Replicants replace salespeople?
26th March 2018 | Professor Nick Lee
“More human than human”: The motto of the Tyrell Corporation, from the 1982 film Blade Runner. Recently watching that movie again, in preparation for a showing of the sequel Blade Runner 2049, I was struck (as I often am) by the interesting differences in what we can imagine the future to be like for different technologies. Of course, Blade Runner was set in 2019, just one year from now.
The film’s world contained flying cars, off-world human settlements, and virtually perfect human replicants; all things which we are many years from even being close to (despite Elon Musk’s dreams). Yet, views from within those flying cars show almost laughably simplistic computer graphical displays and control systems, the likes of which the phone in your pocket is now generations beyond.
In some ways, I feel our current understanding of the potential impact of artificial intelligence on the future world, and by extension the future of the sales profession, is similar. Many are nervous, even downright scared, of how AI might change our world for the worse, creating vast social problems as it makes thousands of different human job roles completely redundant, without creating new roles.
Many in sales and similar sorts of roles that I speak with seem to be primarily concerned about AI somehow replacing them physically, with stories on “robot sales assistants” and the like appearing in the technological and business press. While I think there is some legitimacy to those concerns of course, I suspect we are not that much closer to a genuine AI “replicant” today, than we were ten or 15 years ago. Indeed, while I feel that AI will fundamentally change a lot of roles in the sales profession, I suspect that the changes we imagine today will – like the flying cars and replicants of Blade Runner – be significantly different to how things will actually pan out in reality.
In order to be successful in the future, salespeople and sales managers need to understand exactly what this “thing” called “AI” actually is, and what it can do in the near future. I feel there are far more opportunities for smart sales professionals than there are potential costs – AI can be a massive enabler for the best sales forces, not something that will “replace” salespeople with robots.
So, if AI is not human replicants, what is it? Right now, what most people refer to as AI is actually just one aspect of it, something called “deep learning”. At its core, deep learning is simply a technique for finding patterns in large data sets. We feed in a large amount of inputs, and train the system to provide a correct answer.
This is best explained with examples, and one of the best is in the analysis of images. Deep learning methods are currently extremely popular for tasks involving the use of visual data to categorize images – for example as “cat or not cat”, or “cancerous or not cancerous” to take a medical example. For many reasons, deep learning methods significantly outperform humans at this kind of categorization task. However, this should not be misinterpreted as the AI “understanding” what the images actually are. This approach works best on problems where there are clear inputs and clear outputs, making them ideal for categorization.
Of course, if that sounds like a small thing, it really isn’t. In fact, many of our current jobs have quite significant components of pattern analysis and categorization, or at the least can be rethought in such a manner. Indeed, from one point of view, all of our actual interactions with others can be cleverly represented as information-processing, categorization and decision-making tasks. For example, our brains are constantly faced with multiple sensory inputs, which our brains categorize – at its most simple as a threat or not – and then decide on a course of action (eg approach or avoid).
Even our complex society today can be broken down into a similar sort of process. This is where the huge potential of deep learning comes from. However, AI in this form has significant weaknesses, and is certainly nowhere near what most of us think of when we use the word “intelligence”. Deep learning works well on specific tasks that it is trained for, but is dependent on the data used to train it, and is also poor at abstraction or inference – things humans are very good at. Obviously, technology will advance, but in most cases, we are far from the “replicant salesperson” right now.
Replacing sales-related jobs
The areas where AI is most likely to outright replace sales-related jobs are those where direct physical human contact is least important, and where tasks are most structured. In these kinds of environments, the tasks lend themselves ideally to deep learning capabilities, and our current lack of ability to create a convincing physical replication of the human form is unimportant.
Many sales and service call centre environments may find that AI agents can be used to service simple inbound and outbound call tasks. The trick will be in somehow separating out those calls most likely to be simple. In this sense, the AI agent will be in essence a real-time auditory chat-bot. Of course, many firms now use essentially the same technology to service online enquiries. The key to moving this to environments where voice is required is, of course, convincing voice synthesis technology – but this is not really an AI issue (just as the physical replication of humanity isn’t really an AI issue).
Another question is that of ethics: as representations become more convincingly “human”, will it be required that customers are notified when they are not speaking to a human? Will that then become a point of differentiation for some companies, as for example the call centre location is at present?
Enabler for complex sales
In terms of high-value professional or solution selling, however, AI as it is now is much more likely to be a massive enabler than a replacer of salespeople. Deep learning technology is already hugely useful in speech recognition and transcription, enabling for example much simpler and quicker interaction with CRM systems by salespeople. Not exactly what most people think of as “AI”, right?
Further, deep learning techniques are already in some firms helping match the right leads to the right salespeople, and even coaching the interaction in real time, using speech recognition technology. While this is currently most easily implemented in a telesales environment (eg in so many software-as-a-service environments), it is not hard to see how technology can enable real-time coaching of face-to-face interactions using small microphones, earpieces, and even video glasses.
Again though, we must wonder if there are privacy and disclosure issues of importance here, if sales reps are being fed information about their clients in real time, and suggested conversational strategies based on their client’s facial expression and the like. Even so, the potential is clear, and when one matches this to the stream of data available from wearable tech, we are far from exhausting the capabilities of even today’s technology, let alone what may occur in the future.
Sales manager role
Indeed, it is the role of the sales manager that may actually see the most change in the future. Recently, with the rise of real-time AI coaching technology (based on analysis of successful sales calls, and suchlike), and the clear potential for deep learning to do many of the “technical” aspects of a sales manager’s role, I have begun to wonder just what sales managers will be needed for in the future. This to me is one of the key challenges for sales professionals in the future.
The most valuable sales managers must offer something that AI cannot. Right now, a lot of that has to be about the “softer” skills of motivation, and understanding the core drivers, learning how to get the best out of them. The areas of the role that will be least replaceable by AI are those that require abstract thought, flexibility of connecting different inputs to make new outputs, and reasoning/decision making with incomplete data.
In fact, the large amounts of data needed to train deep learning algorithms may be a key limiting factor in some sales contexts – especially face-to-face solution selling as opposed to telesales and online sales where large data sets are often automatically created. Once we break a task down into input data and decision rules, and can collect big enough data sets to train the algorithms, deep learning is ideally suited to do this better.
The task for all of us is look at what we do, decide which parts of it can be outsourced to an AI to make us more effective at the other – more valuable – parts. In that way, we can help AI work for us, not replace us.