MENU
You’re in the board room, the CEO walks in with the new CTO or CDO and bangs their hand on the table saying “AI! AI! AI!” like they’re the second coming of Steve Ballmer shouting about developers. Everyone in the room nods their heads furiously in agreement. “Oh, AI will change everything! It will write our code for us! It’ll scrape our data for us! It’ll make julienne fries for us!” There is nothing that Artificial Intelligence can’t do and so the directive has come down and we’re doing this.
Six months later, nothing has been done. You’ve hired two data scientists who are already so frustrated they’re thinking about leaving if they haven’t already. Your new data lake implementation is a mess. You did all the things that the Harvard Business Review told you to do to get your organization ready for AI but nothing got done… why?
There is a nasty little truth about Artificial Intelligence and Machine Learning that too many organizations don’t want to recognize.
They aren’t ready for it.
The organization isn’t trying to walk before it can run. It’s trying to launch a spaceship thinking they’re playing Civilization and haven’t gotten out of the Bronze age. I get it. We all want magic bullets and as Arthur Clarke noted in 1962, any sufficiently advanced technology is indistinguishable from magic. Therefore, AI is the magic bullet, right?
No.
There’s a reason why in Star Trek, the Federation believes in the Prime Directive. There is a minimal level of technological achievement that is required before you move on to what could be an incredibly huge technological leap, otherwise you destroy yourself with what you haven’t earned. Organizations that rush into AI and ML are running headlong into an oncoming train with no idea or vision about how hard they’re about to get hit.
So then, as an organization who is dedicated to helping our customers succeed with their data and analytics from prescriptive to predictive, what do we see in organizations that have been successful with AI / ML implementations?
An organization that is going to implement AI / ML must be an organization in which people are encouraged at every level to ask tough questions and be encouraged to find honest, genuine answers, even answers that may make the organization uncomfortable or even more scary, question its priors.
However, even that seems premature. As I write that I can hear a thousand voices ring out in unison from business schools across the country, “Culture is earned!” Which is true and its also amazing how they understand that culture is earned but miss that implementing the largest technological change in an organization’s history also must be earned. So where does fostering a genuinely inquisitive culture begin?
That requires data democratization. Data democratization requires a method by which end users can access data and have it be comprehensible. Data democratization begins with technologies which aren’t fancy or eye catching anymore, but are critical for business success: Data Warehousing and Business Intelligence.
Data Warehousing? Business Intelligence? We’ve been doing that for decades! Yes, we have.
How many organizations have done it well? That is a significantly smaller number. Warehousing source data is like the first step of our pyramid. To paraphrase the Gospel of Matthew, it’s the rock on which we must build our house. Data warehousing technology has never been in a better place with several competitors in the market who continue to push each other, making warehousing easier and cheaper than ever.
Business Intelligence represents the walls of our house, they show the shape of what we are building and allow end users to find shelter against the oncoming storm of necessary business questions. A good business intelligence implementation makes data consumable across the organization and encourages data consumers to investigate questions with that data.
When your accountant feels confident that the data they bring in from the warehouse is 100% accurate, that allows them to close books one day earlier each month, that builds confidence and trust.
When purchasing can get a list of its outstanding purchase orders and quickly coordinate with Accounts Payable to make sure an important invoice gets paid, you avoid supply chain issues and that builds confidence and trust.
When nurses who are exchanging shifts can quickly review which patients have special hygiene needs and ensure they get the extra attention they need, you not only have happier nurses who are more confident in their skills, lives are being saved.
When your favorite basketball team can tell the coaching staff that in their next game, the opposing team’s star player goes to their left 80% of the time with a couple days to prepare, the coaches can then gameplan appropriately and give themselves a better chance to win.
None of those examples involves artificial intelligence or machine learning. Business Intelligence is not just about pretty dashboards. It’s about access, data dictionaries, data governance, and yes, its about culture. When you have a great business intelligence implementation it pushes the organization to find answers to questions. The beauty of that cycle is that it doesn’t really end. You find the answer to one question, that leads you to another more important question that you may not have even thought to ask previously. Eventually that process leads to a terminal insight that changes your business.
That process of data democratization allows those insights to bubble up organically which means they’re also happening faster. Human intelligence when combined with business intelligence is still a key driver of intra-organizational innovation. When the organization rewards answering difficult questions, you get more people willing to ask and answer difficult questions.
This business intelligence implementation and growth step is far too often overlooked. They try to skip this important, but less sexy, part of both the technology chain and the cultural benefits it provides. These are both critical keys to ensuring that your future AI / ML implementations have the chance to be successful. However, what is also not discussed enough is that the technological steps you put in place to have a successful BI implementation are all massive accelerators for your AI / ML implementation!
You have a strong semantic layer with a great data dictionary, well defined table relationships, and data marts that functionally answer a lot of prescriptive questions already? Your data scientist is going to jump for joy. First, it gives a data scientist confidence that the organization they’re working for has worked through these key important steps. Second, those tools help them get models off the ground more quickly. It gives them a great starting point to understand how the business has looked at and categorized their data. Third, it also allows them to find where they can add value by not repeating previous work or finding work that they can enhance with their skill set.
For the fifteen plus years I’ve been involved in data analytics and data science, there’s always been our own version of the 80 / 20 rule. 80% of a data scientist’s work is about collecting, manipulating, and shaping data for cohort construction. 20% of a data scientist’s work is actually in model design.
In organizations that have strong BI implementations, you can double a data scientist’s value if you can get them to 60% of their work being in shaping data. Think about that, you don’t have to swap the 80 / 20 rule to get double the value. Part of why data science is hard is admittedly finding novel combinations of data that exist that the organization may not have considered combining previously. We’re never going to get rid of the data exploration part of their work, but if we can cut that by 20%, we are giving them twice as much time to design, test, and deploy those AI / ML models that everyone is so excited about.
This then allows us to get back into a cycle where those AI / ML insights get fed back into our warehouse and allow us to get predictive analytics that we can then integrate into business intelligence implementation. Now, the organization that has a culture of asking tough questions and investigating data can do that not just on prescriptive questions, but can ask questions like “If X is true and the model predicts Y result, what are the implications of that and how do we get ahead of it?”
Whatever you replace X and Y with, that question is going to be tough, but the more minds you put towards that question the better your business’s answers are going to be. When that storm comes for your business, your organization’s data practices were either built upon the rock or upon the sand. It takes a lot of time and effort to build upon the rock, but the benefits are obvious. You don’t see the flaws of building on the sand until it’s too late.
At Onebridge we have our MAP Assessment process which is a wonderful tool. We must assess technological readiness across several different dimensions. Often, we get brought in because an organization wants to talk to us about AI and ML and we are excited to do that work. We will go in, do the assessment, and come back with a simple statement, “your organization isn’t ready.”
When we do this organizations usually respond one of two ways. The first is the response we get most often and it's also the disappointing one. “Thank you for your time, we’ll move forward with someone else.” Those organizations years later, still haven’t made any progress.
Then we have the other type of organization, the one that looks at our assessment and studies our recommendations and says “Let’s build a plan to get there.” Those organizations are infinitely more successful, not just with their AI / ML end goal, but they get all those benefits that come with the work along the way.
Is it easy? No.
Are the CEOs, CTOs, and CDOs all happier with their business’s performance? Yes.
So next time when you’re going through your data technology alphabet, don’t sing your ABCs.
Sing your BCA’s.
Good Business Intelligence leads to a great data culture which gets answers to questions which will prepare you for Artificial Intelligence and Machine Learning. If AI and ML is the warp drive, we have to go through the necessary steps to prepare for using it effectively. A good BI implementation teaches you to walk, run, drive, and fly.
Once you can get to the sky, outer space doesn’t seem so far out of reach.
