AI is real this time. It might be a toy. You might have gotten bored of Prisma already. But it's there. But there is also a tendency to paint everything with the broad brush of AI. There is a new category of companies like 6sense, Lattice Engines, PipeCandy, Infer etc. that are loosely clubbed under the category of 'Sales Intelligence'. Some of them have kicked themselves upstairs by calling themselves AI-driven sales platforms.
Be that as it may, has AI really arrived at the sales world?
There is a saying in the world of AI - Don't model the world; Model the mind. When you model the world, which is a very hard undertaking that deserves more than pat on the back, "you" are still modeling. In other words, "you" are still programming. So it's a program, at the end of the day - a very complex program that is indeed a marvel, but a program nonetheless.
A step up is "Machine Learning", where the machine has to learn to update the models the programmer has built in response to the real world signals it observes whereas if you really are doing AI, you are modeling the mind. You give your program a mind of its own and it learns to build models of the world it observes.
For example, we can build a program to track executive hiring in companies as a leading indicator of upcoming procurement events. This is squarely in the domain of programming. On the other hand, we can also pick algorithms that find underlying patterns in historic data on company funding, hiring, and other parameters that impact procurement. The latter is Machine Learning. But then these two are not AI.
You might have heard of this term 'Deep Learning' thrown into good measure when we talk about Artificial Intelligence. However, true artificial intelligence has to be intelligence that develops on its own (as against programmed). In other words, it is what we can call as "unsupervised deep learning".
Staying with the above example, AI-ification of sales means that we should be building programs that not only analyze large data sets to detect historical patterns, but are also intelligent enough to quickly adapt to new data and emerging patterns and evaluate their impact using a set of predictive frameworks. In essence, these are programs that "build" predictive models on their own.
But unsupervised deep learning in the area of sales is a hard problem. Even in highly rational buying situations like in the case of the B2B world, the sales reps and the buyers are conditioned by their earlier experiences. Biases and stereotypes greatly impact decisions. It's hard to find data to detect these stereotypes, let alone model for their interplay.
So, the pragmatic expectation to have is to expect AI to mature in the sales domain to the extent of how AI has evolved to beat a human in chess. In other words, in the narrow domain of sales, AI can evolve to do better job than a sales rep in analyzing narrowly scoped business information and identifying potential opportunities.
For instance, in our own experience, we've seen analytical models recommending retail businesses to retail technology sellers based on esoteric parameters such as geo-location, census and socio-economic information, neighboring businesses - an area where only highly experienced sales reps can intuitively connect the dots.
A technology buying propensity model that we recently developed based on observing prospect company's website helps deliver 5x better campaign responses from prospects (for an A/B testing software company) based on modeling the impact of various parameters like relative traffic growth, funding deployment, technology clusters used, TLD, intent of the website (to convert vs. to inform). It's entirely possible for the model to improve itself based on the prospect responses (and even then, it is still machine learning).
In spite of these promising applications of machine learning techniques in sales intelligence, it is still very much "program the world" kind of AI that then gets improved and not the "program the mind" AI which would be needed to improve sales performance across a variety of industries. In other words, the models have narrow applicability for specific industry domains (based on objective sales nuances of those domains) and do not do well when you take it to other domains or go deep into the realm of subjectivity introduced by human experiences.
So if your sales intelligence vendor claims to be AI-driven check how narrowly they apply AI to help improve your prospecting capabilities. If they can't provide evidences they are giving you a very clever program (which is not bad) but it may not be AI as yet.