10 Steps to Finding your First AI Use Case
Listen to the blog here:
Artificial Intelligence, machine learning and data science seem inevitable to unlock the digital future. Do you want to jump on the bandwagon? Slow down, take a step back and ask yourself: Do you really know where to start? Does your idea really fit your organization? Sneak Peek: Failure is part of the process and concrete business value doesn’t come easy.
According to Gartner, by the end of 2024, 75% of enterprises will shift from piloting to operationalizing Artificial Intelligence (AI), driving a 5 times increase in streaming data and analytics infrastructures. Is your digital transformation journey about to begin? Great! However, getting started can be tricky. The big question is: how do you find a suitable AI use case and how do you know it’s the right one to kick off your digital journey? Our AI experts from Siemens Advanta identified 10 steps to find your first AI use case.
1. Select a real business problem
First off, don’t give in to the temptation of just using that cool technology you saw at a conference or of copying that fancy solution your competitor went for. Go for an ACTUAL challenge of your organization! This will help you build an ACTUAL project as opposed to a hobby: have a real “customer”, get a commitment, search for support, and a pull from the organization. Also, keep in mind that hobby projects are the first thing to be cancelled once operational pressure builds up. Working on a real pain point insulates your project from being stopped prematurely.
2. Define a clear “Why”
Before even looking at the data, there must be a shared understanding of the goal of the AI use case within the team. Merely “analyzing the data” is not going to work! The common understanding of what you want to achieve needs to be directly attached to the business problem: Go back to your organization’s strategy or major pain point and align potential use case outcomes (e.g., a strategic goal of a more bespoke marketing can be achieved by understanding different types of customers based on their purchasing behavior). New ways of ideation and innovation such as design thinking can be the right approach here.
3. Know your stakeholders
The healthcare industry holds one of the most significant investment opportunities for IoT technology ventures. We have observed its growth for years as investors keep leveraging better diagnostic and treatment ideas. The outbreak of the COVID-19 pandemic brought a lot more attention to the sector, and several digital health innovations emerged in 2020 in response to it. These IoT related solutions include applications that help with contact tracing, remote diagnostics and remote patient monitoring. We anticipate that the IoT new trends in this sector in 2020 will still be relevant in 2021, especially with the emerging need for vaccine cold chain management. These innovations will stay with us even beyond 2021 and will keep addressing various other health care challenges – supporting our ability to manage healthcare systems effectively.
4. Specify the data problem
Will we ever have an Artificial General Intelligence – an AI that can solve any problem, just like a human brain? This is highly disputed. Still, nobody contests the fact that today’s AIs perform well on specific, and sometimes very complex, problems. Make sure the selected challenge you want to tackle is defined in a way that is susceptible to AI. It has to be a data problem, such as:
- identifying anomalies
- grouping together entities that have similar features
- predicting outcomes
Formulate your problem as a data problem!
5. Get access to the right data
The data you use must fit the task, it must be sufficient in volume and quality and it needs to be accessible. If the data is not in shape, your use case will not get off the ground. Problems with the data can typically be fixed, but the fixes are all too often neither fast nor cheap. Processes might have to be changed and systems or IT landscapes adapted before you can get started. Plan significant effort to get the data in shape and ask the right questions early!
6. Calculate a business case
If you’re not in a research and development department, you are unlikely to get your AI use case funded or supported beyond a certain stage, unless there is a reasonable business case for the envisioned solution. This can be a daunting exercise. Estimating the potential benefits is already quite hard, determining the costs correctly is typically even more tricky (see our ROI whitepaper). If you manage to specify the benefits of your use case, you enable your whole organization to see the potential of the investment. For instance, increasing process efficiency or reducing manual efforts in your organization will provide you with the buy-in you need.
7. Prepare to fail fast …
Your digitalization journey is a learning journey – not only for you, but for most of the organizations out there. There are chances that your first use case might turn out differently as originally envisioned. Although total failure is a somewhat extreme scenario, partial rather than full success is the likely outcome. Be prepared to adapt! You might not end up with what you initially craved for, but when adapting learnings early on, your partial success might turn out to be exactly what your company needs.
Remember: failure is part of the process and concrete business value doesn’t come easy!
8. … but make something work
Since this is the first step in your digitalization journey, nobody expects a full-blown victory. Nevertheless, flat-out failure will not help you getting support for the next idea. Make sure to learn on the way so that you can either stop early or adjust. This will help you to give positive traction to digitalization efforts in your company and gather promoters to support you.
9. Start small …
Don’t let overwhelming organizational complexity hold you back in kicking-off your use case! Rather select a small and manageable subset, like a small region, a specific business unit, a selection of products or a particular type of customer. This will allow you to prove the value of your solution to the stakeholders while you learn about the important opportunities and threats at the same time.
10. … and be ready to scale
If your use case proves valuable, you don’t want to stay limited to this small subset! Ideally, your case can be swiftly adopted to be scaled to larger regions or more business units. This also means you need to balance efforts to standardize processes, document your initial steps and have productive infrastructure and operations in mind early so they can easily be applied to the expanded scope.
Huge potential to be leveraged with the right partner
For many business stakeholders, the journey of finding the first AI use case still needs to be demystified. Surely, there is no one-size-fits-all solution. The success of an AI project heavily relies on individual characteristics of your organization, its stakeholders and available data. As every journey is unique, it is important to start by answering the right questions early on and manage expectations. The potentials of Artificial Intelligence remain huge. However, often a trusted partner is needed to create transparency on the benefits of machine learning and other technologies for your business. Learn how we can leverage advanced analytics to create value here.