Why Is Enterprise AI so Hard?
Written By: Celia Wanderley
Published: March 13, 2023
Unless you live in a bubble, it is nearly impossible not to be bombarded with the promise of how artificial intelligence (AI) will change every industry and every area of our lives. Nevertheless, when we look at what is actually happening on the ground in the corporate world, the reality seems to be somewhat different.
I recently attended a conference which focused on the intersection of AI and the Internet of Things (IoT). As I attended presentation after presentation and walked around the expo, it made me empathize even more with my clients: how should one make sense of this gigantic promise, make sure you are not missing out, decide between platforms that seem to do pretty much the same thing and know where to start, how long it is going to take and where you are even trying to get to? How is your legacy data and legacy technology going to integrate with the new? Will you have the right level of connectivity if you have remote plants or manufacturing floors? Where will you find and retain the highly specialized talent that will be required to do this work?
You will hear examples of what others may have done and honestly, a lot of it sounds more like the proof-of-concept stage. Then come the major concerns with adoption: how does one ensure the humans involved in the process are not left behind, can we actually be active contributors and, if possible, evangelists of a new way of working? And, of course, on top of it all comes the very valid question of how privacy, ethics, and governance will not be forgotten so that the solutions built are not big black boxes that perpetuate or exacerbate the bias that has historically existed in the data that we create and capture.
Enterprise AI Seems Hard Because It Is
The matter-of-fact reality, whether we accept it or not, is that enterprise AI is really hard. But if you only listen to the vendors trying to sell it, chances are you will rarely hear that. I am also in this business, so I can safely say I have fallen in that trap and need to constantly remind myself how others might feel on their side of the table. To be clear, that does not mean I do not believe the promise of AI is not real; I see evidence every single day and could not be more excited to help make the path easier for others any way I can.
A Typical Enterprise AI Journey
To put things into perspective, a common pattern my team and I see in the corporate environments are demonstrated below:
Figure 1 – A Typical Enterprise AI Journey
In many cases, it all starts with the fear of missing out. It seems like everyone is doing AI, everything you read says there are millions, billions or even trillions of dollars at the table, every vendor is offering you a platform that looks amazing and they're all telling you how successful your competitors are.
But, before you can dive in, you need the people who can do this type of work. So, you say, 'Let’s go out and hire.' Soon though, you realize this step is much more difficult and expensive than you anticipated.
Candidates seem to work in a very different manner and with very different tools. Adding to the confusion are the questions:
- Should this be a completely new intake process?
- How do I get the ideas that are a fit to this type of solution?
Before you know it, enterprise AI starts to seem like a solution looking for a problem and competing for the dollars you already earmarked for everything else. Is this “business” or is this “technology”? Who is best prepared to lead it?
Next, many decision makers pick some initial ideas and assign them to current teams while attempting to figure out intake. Six months pass, then a year and hundreds of thousands of dollars, and you’re thinking, “What?! Nothing in operations? Where did time and money go? Let’s move to a different idea and let’s see if we can take this one to operations.”
But, before you can, you lose your data scientists to a higher-paying job, to a tech company and/or to a more challenging endeavour. So, you post, hire, and onboard and the cycle starts again.
Yes, I know, this is a generalization and not necessarily what happens every time in every company. That said, it happens a lot. Having seen it first-hand, over and over, made my team and I scratch our heads and think, 'There has to be a better way.' If our goal was to proudly wear the “trusted advisor” hat to our clients, we needed to at least try.
Accelerating Enterprise AI
For the sake of simplicity, I will use the diagram in Figure 2 to contrast Figure 1 with what I call “An Accelerated Journey.”
Figure 2 – An Accelerated Journey
The key premise behind this accelerated journey is that one plus one equals three.
What if:
- You could reuse the parts of the process that are repeatable?
- You could have a bench of highly skilled resources that you can swap in and out based on the specific skills you need (e.g., natural language understanding, computer vision, time series, knowledge graphs)?
- Your resources felt like they did not have to go elsewhere to work on challenging problems?
- Your resources have peers with whom they can speak the same language because they can work closely with this big bench of specialized resources?
- You did not have to reinvent your intake and operating models because you can leverage what worked for others?
- You did not have to reinvent methodology because you could reuse what has worked hundred of times?
- You had an honest answer, based on real experience, of what it takes, how long it takes and how much it costs to move from A to B?
- You could also have an honest and unbiased opinion of which technologies and platforms you could work with?
It was only after my team and I asked ourselves those questions that we set out to package what we learned by building solutions for our own ventures and for clients in multiple industries hundreds of times.
Conclusion
So, why does enterprise AI seem so hard? In short, because it is. But, it doesn’t have to be. By having an honest conversation with our clients – not a conversation that says “you are missing out” and “I can’t believe you are not further ahead,” but a conversation that acknowledges the challenges of enterprise AI – my team and I have found there are ways to both de-risk and accelerate the journey.
