Over the past two months, we talked about getting ready for AI, both for yourself and your co-workers. But what about the organization as a whole?
In the third installment of this series, I want to talk about what has to happen from a company point of view for AI to make a grand entrance and succeed.
It Starts with a Desire
I’m going to assume here that you’re not the CEO. As a result, you can’t just say, “Hey, we’re going to do AI” and have everyone snap to.
You need to start by getting some people in the company interested in AI. Frankly, that shouldn’t really be hard to do. There are articles everywhere that hawk the importance of AI, stating that it’s going to be the “it” thing in the future. Rather than getting pushback, it’s more likely you’ll almost have to hold people back.
Naturally, you want to make the C level of your company the main target of this sales cycle, but you don’t want to ignore everyone else. As we said last month, you can get some of the best and most practical ideas from people who are at a more “line” level. Of course, the C level has all the cash, so you probably want to get them interested early; you’ll want to start by canvassing people who are on the firing line.
I know, C-level people can have ideas too. The problem with their ideas is that they tend to be more strategic than tactical, and strategic ideas tend to be broader in scope and therefore more time-consuming (and expensive) to implement.
A Simple Desire if Possible
The hard, cold truth is that AI is not easy to implement. And now you know why I suck at sales.
Anyway, the web offers tons of examples of companies that turned to AI to solve complex problems. And some of them succeeded, at least partially. But in every case, they spent a boatload of money, took a ton of time, and were only able to keep going because they were lucky that the ultimate management group felt it was important.
It would be very hard to argue that AI is not still in its infancy. And I think it would be easy to argue that it might be 20 years before it really begins to mature (that’s my opinion, not something I have seen somewhere). And, in my opinion, at this time it makes more sense to do something small with AI, something that doesn’t cost much and won’t take two lifetimes to achieve. And these are the kinds of projects that many times lower-level people are more in tune with.
And Some Knowledge
Either way, big project or small, I want to go back to what I said last month (and the month before). People, even highly intelligent and technically trained people, have many misconceptions about what AI is and how it can be used to resolve problems that are common in your company.
No one needs to be trained as a data scientist (unless you are doing analytics), but you should have some sort of training to make sure people understand a few things.
The first is what the different components of AI are. I mean, it’s more than robots carrying laser guns and roaming through the corridors of a building looking for people who don’t have a hall pass. Robots might not even be involved, and if they are, they probably shouldn’t be given weapons, at least not until the beta test is over. Instead, the components could be chat bots, or analytics, or visual inspection, or some other facet of the AI revolution.
The second, and perhaps the most important, thing that you will need to stress is that AI is not something you implement overnight. It’s not a “one and done” kind of technology. Again, I encourage you to start slowly, with a small project that has a modest cash requirement and doesn’t change your existing systems much but can deliver a meaningful and measureable punch. The measurable part is important. You’ll need to be able to justify this thing by showing tangible benefits that couldn’t be obtained any other way.
Third, many AI solutions require extensive training to learn how to perform the task they are assigned and make decisions in an appropriate manner. It’s probably the most difficult part of the process. Never underestimate the time required, and don’t undersell that fact.
And a Plan
Any corporate plan is like Adrian Monk’s ability to solve crimes. It’s a gift and a curse. On one hand, it’s impossible for any large organization to commit to a goal and reach it without having a solid, detailed plan for how that’s going to happen. On the other hand, nothing seems to paralyze an organization and ensure that nothing really good will happen like a solid, detailed plan.
The difference, of course, is in just how important top management considers AI to be.
If this is just one more project standing in line with moving to Windows 10 or retiring obsolete hardware and software products, then it may not turn out to be the revolution that everyone was hoping it would be. There will be delays, redefinitions, and compromises that will reduce the effectiveness and perhaps even the distribution of this project.
Strong corporate sponsorship and involvement will be critical to the overall success of something that is potentially as transformational as AI, and that commitment must exist from day one.
And a Staff
Once commitment is established, a broad-based project team must be created—maybe not huge, at least to begin with, but broad-based. This is not just an IT initiative. IT must be included, but user involvement should be equally or more important. And top-level management needs to be in the team as well. Not as a steering committee or as “observers” but as actual participants with roles and responsibilities.
Perhaps the key member(s) of this project team will be the machine-learning trainers. As noted above, many AI solutions—for example, anything analyzing pictures or video looking for product defects, or anything dealing with text or human voices—will need extensive training on just how to interpret that data. And setting up and evaluating this machine learning is a specialty all its own. Whether you bring in someone from the outside to do this or have one of your own folks trained is up to you. What’s really important is that you do it and then take the time to do a proper job of letting them train the machine; don’t expect instant results.
The question then is, who should lead this effort? A top manager? An IT person? A user?
Truth is, there is no ideal title to put in charge of this, Instead, choose someone good. Forget about who they are, just pick someone good.
And by “good” I mean someone who will stay focused on the goal, will be detailed enough to verify that vital plan steps are properly carried out, and will have both ownership of and a passion for what AI can bring to the table. Again, it could be someone from outside the organization, and sometimes that works best in terms of minimizing turf wars.
In the End
In the end, you can’t do AI alone. You need some knowledge yourself, the support (and knowledge) of your co-workers, and the support of your company. And the latter is very important. AI can take a long time between the initial decision and the final results. And unless your company is fully invested, you will not get there.
Remember, AI is in its infancy. It will be a decade or two (in my opinion) before we’re talking to AI intelligences like they’re friends who are familiar with our idiosyncrasies, are able to recognize sarcasm, etc. But you can take the first steps today…and see some tangible, useful results.
MC Press Online