How to introduce AI and Data Science in your organization? A CxO guide

It’s easier than you think

Once upon a time making changes in an organization was like pulling teeth in the days before anesthetics. Sometimes it was necessary; everybody knew it was going to hurt like hell; people got it done anyway because, if they didn’t, an abscess could mean literal death.
Dwelling in the past, being reluctant to embrace the present (sorry dear reader, but it is no longer “the future”) is a costly error. Artificial Intelligence (AI) and Data Science (DS) are now facts of everyday life and business. Most of your competitors have already made the leap in one way or another, so by dawdling, or being throttled by co-workers that deliberately demand studies and reports to thwart the changes, you’re letting more progressive companies gain a significant lead on you.
Maybe you’re worried that you’ll meet a lot of resistance from workers when confronted by the changes. The good news is that unlike in past decades, we now have Human Resources (HR) as an intermediary, and they are experts at interfacing with all the company staff. It’s not like Henry Ford rolling up his sleeves and “getting greasy with the boys” to solve problems. You have the luxury of making the decision and then turning it over to the experts to implement.

The Hard Sell

But let’s say you’re the CIO (Chief Information Officer) and you know right down to the soles of your feet that this is the right plan. How can you “sell” the idea to the other C-Suite execs from which you need support?
How are you going to sell it to all the CSRs (Customer Service Representatives) that are fielding phone calls; the advertising department who have always done it “their” way; and even the Data Scientists who continuously comb through your Big Data, looking for relationships and opportunities to exploit discoveries, and from which they derive their greatest job satisfaction?
These people are heavily invested in doing their jobs as best they can, the way they have always done them, and any change can be threatening. Fortunately, it isn’t as difficult as suggested—in fact, it’s rather easy…

The Easy Sell

Let’s examine the example of a telephone support worker who answers one call after the other, often with the same questions over and over again. They have 450 minutes per shift, and answer between 80-100 calls per day. That means they have between 4 and 6 minutes per caller to solve their problems, answer their questions, or redirect them to a higher authority.
Pressure, pressure, pressure! Aside from the fact that there is always someone in the queue, that next caller has been endlessly assured that “your call is important to us, and you will be connected to the next available operator” for many minutes so that they may be in a bad mood, too.

Now try telling that operator that an AI-driven phone answering system, or a text-driven ChatBot, will take Inform them and that it is capable of answering the most common questions. Tell them that the current onslaught of phone calls will diminish to a trickle so that they can now turn their creative abilities to answering the questions that are beyond the capabilities of the machine.

Sometimes solutions need a creative touch that is beyond an AI’s ability. If workers now need to spend 15 minutes solving a problem with a delighted customer that got to speak to a human almost instantly, they’re going to feel better about their job. The customer will be thrilled, and the phone worker wasn’t stuck providing a “stock answer”; they got a chance to be creative and craft a lasting solution.

A Universal Solution

We can say the same about the Sales Manager who spends most of his or her day tracking what all the salespeople are doing. Who sold what? When can production be scheduled? Does the customer meet the credit requirements or have the resources to pay for this order?
All of this information is amenable to being collated by an AI. That manager no longer needs to spend four hours of the day writing reports and entering them in a computer/database. The AI can manage all that.
Now that Sales Manager has four more hours to devote to questions like “I’ve got three hours of unscheduled time on Production Machine #6… Hmmm, can the ABC Manufacturing order be increased by 20% if we offer them an 8% discount? We know they’re going to need it…”
Meanwhile, the Production Manager, experiencing the same relief from mindless paperwork is thinking: “Line #5 is scheduled for maintenance in two weeks, but I have a hole in the schedule now. Can it be done early, so the machine is in good shape for the busy season?”

HR benefits from AI, too, since an AI can automatically create a customized onboarding experience for a new employee. It can program a laptop with the tools needed to do their job; it can schedule office space near people & resources that will be necessary; it can monitor performance and offer skill upgrading in areas where performance is lower than expected.

Every single employee can benefit from “meeting scheduling” handled by an AI. Just tell the AI you need an appointment with someone and it will find something convenient for both parties (or a dozen people for that matter).

Data Scientists

They may not actually wear white lab coats, but your Data Scientists have ferocious intellects that like to design algorithms to cull fascinating tidbits of data to discover relationships. It might be represented by a query that says “find every instance where condition ‘A’ exists with condition ‘B’ and create a list showing all prevailing circumstances that co-occurred.”

Then they will painstakingly go over that list and manually discover some relationships. There is no possibility that they will recognize all of the connections—the human brain just isn’t built that way.

Why AI is Better

One of the best examples of this occurs in medical imaging. More and more analysis of medical images is being turned over to AIs. The reason for this is that on an MRI, PET scan, or chest X-ray, a radiologist or physician might look at the image and not notice several faint shadow-like anomalies that are distributed far apart.
They’re faint; they’re separate; they’re almost indiscernible, but each has a similar, distinctive shape. Because one appears at the very top of the image, and the others are at the bottom or left and right sides, they’re too far away for a human eye-brain combination to make the connection.
AIs have greatly enhanced the early detection rate for clots, fibroids, malignancies, hyper-vascularization, or vascular atrophy. When you have the immense computing power of an AI, it can look at the whole image and examine each part, pixel by pixel, to find relationships. And this occurs in just a split second—faster than a human—and in more detail than we are capable of!

Selling it to the C-Suite

There are only two types of businesses: those that sell a product, and those that sell a service. It should be self-evident that companies which sell a service stand to gain a great deal from Artificial Intelligence and Data Science. Those that sell products stand to make immense gains as well but you might have to look a little closer to figure out how it makes a difference.
Service Providers
AI and DS will provide you with the opportunity to update your knowledge about your customers’ needs constantly. Did your client open a brand new office in another city? That’s an opportunity to expand the service that you offer to this client. But how did you come by this information?

AIs using Deep Learning (DL) can monitor social media, conventional media, and business news of every possible type. Computer-vision now allows an AI to analyze video and images to extract relevant information and make extrapolations. As we all know, the more often it does it, the better it gets at the task.
Consider the airline JetBlue, and how they continuously monitor social media such as Twitter, responding gratefully to positive comments, and immediately tasking a human being to deal with a negative remark.

If you see “I hate JetBlue! My flight is delayed, and now I’m stuck in Seattle airport for the night!” Within 15 minutes you’ll also see “Wow, Mary Smith from JetBlue just contacted me and gave me meal vouchers and a hotel room reservation because my flight was delayed. She said she has already arranged a new flight for tomorrow. I love JetBlue! I’ll never fly any other airline”.
Who wouldn’t want that level of loyalty and favorable public opinion? More and more companies are hopping on that bandwagon because they see the ROI (Return on Investment) of a positive public image. You can advertise until you’re blue in the face and only show a modest increase in business for all your work. If you get one real human being to extoll your virtues to friends, family, and associates, you will see a massive return for the most trivial investment.

The problem is that it takes AI, ML, and DS in combination to make that data available in the first place so that you can respond to it. Without those systems in place, you are fumbling around in the dark. Your rivals will walk all over you with their responsiveness to their customers.

Product Providers

C-suite executives always want to know about ROI. Stockholders are still going to demand that they explain investments and how that is going to earn the stockholders more money. ROI has become such a buzzword that just using it in a sentence makes people happy for no discernible reason.

When you try to explain to an investor that you spent five million dollars on an AI upgrade to your plant that prevents downtime, they would immediately want to know how much money that will make. It’s hard to explain to people that avoiding downtime doesn’t create new income.

What it does do is prevent the loss of existing income. When they ask “If you didn’t spend that money, would we still make the same profits?” the answer has to be “maybe” because theoretically, the production might not have experienced a failure. Eventually, unregulated, there will be a failure, and there will be a significant financial loss. It’s part of life, and it is inevitable.

But a modern IIoT (Industrial Internet of Things) system can report small problems that will lead to significant disasters long before they become actual problems. The philosophy of “If it ain’t broke, don’t fix it” is just profoundly stupid in this modern day and age. Asset failures can destroy a business in much the same way (as discussed earlier) that ignoring an abscessed tooth can actually kill you.
Who doesn’t remember beloved entertainer Jim Henson’s “chest cold” that he decided to battle through rather than visit a doctor? He could have found that he had pneumonia that could have been treated before it became fatal.

We need to learn to speak to investors in a way they can understand. We need to be able to explain that assuring income is just as important as discovering new income. Both tasks are made easier when you have the right tools at your disposal.

AI and DS Make Money

The key to success in this area is that you cannot jump in with both feet and no direction in mind. AI and DS are not some magical sauce that you can pour on your business that will suddenly make it immensely profitable. Companies that decide to invest are hopefully not so naïve that they go to the IT department and can say “Here’s half-a-billion dollars. Make sure our bottom line improves by 25% next quarter”, because that will never, ever happen.
The lack of a real plan is the primary reason that AI implementations fail. Most experts agree that small pilot programs can win over the C-Suite executives with POCs (Proof of Concept) that have solid, substantive results. The most important aspect is that these POCs must be easily perceived as scalable.
Data Scientists are Expensive
At a starting salary of $100,000, we’re already looking at $1,000,000 a year for a basic group of 10 Data Scientists. Depending on the size of your company that could be 10 or 20 times larger.
For most smaller companies that kind of investment is out of reach anyway, yet companies of all sizes are using AI, ML, and DS quite successfully. How?
Small Necessity—Giant Strategy
An Enterprise can use a small business strategy quite effectively. Instead of setting up your software and hardware AI solution, which would be a significant investment, maybe you should be thinking like a small business and investigate AIaaS (Artificial Intelligence as a Service).

Maybe you’ll use WATSON services from IBM; perhaps you’ll look at ALPHA GO from Google; there are innumerable companies that are now providing AIaaS for commercial consumption. You’ll see familiar names like AMAZON, or eBay, but smaller more agile companies will be interspersed amongst the giants.
Who to Choose?
If you want a full-service package, including Big Data analysis by their resident data scientists (saving you all those salaries), the bigger guys are probably a good choice. If you have your DS team and need the AI resources, one of the smaller guys might be a good money saver. Using several companies for different POCs can also help you develop insight into which provider you might like to use on a full-time basis.

Late to the Party

If your company is still without a valid AI/DS strategy, there will be an urge to “rush” and try to “catch up” to the competition. You must resist this. You can take ambitious steps, more substantial steps than you might have five years ago, but remember you are still in POC-territory and have to develop systems that work before you attempt to go full scale. If you can’t produce a good business case for a POC, then it’s nothing more than a waste of money. You need understanding, funding, and a valid business case, or you should not start the project.
CIO vs. CDO

Data will transform the way you do business in the future, of that there is no doubt. The CIO and CDO can cooperate, but most experts agree that one will vanish in favor of the other in various organizations within the next decade. The CIO’s function remains the same, making sure the organization has the info it needs to function effectively.

The CDO (Chief Data/Digital Officer) has the responsibility of making sure that digital information about customers is secure; that digital information is handled in a way to protect it from abuse. S/he is also responsible for the utilization of internal data and how to exploit it to enhance the company’s value or abilities. The CDO must also control, as much as possible, information about the company held by outside third parties, whether it is levels of government, banking institutions, partners, or clients.
While the CIO is focused on making sure the organization has the computer support and information systems it needs to realize its goals, s/he must also inspire the IT staff and balance the IT budget. They must retain a healthy relationship with the CFO (Financial), CMO (Marketing), and of course the CEO. The most successful companies have strong relationships with these four people.

As you can imagine the CDO and CIO can have a lot of overlap. Without a good working relationship, this can lead to competition or confrontation so both need to agree on how data will be curated. In all likelihood, eventually, the roles will be blended so that you can look forward to a new name CITDO, or something similar.

However, it ends up, one or both of these people will be responsible for the AI, ML, and DS infrastructure, development, and exploitation. The problem either or both will face is compartmentalization, or as it was called in the old days, the “Silo Model.”

Data scientists are tasked with creating models to test new ideas. When the project is completed the techniques are simply saved to a file, eliminating the possibility that they can find new life in a different task.
This is not an isolated problem because it occurs across all enterprises. Information is not correctly curated. Instead of taking these good, tested, successful ideas out of circulation, there needs to be a central DS repository, which is also accessible to the AI, so that we’re not always reinventing the wheel.

The Takeaway

Where does all that leave us? In 2017 only 45% of CEOs were involved with AI (24% using, 21% instituting) but the remaining 55% were only “aware” of it, or evaluating it to see if it was useful. The trend seems to be that small companies are gung-ho about AI, ML, and DS, and are taking advantage of it, whereas big companies, suffering from massive resistance and lack of agility, are dragging their heels.
Why does this happen? It’s the old “too many cooks spoil the broth” problem. There are too many people working at cross-purposes in big organizations, and no one seems firmly in-charge of AI on behalf of the company. The AI “ship” can race its powerful engine, churning the water into foam, but if it lacks a rudder, it will never get anywhere useful.
Everybody seems to have their budget for AI—marketing, finance, product development, sales, CSR, HR—and it’s all outside the purview of the CDO or CIO… Someone needs to step up and steer the ship!
Once you have identified that person, they need to sort through the available projects, determine which are most likely to benefit from automation, and get them rolling. What will change company revenues the most and hence get support from the C-Suite? What will improve the work-life of people the best by eliminating work that they hate? What processes would become more accurate if they had less human interaction?
Once you identify those targets, the rest becomes obvious… Once you have “moved the needle” and shown some positive results, POC or otherwise, you can continue with other projects. They may be less dramatic but are still useful. Just make sure you start was something that proves the value!