In our introductory blog in the GenAI series last week, we mentioned six success factors for every AI project. This week we will discuss the first success factor; selecting a GenAI use case. Choosing a successful AI use case is crucial to the success of your project.
You want to create as much impact and value for the organization as possible. But how do you find a good use case for your Generative AI project? And what should you pay attention to? In this second blog from the blog series “Getting started with generative AI” we answer these questions.
This blog is an English translation of the original blog on our website.
Data Discovery Sprint Methodology
At JoinSeven we work with the Data Discovery Sprint methodology. In finding and formulating a suitable use case, the first step “Launch” is crucial.
In this first step we dive into the challenges and investigate the problem in all its facets. We formulate the challenge as a “How can we..” question. Based on this, we brainstorm solutions and choose the one that we consider most promising and impactful. In addition to collecting ideas, in this step we also define the technical principles and necessities that will be crucial later in the project.
We visualize the outcomes of the Launch phase in a value proposition canvas, which gives you an overview of your entire use case for your AI project in one overview.
How might we..?
A classic mistake is that organizations, based on their enthusiasm for a new technology, look for a problem for a solution, instead of the other way around. Questions such as “How can we apply Generative AI to increase customer satisfaction?” at first glance it may seem like a good starting point. However, by focusing on AI in principle, you may limit yourself to other solutions that contribute to increasing customer satisfaction. You run the risk of developing Generative AI that has little connection with your organization and therefore has limited impact.
How might we..
We believe in the power of “How might we..” (HMW) questions. HMW questions open us up to new perspectives and innovative solutions. In principle, you can formulate any problem, issue or bottleneck positively in an HMW question.
..Answer parliamentary questions faster, more accurately and more consistently?
In 2021 we won the “Parliamentary Questions” challenge of the Startup in Residence InterGov program. The original challenge was “Develop a tool that improves and accelerates the answering of parliamentary questions”. We chose to take a step back and first investigate what exactly the problem was. In this way we increasingly found out what the problem was and how policy staff experienced answering parliamentary questions. The following problems and bottlenecks came to light:
- Policy staff experience a very high workload and answering parliamentary questions is often an added burden
- Parliamentary questions are often asked several times, in the same or slightly different context (sometimes to different ministers).
- Parliamentary questions must be answered very accurately and consistently
- To answer a parliamentary question, policy staff spend a lot of time looking for information about the latest state of affairs
Based on this, we reformulated the challenge into “How can we answer parliamentary questions faster, more accurately and more consistently?”. We shaped the rest of the project based on this HMW question.
Crucial parts of every successful AI-project
As we show in the previous section, a successful AI project starts with a broad understanding of current and relevant challenges, objectives and bottlenecks within your organization. The better you connect your project to this, the more impact you can make with your AI project. That is why we recommend that you first sketch a clear picture of what you want to achieve and specifically investigate what role Generative AI can play in this. AI is a powerful technology, but that does not mean that it is the best or most suitable solution for every issue (and vice versa).
A suitable use case contains at least the following components:
- A clearly defined need or problem: Make sure you are clear about what you want to solve, who you are doing it for and why.
- Measurable objective and success criteria: Also make the desired situation, objectives and success criteria concrete and measurable.
- Suitable within the technical possibilities: Make sure you have the preconditions in order. Think about your data, systems and processes.
- Support from stakeholders and those involved: People must want to participate in the project. Without sufficient support, there is a good chance that you will get stuck in the meantime.
Use cases for Generative AI
As soon as the issue is clear, you take the step towards exploring possible solutions. Depending on the perspective of your organization and the end users, you choose a (combination of) specific functionalities. Generative AI can be applied for various purposes (and therefore use cases):
- Generating, modifying and explaining text.
- Generating, adapting and explaining software and code.
- Generating, modifying and explaining speech.
- Generating, adapting and explaining music.
- Generating, modifying and explaining images.
- Generating, modifying and explaining video.
- A combination of the above.
Combine your Generative AI use case with “traditional” data engineering and data science
The possibilities with Generative AI are endless. Especially when you combine it with other forms of data engineering and data science. For example, an AI model can make even better predictions about relevant answers to customer questions if you tag the data about historical conversations (for example using topic modeling). And an AI model that is intended to summarize internal reports can do this best when you give it good examples of summarized reports in the specific domain (or domains) for which the AI needs to do that. A good summary of a tender document may not be the right format for a summary for a marketing strategy, or the business case for a technological exploration.
Think about the role of AI in the process
It also helps to ask the question how far the role of AI extends in the process. Do you want this person to have to take over a process, or support people in these processes? In some situations you mainly want to give Generative AI a supporting function and not let it make decisions autonomously. Consider situations where precision and reliability are crucial, such as in legal, financial or medical decision-making. The judgment of people is very important here and cannot be separated from the AI. The same applies to the role of AI in answering parliamentary questions. We consciously choose to support policy makers with AI by finding and presenting relevant information quickly and efficiently. We consciously leave the formulation of an answer to the policy staff.
Conclusion
In this blog we have taken you through the importance of choosing the right AI use case. We walked you through common pitfalls when choosing a use case and provided an overview of how to find the right use case in your organization.
Some key-points:
- Choosing a suitable AI use case is crucial for the success of your AI project.
- A classic misstep is to look for a problem for a solution, instead of the other way around.
- A successful AI project starts with a broad understanding of challenges, objectives and bottlenecks within your organization.
- The Data Discovery Sprint methodology can help you create good use cases for your AI project.
Outlook
In the following blogs we will take you through all the steps and considerations for a successful implementation of GenAI in your organization. We provide you with the necessary knowledge and tools to set up a successful AI project in your organization. So keep following us and register for the next blogs or the Whitepaper via the form on our website!
In the next weeks, we will publish the following blogs:
- Blog 1: Introduction in Generative AI
- Blog 2: Generatieve AI use case selection
- Blog 3: A solid ground for your AI-project
- Blog 4: Select, train and tune your AI-model
- Blog 5: Ethical considerations in AI-projects
- Blog 6: Validate your AI-solution
- Blog 7: Guide your organization in working with AI
- Blog 8: Recap — Lessons learned from this blog series
