What started with maverick software developers has actually become, during the last few years, a growing reality. Software systems can now be implemented to function in lieu of humans. While this sounds great (or not) on its face, there are a few key reasons why AI—that is, Artificial Intelligence, won’t be taking over your job any time soon.
Well, maybe it already has. If you are a checkout clerk, you may have witnessed an exponential growth in customer self-service stations at your workplace in recent months. That is not what I am talking about. When I reference AI, I am specifically talking about phone interface systems which function on behalf of a user or a business.
Having made that clear, let’s take a look at some of the systems that are already out there. Google Duplex is perhaps the most well known, simply because it’s Google. In one video, Sundar Pichai gives a jaw-dropping demonstration of the system showing what it can do for customers and businesses.. although he seems to admit that some calls don’t go as planned.
In another demo available on YouTube, Gridspace shows us what they can do with AI technology to create a seamless customer experience. While we are prepared to stand and applaud at the way everything ties together, again it becomes obvious that the representations are a bit fanciful.
Of course, we know that demos are meant to sell product. We get that. You always put your best foot forward when you are showing a demo; and whatever problems or glitches may lie back of your system are usually not discussed at that initial stage of presentation.
But there are reasons why companies are still very hesitant in investing in these systems. And it’s not simply the high cost of implementation. If you are a B2B or B2C company, you know that contact between your company and its clients and/or customers is simply one phase of its day-to-day business paradigm. When we consider someone calling in to purchase a product or request a service, we must also reckon on things such as data capture and data entry. And that’s where AI runs into kinks.
Not to even speak of things like data security (that is a totally different discussion), we all know how essential it is for the information tendered by the end-user/requester to be accurately captured by a system which purports to replace a live user. The videos make it look so easy. But we suspect that developers are not telling the entire story.
If I have to provide information relative to a transaction, what guarantee is there that the information will be correctly captured? After all, it’s just me and a computer program, and no one else is watching. Siri still doesn’t understand me when I tell her to turn off my flashlight. Are we really there yet?
Also, have we fully fleshed out issues like data entry? It is one thing to accurately capture the information. But how exactly will the system transcribe that information to third party software systems so that it can be internalized? AI developers must face the fact that virtually every organization uses multiple software systems to record and store data. Will AI cooperate efficiently with those systems? If so, how??
Suppose that the B2C company is using Oracle to store information regarding customers’ orders, and Salesforce to store internal notes. Suppose there is yet another program needed for tracking loyalty points. Can the AI software access all these systems simultaneously, and do all that is needed in one phone session, without any IT hiccups? Remember, there may be calls holding.
Until software developers take it to the next level, human users will be needed to transcribe data, because systems are still unable to do all that a live user is accustomed to do. They can’t, and we know they can’t. Even in a hybrid environment which used only some AI, data would still need to be gathered from the actual calls themselves, if only for quality assurance purposes; i.e., to make sure that AI was doing its job correctly. Hence, you can keep your call centers busy. If I have to listen to the same call to get the information, I may as well just take the call.
One solution, of course, would be to build “all-in-one” systems developed not only to take calls, but capture and store data as well. That would obviate the difficulty of dealing with multiple programs, and bring everything related to the transaction under one banner. Due to the nature of things, there will always be some amount of human interaction needed, because problems occur. Systems fail, data must be retrieved, glitches must be worked out, and what have you. But comprehensive software systems would at least float AI out of the novelty phase and into the mainstream.
Naturally, the more complex a company’s business operations, the more distant are its prospects of accepting AI as an alternative solution to an ‘analog’ work environment. Will there be storm signals before actual implementation? Of course. From the business side of things, company policies and operations would probably have to streamline, if not simplify, in order to accommodate themselves to the software. Thus, you would likely see a ‘dumbing down’ of the organization’s internal processes preparatory to any wholesale (or retail) AI invasion.
The conclusion, then, is clear. AI is still a way away. It’s in the offing. But it’s not right around the corner. Andrew Yang is still ahead of his time. Despite all the hype, I believe that much of the AI talk equates to “The Martians are coming!”.. because, after all, when a company develops a software system, they need to get it out there. Their business is not to present a realistic, but an ideal, scenario. And that is where we must stop and bring the conversation back down to earth. While we, too, look forward to the revolutionary changes that AI will bring, we must admit that, currently, HAL is still somewhere in space.