An Artificial Intelligence Capability Model

Abstract

An Artificial Intelligence Capability Model will allow us to understand the potential breath of what AI could do within an organisation. It would be a high-level map of what could be possible, and it will allow us to choose any capabilities that are important to us, without losing the big picture of our AI growth potential. This unified and technology-independent map will assist us in defining a top-down overall approach towards AI implementations. This artefact can also guide the AI maturity assessment of an organisation.

?

Disclaimer

The content in this document reflects the view of the author and not a view of organisations where the author is or has been affiliated. In addition, the content should be seen as an opinion and not as an advice.

?

Acknowledgement

The book “Artificial Intelligence – A modern approach” by Russell and Norvig has been invaluable in writing this document. This architectural artefact also incorporates ideas from papers written by the Boston Consulting Group, as well as Enterprise Architecture concepts by the Open Group. Links are available in the Reference section, at the end of the document.

?

What is a capability Model?

According to the Open Group (An Enterprise Architecture International authority), “a business capability delineates what a business does without attempting to explain how, why, or where the business uses the capability”. They also say it “involves identifying and describing what needs to be done by the business in support of its overall mission.”

Business Capability Models (BCM) are graphical representations of structures of capabilities to better visualise the scope of what a business does. Capability models are often guides for mapping business levels of maturity or how applications or technology contributes to a set of business capabilities.

Value Stream

A value stream is a?sequence of activities?that are necessary to deliver a product, service, or experience to an internal or external customer.

In this case, the sequence of activities to deliver AI is listed as below.

• The Perceiving step is the ability for machines to see, hear, and being able to understand natural language.
• Thinking will provide the solutions with Machine Learning AI models and the ability to autonomously solve problems.
• The Acting step is about putting together multiple AI capabilities into a unified front, typically represented by an Agent or a Robot.
• Operationalising is about converting concepts into actual data and technology, including a way to successfully repeat its process and maintain it.

Capabilities

Then, every step in the value stream is supported by a set of capabilities.

Our AI model has ten capabilities in total.

• Computer Vision & Hearing: Capability for computers to establish an understanding upon digital images, sounds and video.
• Natural Language Processing: NLP enables computers to comprehend, interpret, and manipulate human language.
• Reasoning & Problem Solving: Thinking process about something in a logical way to form a conclusion or judgment. It involves the ability of the mind to understand and analyse information systematically.
• Machine Learning: Developing algorithms that allow machines to learn from data and make decisions or predictions without explicit programming.
• Generative AI: This is about being able to generate new content (images, sound, text, etc.) by using algorithms or AI models.
• Automation & Autonomy: Automation is a range of technologies that reduce human intervention in processes. Autonomy is the ability to make one's own decisions without external influence or interference.
• AI Platforms: AI Platforms are the technical components where we develop and run AI models.
• AI Governance: Through Governance, we can ensure there is accountability for the AI models, and that they adhere to predefined ethics and regulations.
• Operations: To be sustainable, AI models need to be properly developed, monitored, and supported.
• Security: We need a secure environment to protect AI models, insights, technical interfaces and networks.

Capability Enablers

The next level down in the AI Capability Model is a set of elements that will enable the AI capabilities. Those enablers will be segregated in four categories.

• Models & Algorithms: All mathematical and statistical models that are required for a capability to exist. Most of the AI capabilities will be propelled by models and algorithms.
• Technology: Some of the AI capabilities will require actual technology components to be enabled. Cloud Platforms would be a typical example.
• Policy: This is a combination of legislations, standards, and principles to have a successful and sustainable AI environment.
• Technical and Non-Technical Processes: Technical Processes would involve development and support of AI models (e.g. how code is written, tested, and deployed). Non-Technical Processes would focus instead on human interactions, communication, and decision-making.

Measuring AI Maturity

The Capability Maturity Model Integration (CMMI)? is a proven set of global best practices that drives business performance through building and benchmarking key capabilities. The CMMI framework is a typical and recognised approach to measure maturity levels within an organisation.

By overlaying the CMMI framework over the AI Capability Model, we could assess our AI maturity.

An AI Maturity Assessment Use-Case

Here is a use-case example for using the AI Capability Model to measure an organisation’s maturity.

Let’s imagine a commercial airline called AI Airways. AI Airways is a small but very technology savvy airline. AI Airways is proud to have advanced Marketing capabilities that provide customers with customised offers according to their consumer behaviour. AI Airways also has acquired robots to improve their baggage services at airports, and is only introducing facial recognition technology for check-ins. They also have a Natural Language bookings app, and a virtual expert travel agent.

The organisation also has made a significant investment in Cybersecurity, and their Agile capability is fairly developed.

Unfortunately, AI Airways has appeared in the newspapers and social media with bad publicity. People complained their Marketing engine is discriminating customers based on race and postcode.

The AI Airways board requires an assessment to learn not only how to avoid bad publicity, but how to become the most advanced technological airline in the world.

Enterprise Architects at the organisation use a Capability Model to assess the maturity levels the organisation has in the AI space. As a result, they created the heatmap on the left. This assessment will allow the AI Airways board to make an investment decision for the immediate steps and road ahead.

Afterword

An AI Capability Model could be an important tool to put in context a structured big picture of what AI can do for your organisation. The AI capability model could be used to be mapped against applications, org charts, projects, goals, objectives, and even the Enterprise Business Capability Model your organisation has. That will provide you an Enterprise Architecture holistic view; it will also help you and your leadership to better understand how AI adds value, and how that value is derived.

References

Expanding AI’s Impact With Organizational Learning by MIT Sloan & BCG (2020)?? https://web-assets.bcg.com/1e/4f/925e66794465ad89953ff604b656/mit-bcg-expanding-ai-impact-with-organizational-learning-oct-2020-n.pdf

BCG Robotaxonomy (2023)?? https://web-assets.bcg.com/30/ac/fb11b7cb4ed3bda99f1598d71c19/bcg-robotaxonomy-chart-2023.pdf

Artificial Intelligence by Stuart Russell & Peter Norvig, A Modern Approach, 4th Global Edition (2021)?? https://www.booktopia.com.au/artificial-intelligence-stuart-russell/book/9781292401133.html

TOGAF Business Capabilities – The Open Group (2022)?? https://pubs.opengroup.org/togaf-standard/business-architecture/business-capabilities.html

Introduction to CMMI?? https://cmmiinstitute.com/cmmi/intro