Five elements of an AI business case
More and more business executives are facing their first AI business case. Executives such as CHROs (Chief Human Resource Officers), CFOs, CMOs, or CSCOs (Chief Supply Chain Officers). They need to create an AI business case with limited experience, if any, within their organization. The IT organization will likely handle the business case’s complex technical aspects, whereas the business executive is left with detailing the AI solution’s business impact.
There are five critical elements the business executive needs to include in a business case for an AI solution. They are 1) the business problem solved by AI, 2) the details of the AI solution, 3) Scalability & maintenance, 4) AI Financials, and 5) the stakeholders. AI implementations fail if the responsible business executive does not detail these five areas.
However, before looking at these five elements, it is essential to understand that an AI project’s business case will deviate from classical business cases in four distinct areas. These four areas are:
- AI is a “discovery-driven” investment, resulting in a business case with no clearly defined KPIs. The reason is that there is a limited history of use cases, knowledge, and data about AI implementations in most organizations. As a result, business executives need to address two challenges. First, the executive responsible for the AI project and the CFO or finance organization must understand that this is not a traditional business case. The risk profile of an AI business case is different. If the AI project risks are not made explicit, the project may get defunded. Secondly, the business case’s key executives will need to determine the business case metrics and track them. These KPIs will likely be less standard KPIs or less robust in the initial business case.
- AI projects are cross-functional projects. They often involve more business functions in addition to the IT function than the average IT project. Therefore, project and business case governance is more complex, the budget impact is broader, and the need for collaboration higher. When creating an AI business case, ensure that the project’s governance is detailed, the effect on budgets across the organization is agreed upon, and the implementation plan includes dealing with the need for communication across the organization to enable frictionless collaboration.
- Ensure that the data environment is ready for implementation. AI projects far too often stall or even fail because the data is not available or not usable as-is. Information is collected for a purpose and in a format that supports a specific analysis. As a result, the data may not support new forms of analysis intended for the AI solution proposed. Every AI business case needs to highlight the importance of pre-existing quality and relevant data or include a budget to ensure that such a dataset is created as part of the business case.
- The AI business case needs to include provisions for the appropriate skills and leadership knowledge to execute the project and subsequently manage the environment when in production. This may result in the need to hire more data science or developer expertise and include costs for developing organizational and leadership competencies.
Once an organization has internalized these four differences in AI business cases, they can work on the details required in the five elements of an AI business case.
Five elements of an AI business case
There is extensive coverage of AI business case development from a technical perspective, including the challenges around the data environment, what a business case needs to consider concerning the overall system architecture, and critical security, risk, and governance. But there is only limited coverage of the business elements for the AI business case. Below are five factors to consider as part of every business case for an AI implementation.
The business problem
The first but obvious question is what business problem or opportunity are you solving? While this seems a trivial question, far too often, the actual business problem is not clearly articulated. As a consequence, it becomes harder to measure the impact or result of the project. This statement applies to any project, but because most organizations are still learning about AI’s use and effects, it is even more critical for AI projects. A clear articulation of the business problem to be solved and how to measure the project’s impact is the first step to get approval for an AI project, especially if the finance organization is not experienced in “discovery-driven” investments.
Spending time on detailing the actual AI solution is the second area of focus. This is critical for several reasons. These include the fact that most projects are new to the organization, so what the solution is, how it will work, at scale, and who will be involved in using the solution will be new to the organization. Ensure that the solution fits as part of the organization’s overall strategy and is linked to the key leaders’ objectives and their organizations implementing the solution.
Scalability & maintenance
Most AI projects are initiated as a smaller POC project with the intent of scaling the solution rapidly. The business case must outline how to scale the solution, technically, business-wise, and on the required skills. Lastly, the necessary training and development needed within the organization implementing the solution. Furthermore, maintenance of the full-scale solution is more complicated with AI projects since the algorithms evolve as well the regulatory and legal environments around AI. Lastly, be realistic with the timelines for the scalability of the AI project. An initial review of projects over the last three years shows that most organizations are overly optimistic about implementation and scalability timelines.
The AI business will need to be approved by the finance organization, as is the case for any project. Given that most AI business cases lack well-proven and standard data and metrics used in traditional business cases, the finance organization will often challenge the business case because it is a “discovery-driven” investment. Consider whether simulating data where non exists could be a possible option. Approaching the AI business case as a venture capitalist is a wise approach, thus assuming less qualified KPIs, a clear, quantifiable identification of risks, and an iterative approach to proving the business value as you scale the solution. Lastly, the business case’s financial part should also define possible failure criteria, so these are clear from the outset.
AI projects involve more stakeholders than traditional IT projects. In addition to several entities within the IT organization, there will be several business stakeholders. The critical stakeholder will be the organization implementing the solution, but the HR organization and the legal organization will likely need to be involved. Thus, ensuring a broader stakeholder communication plan and continuous reporting is a critical part of the AI business case.
There will be other factors to include in an AI business case, but the five above are the most critical to have in the business case any organization builds. Also, understand that all business case documents should be treated as live documents, especially AI business cases.
CEOs and senior business executives need to appreciate that AI business cases will differ from traditional business cases in several ways. This is one of several critical areas that need to be addressed if AI implementations are to scale across all organizations’ sizes.
About the author
Chairman of the Board, 2021.AI
Peter Sondergaard is currently Chairman of the Board at 2021.AI and Owner of his Executive Advisory company, the Sondergaard Group. Before this, Peter worked as the Executive VP and member of Gartner’s operating committee for 15 years. Peter is a well known and sought out speaker covering many topics within IT, AI & ML.
You might also like…
At some stage over the next ten years, the average leader will be responsible for a business area where they are responsible for several employees, a dynamic evolving data platform,...
MELDS is an abbreviation for the five components: Mind, Experimentation, Leadership, Data, and Skills. These five components are invented for management purposes and should be considered when embarking on an...
The Ethical AI Newsletter
It’s not fake. It’s not artificial. It’s real news! Sign up for our Ethical AI newsletter and get the latest AI insights from our data science and AI experts.