Watching the 2025 Horizon
Where innovation in AI shows strong promises of growth
Oh hey, hello there.
It’s been a beat since we last spoke. I’ve been busy unplugging with family and enjoying a slower pace for the holidays with friends, as I hope you’ve had the chance to as well. Although the time off was very necessary, I’m excited to be back and haunting your inbox with a fresh batch of insights, advice, and zany AI metaphors.
2025 promises to be a big year. If you’re anything like me, you’ve been perusing the predictions for where AI will be going this year, and maybe even forming some of your own.
If you’re also like me, then you have come to the conclusion that it is just too early to tell where in the world AI will go over the next twelve months. I may have some hunches, but I prefer to attentively watch and patiently wait for a few major developments to finish unfolding first.
So, I decided that instead of sharing with you my AI predictions for 2025, I’d instead share with you the areas I’m paying special attention to as ample spaces for new innovation this year.
It’s far more fun to have someone to watch and discuss with than pretend I know the answers to everything 😉
TL;DR - Areas ripe for innovation in AI include: process, use case and information.
Bonus: Tune in Tuesday the 21st for a discussion with Philipp Adamidis and Antoine Gautier, founders of QuantPi, to learn how they approach building trust into the AI lifecycle.
Innovation can be a fickle mistress. Constantly in flux, it can be difficult to pin down and define what will be the determining factors for success.
But, just because we may not be able to predict what will make innovation grow, doesn’t mean we don’t know where ample ground is being sown for it to flourish.
Of course, when it comes to AI, my definition of fruitful ground for innovation is heavily influenced by the fact that AI has, and always will be, a human story. And, as an ethicist, in order for that story to be successful, it must be rooted in our human values.
Which means that when I am talking about opportunities in AI, I’m specifically talking about the spaces I see opening up for true innovators to align our technology with our values and redefine what it means to be successful in AI.
So, without further ado, I present to you my three hotspots for innovation that I’ll be watching with particular attention unfold throughout this next year.
In 2025 I’m watching for innovation…
…in the Process
The idea that any AI can simply be a plug-and-play tool is nothing more than an empty promise.
One of the most common blockers I heard last year from business leaders looking to enable AI within their teams was that their old processes were getting in the way of adoption. No matter the size of company or team, everyone seemed to be arriving at the same conclusion:
AI will always require some form of change management if you have any hope of making it stick.
This fact leaves us with an interesting spectrum for AI adoption:
On one end, companies can look for solutions that target specific task sets with relatively low implementation needs. These solutions are typically designed to automate pre-existing tasks, with the heaviest lift being to train the humans how to use it. The challenge on this end of the spectrum is that unless the AI shows immediate return on investment and recurring use, the chances of it sticking quickly decline as the ‘cool new tool’ factor fades.
On the other end, companies can look at their current processes as a whole and use AI as a catalyst to reinvent how they do business. Instead of looking to automate pre-existing tasks, these solutions look to pioneer new ways of working and so capitalize on the depth of potential AI brings to the table. The holistic approach to this end of the spectrum creates the environment necessary for AI to stick. However, it does significantly increase the time and resources needed to complete the change.
Essentially, we are looking at a tradeoff between AI stickiness and investment into change management. The more you invest into change, the more likely AI is to stick, the less you invest, the harder it is to tap into the promise of these tools.
Which leads me to the first of my 2025 innovation hotspots: The Process.
I have seen a plethora of companies target the lower end of this spectrum. I have also seen these solutions result in poor adoption rates and exposure to unnecessary risk.
So as we dive into 2025, my interest is not in the latest AI development that promises to improve efficiency on a singular task within a pre-existing process by X amount. What I’m watching for are the companies playing the long game by going all the way to roots and innovating on their processes with AI.
After all, if AI promises to invent new ways of working, then we have to be willing to reinvent the way we work in the first place.
…in the Use Case
One of my favorite mantras coming out of 2024 has to be:
Just because there is a solution, doesn’t mean it solves a problem.
We often make the assumption that every solution is born out of need to solve an existing problem. However, the recent flooding of generative AI solutions onto the market has been brutal in teaching us that not every solution has a problem.
Case in point, when I speak to business leaders about their needs in AI, I rarely ever hear someone say they need more options on the market. Instead, I’m flooded with requests for AI use cases.
In other words, we have an abundance of AI solutions. What we do not have is an abundance of ideas on how to use them.
Which leads to my second innovation hotspot of 2025: The Use Case.
There is no denying that GenAI tools are cool, but cool does not automatically translate to useful. From my experience, innovation that starts with the technology and tries to retrospectively fit it to a need is doomed to fail from the start, and, in some cases, I would even go so far as to say just plain lazy (yes, them’s fightin’ words).
So, instead of getting swept up into the latest model development of Omni-Strawberry GPT 13.500, I’m on the lookout for the innovators who are coming to market with use cases that address well defined needs of clearly identified users.
…in the Information
Last but not least, I would be remiss if I didn’t address the recent trends towards agentic AI that colored our newsfeeds almost as quickly as the autumn leaves covered the fall of 2024.
What is agentic AI?
Agentic AI is a type of artificial intelligence designed for autonomous goal-oriented action and decision-making. The purpose of these systems is to adapt to changing conditions and complex scenarios in real time, making it the desirable choice for enterprise looking to balance operational agility and context-sensitive reasoning.
As we began to comprehend just how massive a limitation the hallucinations of LLMs and generative AI were proving to be, an opening in the AI ecosystem opened up for the next evolution in AI architecture to take place.
And in late 2024, agentic AI did just that.
By the time we were wrapping up both the year and our holiday gifts, agentic AI had not only snuck onto the AI scene, but was clearly positioned to take main stage for 2025.
One of the primary attractions of these systems is the ability to automate decision making, which in the operationally complex conditions of enterprises makes it a true game changer and reason enough to watch this space for innovation in 2025.
However, this is not necessarily what has my interest peaked. Although I will be keeping a close eye on agentic AI developments this coming year, what I’m really watching for instead is innovation in what these systems are enabling us to do.
Which leads to my final innovation hotspot of 2025: the Information.
Think of it this way, through chatbots we are able to “talk” to the data, meaning I can ask a question in plain language and receive an answer back again in plain language. The only challenge here is that when I am using generative AI, I have no way of knowing if that answer is correct unless I utilize my own reasoning and fact check.
Now, let’s add agentic AI. I am still able to ask questions and receive answers in plain language, but now thanks to the reasoning and decision-making capacities of agentic AI I know the answers I’m receiving have been checked against fact and context.
Essentially, I am able to engage directly with the information instilled in the data as if it were a living dynamic organism. Which signals to me ample opportunity for innovators to reinvent how we humans interact with the information of the world around us.
While LLMs introduced us to a new horizon in user interface for AI, agentic AI mobilizes this interface to bring us a new horizon for engaging directly with information unlike we have ever seen before.
*Please note I am deliberately avoiding the term ‘human-like’ here, nor have I likened agentic AI to the human brain. I’ll dive into why, as well as the depths of agentic AI, in a later newsletter edition.
How do you test for trust in AI?
Amid the general boom of AI adoption there has been an underlying concern for the trustworthiness of these systems we are rapidly embedding into our companies, daily lives, and beyond.
Although we may not agree on whether it is the technology or the people building it that need to be trusted, we can all agree that trust is essential to being able to push the cutting edge of AI.
Without trust, we will never have the confidence or support necessary to move forward in bringing AI development to new innovative heights.
Tune in Tuesday the 21st for a discussion with Philipp Adamidis and Antoine Gautier, founders of QuantPi, to learn how they approach building trust into the AI lifecycle.
The first of your major points is a decades long truism for implementing any major technology: 1. Do you pave the cow path? 2. Do you invest in all aspects of implementation (not just change mgt) in order to be successful?
For the second major point: Can we call it a solution if there is no problem? Seems like a nomenclature issue
Thanks for the article Stimulated some thinking