Unlocking the Future of AI: A Creative Framework for AI-Powered Services

Unlocking the potential of AI for service innovation can feel overwhelming. But what if you had a clear framework to guide you? In this article I am excited to share an easy-to-understand model, highlighting nine categories of AI services that span across prediction, adaptivity, and agency. This framework isn’t just a theoretical exercise, it’s a practical tool that helps designers and innovators quickly generate and frame valuable AI-powered services.

Introduction

Artificial Intelligence (AI) has the potential of changing the world. From predictive analytics tools that help us make better decisions to intelligent agents that operate as butlers, AI is increasingly intertwined with our daily lives. But as we move into this brave new world, we need a framework to understand, categorize, and innovate within the space of AI-powered services. This post proposes such a framework, exploring the various use-cases, capabilities, and potential future directions of AI. It’s work in progress, so any feedback is welcome.

Let’s start with a simplified model of archetypical use-cases where AI can provide value.

AI Use-Cases

I consider three main categories of AI use-cases. Looking at AI from these three categorical perspectives allows us to appreciate its vast and diverse range of applications and potential, while also highlighting the areas of human activity where it can be most beneficial.

a) Data-driven Decision-Making: AI services in this category help humans make better decisions by providing insightful predictions. For instance, Google’s PageRank algorithm helps us make better decisions about which websites are relevant to our search query. AI’s predictive capabilities are a breakthrough in decision-making processes. They help distill complex data into actionable insights, enabling better-informed decisions. This leads to increased effectiveness in many fields, from business to healthcare, and from policy making to personal life. AI’s predictive power has changed the game in terms of our ability to anticipate outcomes and act proactively rather than reactively.

b) Efficiency and Precision: AI can automate or augment specific workflows, making them more efficient or precise. Robotic Process Automation (RPA) tools, for example, automate repetitive tasks, freeing up human time for more complex work. AI’s ability to enhance efficiency and precision in workflows revolutionizes sectors from manufacturing to customer service. AI-driven automation takes care of routine, repetitive tasks, reducing human error and freeing up human workers for tasks that require creativity, critical thinking, and emotional intelligence. This is not only cost-effective for businesses, but it also allows humans to focus on the aspects of work that are most fulfilling and valuable. (Yes, there definitely needs to be a discussion about how humans decide on the future of work here.)

c) Unique User Experiences: AI also enables novel user experiences that wouldn’t be possible without it. Speech recognition services like Siri or Google Assistant, for instance, allow for new interaction patterns with our devices. This provides opportunities for interaction and engagement that were previously unattainable. AI-powered services like virtual assistants or personalized recommendation engines can make our interactions with technology more natural, intuitive, and personalized – if they are well-designed. These AI-powered experiences are not just about novelty; they also make technology more accessible and inclusive, serving a wider range of users with varying needs and abilities.

This three-category framework helps us to understand and roughly frame the multi-faceted nature of AI, and its potential to transform a wide range of human activities. By looking at AI through these lenses, we can better understand where and how to apply AI effectively, and envision new possibilities for AI innovation. This view of AI also encourages us to think critically about the ethical considerations and societal implications of each category, and how we can design AI systems that not only work well, but also respect our values and contribute to a better society.

AI Capabilities

From an experience-oriented point of view, we can distinguish between three fundamental capabilities that AI services can possess.

1. Prediction: Predictive AI can anticipate user needs and tailor services accordingly, enhancing user satisfaction. ChatGPT’s large language model that predicts new words and phrases is a prominent example of a prediction machine. Weather forecasts also falls in this category.

2. Adaptivity: Adaptive AI learns from interactions and modifies its behavior over time. Amazon’s Alexa, for instance, provides increasingly personalized service by learning from user interactions. We would also place user-specific predictions, like movie recommendations in Netflix, in this category.

3. Agency: AI systems can have some level of autonomous decision-making capability, where the initiative shifts from a human to a collaborative “butler”-like service. Spotify’s curated playlists that get compiled “while the user is away” is one simple example, a robot stock investor that makes several buy-sell decisions within a defined scope is another. It’s important to note that agentive services typically are goal-based. A human communicates a high-level goal, and the agent finds ways to resolve the goal in the background, while the human can turn their attention to other tasks. An agentive service should be aware of its limits, so that it can alert the human if the goal is unreachable or if something unexpected occurs. There is also an even more powerful version of a goal-based agent: A utility-based agent operates with the primary purpose of maximizing the best possible outcomes from a set of alternatives. Going beyond the mere accomplishment of a goal, it seeks to optimize the path towards the goal, making the process faster, safer, and cost-effective. This type of agent distinguishes itself by its emphasis on the method or ‘utility’ of achieving the goal, using this approach to gauge the success of a particular state. In essence, utility-based agents elevate the AI decision-making process by factoring in efficiency and quality of results.

Granted, these three main capabilities have a tendency to blur, and sometimes it’s hard to draw a sharp line between them. Please note that this is not necessarily a problem. The framework is supposed to be a starting point for framing ideas for value-creating services. Therefore, the distinction between prediction and adaptivity might just be whether the prediction is influenced by a specific user model or not.

It is also important to highlight that there is not necessarily a technical “progression” involved in these steps. A robotic lawn mower definitely provides an agentive experience, while the AI technology in most robotic lawn mowers is certainly less advanced than, let’s say, a prediction machine like ChatGPT. The distinguishing feature is what it means for the user experience. If the service allows a human to give it a high-level order (“keep my lawn nice, please”) so that they can free up their time to do other things, it has an agentive quality to it.

Typically, what we refer to as “teaming” or “human-AI collaborative intelligence”, most often falls into the “agency” category. Another term related to this concept is the transition from “dumb” robots to “cobots”, where the collaborative and agentive aspect of workflows is in focus.

The Framework

When we cross-reference the use-cases and capabilities, we get a matrix that defines nine archetypical AI services. Each cell in the matrix represents a combination of a use-case and an AI capability. By studying existing AI services, we can place them in the appropriate cell.

These are just a few examples and there could be many more.

Putting the framework into practice: A medical agentive service

Let’s consider the field of healthcare, specifically the use of AI in assisting doctors in diagnosing complex diseases. As an example of an agentive and new service experience (3-c in the framework matrix), we can design an agentive medical diagnosis service.

In this service, an AI system, a patient, and a doctor collaborate to diagnose and monitor the patients’ health conditions using a combination of data analytics, wearable, the patient’s needs and goals, and the doctor’s expertise.

1. Preemptive Action by AI: The patient is equipped with a wearable device that continuously monitors vital signs and health metrics (heart rate, sleep patterns, activity levels, etc.). The AI in the wearable device uses advanced algorithms to analyze the data and detect anomalies or signs of potential health issues. When an issue is detected, the wearable recommends that the patient contact their doctor.

2. Initiative shift to the Patient: Based on the recommendation, the patient reaches out to their doctor, initiating a more in-depth diagnosis process.

3. Initiative shift to AI: The collected health data from the wearable is uploaded to the service platform. The AI uses the long-term health data, along with the patient’s current symptoms and medical history, to provide a more accurate and nuanced health analysis.

4. Initiative shift to the Doctor: The AI agent presents its findings to the doctor, including potential diagnoses, supporting evidence, and suggestions for further tests. The doctor reviews the AI’s analysis, combines it with their medical expertise, and makes an informed decision about the next steps.

5. Collaborative Decision Making: The patient, doctor, and AI continue to work as a team throughout the diagnosis and treatment process, with the AI providing continuous monitoring and learning based on the doctor’s decisions and the patient’s progress.

6. Continuous Learning and Adaptation: As the patient continues to wear the device, it not only aids in monitoring the patient’s health but also collects invaluable data for refining its predictive algorithms. It learns from each interaction, enhancing future predictive capabilities and personalized patient care.

This enhancement demonstrates a shift in initiative among the patient, the AI, and the doctor, with the AI offering a personalized and continuous health monitoring and analysis system that greatly augments the doctor’s capabilities. We can imagine this service being truly utility-based because it does more than simply achieving a pre-defined goal (detecting health anomalies). It optimizes for the best course of action based on various factors, ultimately aiming to ensure the health and well-being of the patient. For example:

1. Multiple Alternatives: The system isn’t limited to one type of action. It could suggest contacting a doctor, recommend lifestyle changes, or alert the user to take their medication based on the data it’s analyzing. The AI decides the best action from these multiple alternatives and would ideally be transparent and reason with the user about the best way to act.

2. Efficiency and Efficacy: The AI system doesn’t just work towards achieving its goal of monitoring health, it also focuses on how efficiently and effectively it can do so. For instance, it strives to provide timely alerts and accurate health insights to ensure quick medical intervention, which is crucial in health care. Here, there would also be an opportunity for the doctor to get aggregated insights from the patient as well (with prior consent from the patient, of course) which could influence what the AI will suggest to the patient.

3. Measure of Success: The success of this system isn’t only measured by whether it can detect health anomalies, but also by how well it can improve the overall health of the patient. For example, it might help the patient manage chronic conditions better or assist in early detection of serious illnesses.

4. Patient-based experience goals: The utility function could take into account factors such as patient comfort, ease of use, the impact on patient’s health outcomes, and the degree to which it helps the patient to live a healthier life according to the patient. This would require advanced capabilities in terms of communicating different kinds of experience goals between the agent and the patient.

5. Utility Function: The AI system utilizes a utility function to map the state of the patient’s health (input) to the best course of action (output). This function can be complex, considering various factors like patient’s historical data, current health status, potential health risks, and even the patient’s lifestyle or preferences.

This thought experiment would render an agentive and new type of value-creating user experience as the service takes the most beneficial action in a given state, optimizing for factors like timeliness, accuracy, patient satisfaction, and overall health outcomes.

The Future: Breaking away

In reimagining the future, we can see how AI has the potential to break from conventional molds and usher in an era of regenerative services and sustainable models. Instead of the typical use-cases that dominate today’s AI landscape, we could harness the power of AI to confront new challenges and embrace transformative possibilities.

To date, many AI services we encounter are, in a sense, expected. They follow predictable business cases, models, and application in traditional domains. While these advancements are significant, they merely scratch the surface of AI’s true potential. It’s time to go beyond the familiar and leverage AI’s potential in ways we’ve never done in practice before.

Consider, for instance, the potential for AI to foster democratic decision-making in local communities. Imagine an AI tool that could accurately predict the outcomes of proposed policies or initiatives, providing citizens with a clear, data-driven view of the potential impacts. This could lead to more informed, transparent decision-making processes that truly reflect the needs and aspirations of the community.

But we don’t have to stop there. In the same vein, AI could also be used to grow the commons, perhaps by streamlining the management and distribution of communal resources, or by facilitating cooperative initiatives that enhance communal well-being. AI could also help in the transition to a circular and regenerative economies by optimizing resource use and reducing waste. It would be very interesting to put all our focus on combining the strengths of community-engaged decision-making and AI-powered service design.

In light of these possibilities, below is a second matrix, one that offers fictitious examples of how AI can drive regenerative and sustainable change. These examples illustrate how the framework can be used to conjure new AI applications that are a departure from the usual and expected, offering fresh perspectives on what AI can do. They are a call to action, inviting us to shift our view of AI from a tool for incremental improvement to a catalyst for transformative change.

These conceptual services suggest how the framework can create new services using AI in supporting regenerative and circular economy practices. And of course, you can combine several of these into one coherent complex service. Remember, this framework is just a starting point. Don’t get limited by it.

As we explore the full potential of AI, it’s crucial that we look beyond the obvious and the expected. By doing so, we can harness AI’s potential to address novel challenges, foster regenerative models, and create a more sustainable and inclusive future. This new frontier of AI application is not just exciting; it is necessary for the well-being of our communities and the sustainability of our world.

Conclusion

While this framework provides a robust starting point for understanding and innovating within the AI space, it’s essential to remember its limitations. It does not capture edge cases in AI performance, or go into specialized AI capabilities, or adequately address the emotional, psychological, and societal impact of AI. Also, the distinction between augmentation and automation, as well as the complexity of multi-actor systems, might not be clearly represented – even if I hinted on a multiple actors in the medical service example. However, acknowledging these gaps allows us to continue refining our understanding of this rapidly evolving field, driving us towards a future where AI serves humanity’s best interests.

Hope this makes sense to you. If you have any feedback, thoughts, or questions regarding this, please feel free to reach out.