Cognilytica’s AI Positioning Matrix (CAPM)

“It’s not the position, but the disposition.”

— Susan Sontag, Author

Why is a Positioning Matrix Needed?

The market for artificial intelligence (AI) products and services continues to expand at a dramatic pace. Every day there’s a new development or announcement of a new AI product, service offering, technology, and application of AI. How can enterprises looking to adopt AI understand how these various solutions are placed relative to each other, the overall AI market, and the applicability to their specific use cases? In the fast-paced world of AI, enterprises and vendors alike need a simple way to see how they are positioned in the market versus the other solutions and customer requirements.

There already exist different sorts of market positioning approaches from other thought-leaders in the space, but they all suffer from the same problems. They either focus on how the technology is positioned on a maturity spectrum, ranging from less-sophisticated to more-sophisticated technology approaches, or they focus on general market traction measures such as the breadth of a solution provider’s offerings or how well they are able to execute against their own plans.

However, these measures are inadequate in an industry that’s maturing at a super-fast rate, such as is the case with AI. Rather than looking at a single (and possibly arbitrary) measure of maturity or a company’s ability to execute, we need to look at positioning from the customer’s perspective: what is the “sweet spot” for a particular application of AI in a particular market, and how are the various solutions positioned according to both the applicability of the solution to a specific problem area and its overall sophistication with regards to that application. From this perspective, solutions could be too-sophisticated for the problem areas they are trying to address, or the problem areas might require more sophistication than what is commercially available at the time.

Introducing the Cognilytica AI Positioning Matrix (CAPM)™

To solve the needs of customers trying to evaluate the constantly changing landscape of AI solutions, we’re introducing the Cognilytica AI Positioning Matrix (CAPM)™. The goals of the CAPM matrix are two-fold: identify for each particular application of AI, what the “sweet spot” of required complexity for AI solutions are and how broadly applicable to a range of AI problem areas that solution needs to be; and position various solutions against that same matrix.

One axis of the two-dimensional CAPM is a measurement of that particular solution or problem area’s AI sophistication and complexity. The other axis is a measurement of how applicable that particular solution is to a range of AI problem areas. Placed on this matrix are bubbles of various sizes that address different AI solution areas that have different areas of coverage against both of these axis. Likewise, we position different specific solutions against this same chart, with the size and shape of the bubble indicating how that solution spans in either applicability or sophistication depending on how it is implemented.

Below is a visualization of a CAPM with a few applications of AI illustrated:

In this particular CAPM, we can see that Artificial General Intelligence (AGI) is positioned at the top right sector of the matrix because it has a very broad applicability to all aspects of business and personal life, and at the same time is extremely complex and sophisticated to implement. So much so, we haven’t been able to realize AGI yet, which is also why it’s beyond the currently available technology threshold line — explained further below. On the other hand, we can see that Assistant-Enabled Commerce (AEC), AI-enabled Drug Discovery, and AI-Assisted Driving are areas that are much simpler and less sophisticated to implement, but at the same time have less general applicability to different use cases and situations. In the middle, we can see that chatbots have moderately broad application and a range of complexities depending on the chatbot, while intelligent assistants has a narrower complexity range but a broader spectrum of applicability depending on the type of assistant.

The AI Complexity / Sophistication Axis

With only two fundamental axes, it’s important to understand what each axis represents and how Cognilytica uses it to position various solution or requirement areas. The horizontal axis represents how much advanced research, development, and implementation complexity needs to be put in to provide the required solution. Clearly, as new developments progress, the line between what’s complicated and what’s simple changes, but in this ranking, we simply indicate “relative” complexity. What’s complicated today may be simpler tomorrow, which means that more of what we would like to implement or solve with AI becomes possible.

Often, but not always, the more complicated a solution or technology is, and the more to the right it is indicated in the CAPM, the more “in the future” the technology is. If you’re looking for what is implementable today, or with the least amount of complexity, or with more existing, proven solutions, look for technologies more to the left on the chart. If you’re looking to be on the bleeding edge of AI adoption and looking for more complicated problems to solve, look for those solution or technology areas towards the right side of this chart.

The Breadth of Application Axis, and those Bubble Sizes

Likewise, the vertical axis identifies how broadly applicable a particular solution or technology is. Is this technology or solution area only relevant to a narrow industry or application area, or is this something that can be adopted by a wide range of industries, customer types, and usage scenarios? The reason why this measurement is important is because not every technology or AI use case is good for every scenario. Knowing when, or when not, to apply a technology is highly important if you want to reduce the risk and increase the ROI of your AI implementations.

Just as is the case with the Complexity Axis, being more “broad” is not any better than being more “narrow”. Some AI use cases, such as Assistant-Enabled Commerce (AEC) don’t need to be relevant to non-commerce areas of the business. Likewise, augmented intelligence solutions get their benefit by being broadly valuable to the business rather than useful only in a narrow context. In some cases, the specifics of the implementation matter to determine how narrow or broadly applicable an AI solution or technology is. For example, some intelligent assistants are very narrowly applicable and not particularly sophisticated while others are very broadly applicable and have relatively more complexity and sophistication.

This is why the “bubbles” for each particular solution or technology vary in size. Some of the solutions have a narrow range of complexity and applicability, but other solution areas can vary widely in both AI complexity and applicability, across different ranges. This means that when evaluating a particular technology implementation or solution for an AI use case, you should be aware that if there’s a great range in complexity and applicability for that one particular AI use case, then two different solutions might fall into different places on the CAPM chart. This is what makes the CAPM useful and powerful for customers in evaluating what’s a good fit for their particular AI situation.

The Currently Available Technology Threshold

In many of our CAPM charts, we’ll also draw a red dotted line vertically down the chart that shows where the AI technology capabilities currently stand in the industry. Basically, this line defines where is the “cutting edge” of technology that’s at the frontier of what we’re capable of doing with AI. Everything to the left of that currently available technology threshold are technologies and solutions we can implement today with available technology. Everything to the right of that line are things that are either only being done theoretically or we don’t have the available technology to do them well, or even at all.

In the example chart above, we can see that AGI falls well to the right of this line, meaning that we’re far away from any sort of realistic AGI implementation. We can also see that Autonomous Vehicles (AV) are emerging into the realm of technology being available and implementable. There’s still some aspects of AV technology that are not yet figured out, and that’s why part of the bubble extends beyond the threshold line. Similarly, we can see that while organizational approaches to Explainable AI (XAI) are easily implementable today, the technology approaches to XAI, with automatic explanations for reasoning and decision models is barely possible today, with the bulk of what we need from XAI just beyond the threshold of what our technology can currently do.

Over time, as our AI technologies improve, this currently available technology threshold line will move increasingly to the right, meaning that more of what was theoretical yesterday is becoming possible. This can help enterprises as they plan for their AI initiatives to know what they can implement in a few years as more of AI becomes practically implementable.

Don’t Focus on the Top Right – Focus on being Optimal

In some analyst evaluation matrices, being in the “top right” section of the matrix is the desirable place to be. However, this is most definitely not the case with the CAPM. Each particular solution area in the CAPM has a “sweet spot”. The more that a particular solution fits into that “sweet spot”, the more it will be optimal for the particular problem it is trying to solve. If a solution uses too sophisticated technology than what is required for the particular problem, then it’s not a good fit. It could be too expensive, too complicated to implement, too brittle, or too prone to veering off track from your solution. Likewise, if it is too simplistic, then it might not be AI “enough” and require more manual work or other solutions to fill the gaps that the solution can’t provide. Furthermore, if a solution is too broadly applicable, then it won’t be well-suited for a more narrow task, and the inverse is also true.

So, enterprise implementers should use the CAPM as a way of making sure they source their AI solutions to best fit their needs. Likewise, professional services and consultancies should use the CAPM to evaluate the technology suppliers they will use, as well as the positioning of their own professional services. Vendors should keep an eye on how Cognilytica is positioning them to see which solution areas they are most suitable for and how they stack up against their competition, which might be more, or less, suitable to certain solution areas.

How Cognilytica Uses the CAPM™

The Cognilytica AI Positioning Matrix is an important tool for Cognilytica to understand the market, the technology providers, the use cases, and required solutions that AI can provide. We use the CAPM extensively in all our Briefing Notes to evaluate vendor solutions as well as in all our research to indicate how relevant the particular topic area is to the AI landscape as a whole and how it compares with other AI solution areas. In our briefing calls, we’ll be looking to help our enterprise users understand how their specific needs map to the various applicability and complexity capabilities of different AI solutions, and our vendors to position their technologies in the matrix.

If you’re looking to get a briefing with Cognilytica, hire us for your specific AI advisory needs, work with us for white papers or other contributed content, or engage with us for any of our activities, then you should understand how our CAPM works, how it can work for you, and how we use it to advance the state of AI adoption while reducing risk for our enterprise, vendor, and professional services clients.