. However, other than using the tech to amply and boaster your personal work, AI has many benefits in business.

Business AI is AI applications that exist today, rely on data for learning, are narrow and are measurable. The AI technologies most relevant to business are Machine Learning (ML) (the most applied), Deep learning (CL), computer vision, Edge AI (connected devices) and Natural Language Processing (NLP). ML and DL models ensure the grocery store shelves are fully stocked and help provide insights into sales and customer data. Computer Vision helps detects defects on a production line and helps detect injuries or cancer, aiding in healthcare. Digital assistants (powered by NLP) help us track the status of our orders online and help switch lights in our house. Connected devices help predict maintenance on machines and intelligent checkout in stores and help connect our Garmin or Apple watches to map out our morning run.

Business AI creates the most value when it is scaled across functions. Using business AI can modernise the business environment by streamlining and automating manual tasks, making processes more efficient, predicting future trends by processing and interpreting data across the value chain, providing insights by connecting and making sense of large data volumes and delivering personalised products and services for customers. In a recent report by CSIRO Australia, 28% of respondents said that implemented AI initiatives made an average of AU$ 360k in revenue benefits due to timesaving. However, on average, over 60% to 80% of AI projects fail to reach scale.

To reduce the risk of AI project failure, I want to discuss a five-step approach to finding potential AI initiatives aligned with the business. The process can help ensure your organisation stays focused on the right place to apply AI that can benefit your business.

Here are the five steps for undertaking an AI project.

STEP 1: Understanding the business problem or opportunity.

As a starting point, it is essential to have a relevant business problem that can scale across the organisation. I once worked with a business that’s data team built an AI ML model to predict how many employees would leave. The model accurately predicted 98% of the number of employees leaving. However, HR did not have a process to solve this problem, nor did they use the data from the model. This model then landed up being shelved.

Blindly using AI due to over-eagerness about the technology can make AI projects irrelevant to your business and at risk of failure. When finding AI initiatives, the first step is a clearly defined business problem or opportunity.  Articulating the value of solving a business problem could drive investment in technology. For instance, an inefficient process could lead to employee dissatisfaction due to a manual workaround and solving the this can improve satisfaction.

STEP 2: Suitability for AI

Only some initiatives are suitable for AI. Sometimes redesigning a workflow, simple automation, or a rules engine will suffice.  Therefore, the second step is determining the initiative’s suitability for AI. Doing so could help focus internal resources on the relevant use cases. To resolve this, ask the following questions:

Question 1: Is the problem either manual, repetitive, required complex decision making or needs simplifications?

First, you need to understand if the problem defined in step one is either manual, highly repetitive or needs to be simplified. If so, determine if the problems need to do something, such as providing insights, recognising patterns, classifying, automating a task, or predicting future trends. There are inefficiencies everywhere throughout an organisation, and framing the potential AI initiative can help determine suitability.  For instance, large data volumes of customers make it difficult to recognise customer segments for targeted marketing campaigns. Finding patterns in the data can help optimise marketing campaigns and personalise messages based on customer data.

Question 2: Do you know what data is needed, available, accessible, and of good quality?

AI is all about data. Data is used to learning about the problem, determine feasibility, develop the models and train the model. Any AI initiative will have a prerequisite for data.

The data required for the AI initiative will be directly related to STEP 1’s problem. For example, your problem is that the routing service ticket takes too long because it is manually done today. To do this, you must have a historical ticket and routing data available, stored, and accessible. Manual data, such as service ticket data, is excellent for AI. If the data is unavailable and you have a new problem, then the data must be collected before you can continue.

Side note: Companies should have data management practices for running AI, ensuring data control processes, data quality management, privacy management, and governance practices for the entire AI lifecycle.

Question 3: Can the output be measured?

Determining the initiative's specific outcome, goal, or output and whether it can be measured is crucial. This goes hand in hand with the problem definitions. For instance, if a work process requires a manual workaround, the goal will be to automate the workaround. A manual task that’s now automated can be measured by determining the time saved.

If all the criteria are met, you have a suitable AI initiative to automate, predict, provide insights, or personalise. Continue to the following step.

STEP 3: Rank the AI initiatives based on business value and complexity.

You want to start by ranking the AI initiatives based on maximising business impact while minimising risks. To do so, it is vital to work with an expert to determine the initiative's feasibility and complexity of the AI use case. Experts can be data scientists, AI consultants or technology providers that understand your business.

Once the AI feasibility and complexity have been determined, you should see the business value and score accordingly. If the AI initiative business value is high and the complexity is low (HVLC), plan to do these initiatives first. It can help create excitement about the AI possibilities and enable business users to see the value soon rather than waiting years to change their processes. If the AI initiative business value is high and the complexity is high (HVHC), these could be larger projects requiring strategic relevance, business case approval and proper planning. Ideally, you want to mix quick wins (HVLC) and strategic AI initiatives (HVHC). Any initiatives with low complexity and difficulty articulating the benefit should be parked until value increases. Any low-benefit, high-complexity AI initiatives should be removed.

STEP 4: Determine the AI initiatives and technology fit.

To ensure the AI initiative fits with the technology available. Consider the following:

Option 1: Embedded AI

There could be instances where the AI is already embedded into business processes; however, to do so, it could require an update from current to best practice processes, changing the way of working. For instance, the SAP S/4HANA cloud has multiple processes that are AI-enabled out of the box.

Option 2: Extend the existing process

To keep the application clean without customising the current system, there could be capability available in software currently licensed. Generally, these are quick-win AI initiatives. For instance, SAP BTP offers business-centric AI services to help enable processes such as automatically extracting information from documents and classification, recommendation, and clustering of service tickets.

Option 3: Buy or Build

Leaders must consider whether buying or building the AI initiative is best for strategic AI initiatives. For instance, if the AI initiative is to improve demand planning, purchasing a demand planning solution could be the best approach. Generally, there is not a one size fits all approach, and it will depend heavily on the internal capability of the team and the organisation's AI maturity. Any decision here would need to go through feasibility reviews, prototyping, finding AI specialists and the complexity of the AI initiative.

STEP 5: AI project iterations to scale

The final step is the start of the AI initiative project and goes into project mode until deployment. During this stage, it is essential to focus on the measure success of the AI initiative. However, AI is not only the focus of IT but should also include the entire organisation. Implementing AI into workflows is a change management activity for businesses and requires focus across the whole organisation (that is a discussion for another time). Implementing AI into workflows is not only about the AI but there is also a logical path between people, tasks, structure and technology.

Undertaking any AI initiative requires the entire organisation's support, not just technology. Every organisation should have an AI strategy. The strategy should address a new way of working, how to develop internal capabilities, data governance and responsible AI practices.

In conclusion, AI has taken a massive leap in 2023 with the launch of new generative AI, which has created much hype and conversation about AI. However, beyond these creative tools, AI has significant benefits in business. The Global AI market is projected to reach US$1.5 Trillion by 2030. The opportunities for AI technologies to help modernise the business environment are endless. However, to ensure the success of AI initiatives, businesses should undertake a five-step approach to find potential AI initiatives that align with their business problems or opportunities. By doing so, companies can reduce the risk of AI project failure and create significant revenue benefits.


SAP’s AI strategy focuses on embedding AI functionality across solutions to enable automated, accelerated, sustainable processes. Extending capabilities by offering business-ready AI services through the SAP Business Technology Platform and trusted AI built on ethics and data privacy standards.  Visit SAP AI to learn about our offerings.