Summary

To realize the revenue and advantage of AI, every enterprise must use a prioritization framework to transform a crowded field of proposed AI use cases into a strategic portfolio of initiatives. With so much hype and so many new possible use cases for AI, leaders are struggling to identify and prioritize the use cases that offer the most immediate and long-term value. Enterprises must emphasize collaboration across business, technical, data, and governance stakeholders as they develop an adaptable strategy for generating measurable outcomes.

AI Use Cases Need Proof of Value

Generative AI (genAI) ruled the day in 2023 and became synonymous with AI in the public eye. In Forrester’s September 2023 Artificial Intelligence Pulse Survey, 90% of AI decision-makers said that their firm is using or experimenting with genAI at the enterprise level; 88% say that their firm is using or experimenting with it at the level of individuals and point solutions. These efforts contributed to a better understanding of genAI’s unique capabilities.

AI has already been incorporated into many enterprises’ existing digital and analytic capabilities in the form of robotic process automation (RPA) and industrial process automation. But genAI offers new capabilities, many of which are analogous to personas, such as creative designer (e.g., media generation), customer representative (e.g., virtual assistant), application developer (e.g., TuringBot), or manager (e.g., business and operational simulator). With such broad potential, the problem that many AI leaders face is how to home in on the use cases that will provide value in both the short and long term. It’s not uncommon for enterprise AI leaders to have to prioritize from hundreds of use cases. AI leaders realize they need a strategy specifically to help navigate the plethora of use cases coming top-down and bottom-up from experiments and proofs of concept.

Forrester defines an AI use case as:

A business scenario or process that AI (whether generative or predictive) optimizes or enables, thus improving a business metric or outcome.

The second half of this definition is where many businesses struggle today. Businesses and stakeholders must focus on the intent of using AI and its outcome to the business rather than the technical capability of a machine learning model. As genAI shifts AI from an analytic engine to a semi-robotic workforce that has a job to perform, enterprises need to account for the intentions of AI use cases, how AI capabilities will support business functions, and how they will measure the results of AI initiatives (see Figure 1).

Thread Together Stakeholder Considerations Across AI Use Cases

AI requires collaboration across many stakeholders in the enterprise to evaluate AI use cases. These different perspectives ensure that functional needs, business unit objectives, industry forces, and regulatory considerations are taken into account as they affect the viability, feasibility, and prioritization of an AI use case. The selection and prioritization process must evaluate outcomes and feasibility through the lenses of business, technical, data, and risk stakeholders to determine whether and how much to invest in a potential AI use case (see Figure 2). There are four key stakeholder groups, each of which plays a vital role:

1. Business stakeholders determine the viability of AI to generate value. Business stakeholders guide outcome goals and measurements, such as impacts on costs, revenues, or profitability, and give a deeper understanding of the business readiness for proposed use cases. Change management will have a significant impact on the risk and feasibility of a potential AI project. Managing enterprisewide change requires proper frameworks to help employees trust the new process or tool. Without these frameworks, initiatives risk low adoption.

2. AI governance stakeholders align use cases with organizational values and compliance. AI governance teams must be cross-functional, with participants from areas such as technology, legal, business, and compliance. Governance teams may be centralized or decentralized, and must triage use cases by applying basic compliance, feasibility, and/or ROI review before they are passed to a steering committee for final decision-making. AI governance stakeholders must also have a clear understanding of the current state of the organization’s data, and they should have a stake in testing the prototypes and final application to ensure that the AI can be trusted by customers, the business, and regulators.

3. Data science stakeholders assess the feasibility of AI to satisfy the use case. Business stakeholders often start with a solution, such as genAI, in mind for a use case but realize after discussion with their data science counterparts that a non-genAI solution, such as a forecasting model, is a better option. Data science teams assist help stakeholders consider the most appropriate model for a use case and apply the right risk controls. They support both the business and the governance teams in understanding the quality of data available for AI projects.

4. Tech stakeholders calculate the ability to engineer, deliver, and manage AI solutions. Application developers determine the technical feasibility, architecture demands, and resource needs of AI projects, just as they do with other applications. In selecting AI use cases, application developers’ challenge is keeping up with the ongoing evolution of AI applications and the associated technologies related to infrastructure, enterprise architecture, and integration and skills such as prompt engineering, vector database management, model and feature selection, model tuning, and model training. Technology stakeholders, just like the governance team, need to maintain ongoing education around the best practices and architectures for building AI and particularly genAI applications.

Source, Select, And Activate AI Use Cases for Business Impact

You need a strategy for identifying use cases, assessing their value, analyzing their feasibility, and determining their priority. Adopting a flexible and adaptable strategy allows teams to move quickly and nimbly to achieve business outcomes with AI technology. Thus, a paradigm shift is emerging: Instead of taking a project approach, enterprises are continuously collecting and prioritizing use cases. Use a six-step process to filter use cases until you have a manageable set of the most impactful AI opportunities (see Figure 3). The six steps of this process are:

1. Source AI use cases across all functions and roles. Everyday enterprise users are invigorated about AI and this organic (i.e., grassroots and non-time-bound) enthusiasm allows for use case sourcing, scaling, and adaptation as new AI capabilities emerge. At the same time, traditional (i.e., periodic and formal) use case sourcing is enhanced as AI users are encouraged to bring genAI ideas that are already tested and ready to scale for the rest of the business. Sportsbet reached out to its data science and marketing teams to see how they were experimenting and using genAI capabilities. This approach accelerated use case sourcing by leveraging both experimentation that was already underway and the fact that business stakeholders had already considered the value of genAI to achieve objectives.

2. Ideate the opportunity, solution, and outcomes. Ideation should focus on how AI is applied to a business objective or problem with known factors of success. When ideating, the litmus test for a good use case is whether the business outcomes can be quantitatively measured. The US Marines are using AI technology to simulate and test concepts for strategies and tactics as they undergo significant modernization and explore AI opportunities. These concept-driven simulations allow them to ideate and optimize defense and operational models that apply AI to new problems and objectives with specific, measurable outcomes.

3. Forecast the business impact of genAI. AI leaders should define the financial model and metrics they intend to use upfront to benchmark the existing state, consider viability before making more investments, and use the parameters of the forecast to guide AI training and tuning. The FAA sought to modernize its air traffic control system with AI capabilities to improve safety with individual pilot flight planning and execution for pilots. The cost and difficulty to address data needs in a distributed, secure, edge, and difficult to access infrastructure required a forecast analysis to estimate safety and operational improvements. With a forecast in hand, the FAA had more confidence to conduct limited pilot testing and then expand to scaled deployment.

4. Prototype the solution and operating model. The prototyping step helps nontechnical stakeholders connect AI to their crucial business problems and needs. You want to be able to test the technology and how people will manage and govern the AI, both for the specific use case and as a way hone the broader operating model around AI. The City of Cincinnati experimented with genAI by using drones to inspect bridges and perform maintenance. Working with Deloitte, Mayor Aftab Pureval commissioned a pilot using drones and an augmented reality headset to understand how the genAI-based solution identified bridge maintenance needs. For the data and technical teams, this prototype provided additional details to scope and improve the capability for full development. For the broader organization, it provides a blueprint for the development, governance, use, and ownership of future use cases.

5. Prioritize and iterate based on ROI, risk, and technical and data feasibility. Prioritization within AI isn’t as simple as creating a linear list. Enterprises need to consider multiple overlapping factors like risk and feasibility. A major North American manufacturer is using TuringBots with genAI to support both software development and testing. It uses a prioritization matrix that balances the risk, data feasibility (availability, quality, etc.), and potential ROI of a use case. The enterprise takes this prioritization a step further by maintaining flexibility around each of those three elements, understanding that the organization’s appetite for risk and data maturity change and evolve as earlier use cases are successfully rolled out.

6. Activate AI by putting use cases in the proper buckets and beginning to build. AI use case prioritization isn’t a matter of simply saying yes or no to each proposal. Because of evolving business needs, ongoing maturation of enterprise data practices, and increasing familiarity with managing risks around AI, enterprises should slot their use cases into high-level activation buckets and determine at what breadth across the business they apply. A comparison of use cases within the recommended activation buckets helps business and technology execs finalize investment and resources. Where consensus or determination of AI use case activation is a challenge, the CEO (or a delegated department head) should be the final decision-maker, as they more frequently lead the adoption and expansion of AI in the enterprise. In Forrester’s Q2 AI Pulse Survey, 2024, 34% of genAI decision-makers indicated that the CEO is primarily responsible for their organization’s AI business strategy.