A question I get asked often is how does the journey to adopt AI look like. Where should a company or individual start? What considerations are important to keep in mind as you plan to infuse AI in everything you do?
In this article, I will explain my take on this, with a practical framework I've developed while working with a variety of organizations. In following articles I will provide a canvas you can use to identify tangible ways you can leverage AI in a way that solves for a specific pain point and helps you achieve your business goals.
As a starting point, you need to realize that the adoption of AI is indeed a journey, and you can build your own using what I call an opportunity framework. I'll analyze the opportunity framework left to right, and I suggest that you navigate it step by step.
The simplest applications of AI that will give you immediate results are those that impact an individual's daily routine, by either streamlining, augmenting or automating what they do. Examples of this are any tools that help you consume information quickly, do Q&A on a knowledge repository, extract insights from presentations, images, video, all interacting in natural language. A brilliant example of an individual productivity tool is Google's NotebookLM (check it out if you haven't already tried it). These applications are mere augmentations of an existing consumer journey, but they are very powerful because they allow to go through a routine faster and more efficiently.
The step after this initial introduction to AI-powered tools is to scale whatever has worked for an individual to a team. The tasks impacted are still things like knowledge retrieval and Q&A, help with coding, natural language interaction with a database (natural language-to-SQL type of tools). An example of this is, if your teams use gmail, docs, drive, etc, they can now leverage Gemini to ask questions like:"Summarize all the emails I received from my boss this week". These examples are related to internal productivity, but the same can be expanded to your end-customers: think about adding an agent to your website that helps customers find information and ask questions about what they're looking for.
These sets of use cases and related tools directly tackle some challenges that are technological in nature, in the sense that it's thanks to LLMs as a technology that we can now upload hundreds of structured and unstructured documents to a tool like NotebookLM and ask questions about any of that content in natural language.
Things become more complex once we start approaching domain-specific tasks, use cases and challenges. For example, the question:"Can I use AI to develop a cure for diabetes" is a domain-first question rather than a technology one. Technology can help you solve that problem, but you need to first and foremost understand the domain in order to build anything to tackle that issue. A major consideration when going through this journey is how to bridge the gap between technology challenges and domain-specific challenges, which also includes how to get technicians and domain-experts work closer together to develop a technology-enabled solution that solves for domain-specific challenges.
The low-hanging-fruit of domain-specific applications is the usage of AI to analyse strategies, decisions and ideas from years of operating in a certain sector. For example, imagine a tool that reviews all the products you've ever sold on your online store and related performance, and helps you understand why some products performed better than others.
Finally, we start thinking of how to identify innovative applications of AI that can actually disrupt the market. The most obvious example is to train a foundational model with deep expertise of a very niche set of problems that are unique to your market, using proprietary data, and solving for an issue that affects many, but very few understand or ever attempted to solve. These kinds of applications are complex from both a technical perspective, and in terms of the domain-knowledge required to succeed (not to mention funds and top talent). Fortunately, I do believe there are innovative and disruptive applications of AI that don't involve building and training foundational models from scratch. For example, using AI to find competitive edges in new markets that weren't reachable before, thanks to the fact that you can now offer your product to a wider audience, or for a cheaper price.
With this opportunity framework fleshed out, you can start thinking about where do you sit in this journey and where you want to get to.
In my next article I will go into more details on how to identify promising AI-powered solutions that can help you solve your most pressing business priorities.
Stay tuned!
