BLUF (Bottom Line Up Front)

  • Your AI projects can fail due to a lack of true and business-minded data scientists
  • Your AI projects can fail due to a lack of a vocal C-level advocate
  • Your AI projects can fail due to a lack of dedicated resources.
  • Your AI projects can fail due to detractors working actively to sabotage progress on the initiative.
  • Your AI projects can fail due to unavailability of IT and business resources to operationalize AI models.

Let There Be Buzz…

“AI-driven companies will take $1.2 trillion from competitors by 2020.”


Artificial Intelligence Will Drive The Insights Revolution

“The leaders in artificial intelligence will rule the world.”


“AI will generate $2.9 trillion in business value and recover 6.2 billion hours of worker productivity by 2021.”


Let’s be honest. There’s no doubt it can be exciting doing sales in the field of Artificial Intelligence (AI) and Machine Learning (ML) nowadays. With Ray Kurzweil predicting that the AI winter (the late 80s and early 90s) is over and lots of others echoing with similar sentiment [4], one starts to think this time is different. We’ve come a long way over the data availability and quality issue, as well as over compute power struggles. This time, it seems we finally have a reason to justify our enthusiasm.

But could there be other risks on the path of AI transformation at your organization?

Why Is AI Hard?

You have probably heard of the Data Scientist being referred to by HBR as the “sexiest job of the 21st century” [5]. Now, I am not going to rehash our definition of AI (and to be honest, although I mostly do not disagree with Craig Martell’s definition given the context, it would be wrong for me to not admit that it is a quite narrow interpretation of AI, that is, to equate AI with ML. There are still many rule-based systems in production everywhere, from banking to manufacturing, all of which may be considered as performing AI tasks.) I will simply point out, the true data scientists that have the math, coding skills, and business domain expertise are unicorns. In other words, there are not many of those out there. And to implement our “AI” solutions, we need them.

So the first problem is to hire real data scientists who have real-world ML experience. The global demand for machine learning solutions greatly exceeds the production capacity of all the data scientists in the world. More likely than not, your AI projects are at risk because your expensive data scientists might be lacking any one of the skill sets mentioned above (e.g., Ph.D. data scientists who do not care about business objectives and only want to play with the latest algorithms).

Now, suppose you fix that by either enforcing multiple layers of filtering in your hiring process or by acquiring some automated ML platform that brings ML to another layer of abstraction (so that, for example, your business analysts or app developers can use it), your next concern is you do not have a C-suite advocate.

How so? Well, without clear direction from the highest level, you will have a difficult time getting fully dedicated resources (i.e., data scientists and accountable owners that will plan, track, and execute the projects). 20% time will not cut it.

Moreover, AI projects are complicated teamwork. It takes domain experts who bring knowledge of the business and business problem, as well as knowledge of the data. It needs engineers with the ability to write code to gather data, explore/inspect data, manipulate data. It takes developers to write applications to extract actionable items, to build models, to implement models. And it takes specialists who know foundational statistics, the internals of algorithms, who also have practical knowledge and experience, and know how to interpret and explain models. Without those resources available to you, your AI projects will be at risk.

Now finally, there will be internal detractors who will actively work against your AI initiatives for various reasons. Some might just be against changes. Some might be clinging to the status quo. Some might feel threatened. If you cannot manage detractors, they will sabotage your AI projects.

The AI-Driven Enterprise, A Dream Only?

The logical question to ask at this point is: if operationalizing AI is so complex (it is not just some algorithm and a model), how do I get it off the ground and will I still get the ROI after all the orchestration and change management effort?

Starting with eliminating the aforementioned red flags may ensure you are putting your best foot forward at least. And a top priority in that list is perhaps to get C-level support. With that in place, the rest pieces of the puzzle usually fall in place quite nicely.

On the other hand, this challenge can be turned into an opportunity for AI-solution vendors that are diligent and capable of finding ways to partner with enterprises to solve their particular pain points. A friendly reminder for AI vendors in this respect, is to qualify the prospects thoroughly and carefully — not all enterprises may be AI-ready! (Technology wise, AI winter might be over. But we might not see the AI spring we are all waiting for if companies cannot get over a “psychological” barrier.)


[1] Forrester Predictions 2017 https://go.forrester.com/wp-content/uploads/Forrester_Predictions_2017_-Artificial_Intelligence_Will_Drive_The_Insights_Revolution.pdf

[2] https://www.cnbc.com/2017/09/04/putin-leader-in-artificial-intelligence-will-rule-world.html

[3] https://www.gartner.com/newsroom/id/3837763

[4] AI Winter https://en.wikipedia.org/wiki/AI_winter

[5] https://hbr.org/2012/10/data-scientist-the-sexiest-job-of-the-21st-century



Users who have LIKED this post:

  • avatar