When LinkedIn suggested today I connect with Richard I was surprised. You see I know Richard quite well (not his real name) or, should I say, I knew him. Sadly, he passed away a few years ago.
LinkedIn uses very sophisticated (some would say creepy – read this) algorithms in its connections recommendations engine. It tracks recent activity, looking for people you know who know people you are not connected with, reaching out beyond its own extensive network, comparing education background and work history, all to build more connections. But yet, it can’t look beyond the data it has access to and the algorithm doesn’t know when someone is dead. It does not reach into death records and match that to its members. So it continues to offer connections to dead people. This is a simple example of data science gone awry – it is not endowed with common sense.
On a broader scale, there is a seemingly unfettered pursuit of magical AI-based solutions to automate tasks that heretofore required humans to do the critical thinking. As Kevin Kelly from Wired says “There is almost nothing we can think of that cannot be made new, different, or interesting by infusing it with some extra IQ. In fact, the business plans of the next 10,000 startups are easy to forecast: Take X and add AI.”
This worries me, not because I don’t believe in the value of AI – my first company, founded in 1986 was an AI company, and the first book I had published (in 1988) was called Expert Systems Introduced – I believe AI principles provide extraordinary value when applied in a specific context, and when the AI engine is infused with deep knowledge of the domain to which it is being applied. But I fear these facets are not being considered and we are not all learning from the mistakes of the past.
Instead of adding (valuable) AI, we are sometimes just adding automation and (at best) applying light-weight algorithms that help us screw up faster. There are just too many examples, from the trivial to the extreme, that demonstrate that dumb algorithms that promise valuable outcomes are falling short. Common sense is not something that lends itself to representation in an algorithm.
In 2014, Joan Terry a 71 widow from Staffordshire in the UK (pictured above) received a letter from a car dealership for her late husband – addressed to ‘Mr Deceased’ offering a £250 discount.
In 2016, an experimental Microsoft AI-based experiment called Tay, based on a teenage girl’s profile, quickly degenerated into ‘a Hitler-loving sex robot within 24 hours’ according to The Daily Telegraph. Tay was also seen denying the Holocaust and insulting African-Americans and Mexicans.
When the dating-simulation app Invisible Boyfriend was first released, many users assumed that they were talking to a computer program. Billed as ‘Someone you need to talk to when you need it’, the AI component behind IB was so deficient that it was later revealed that the boyfriends were not bots, but real people who worked for the freelancing company Crowdsource.
While some companies present their offerings as Artificial Intelligence solutions, Amazon, with its Mechanical Turk project, makes a virtue of artificial AI, using humans to do the work that computers are not necessarily good at. “From a philosophical perspective, it’s really turning the traditional computing paradigm on its head,” says Adam Selipsky, of Amazon Web Services. “Usually people get help from computers to do tasks. In this case, it is computers getting help from people to do tasks.” As Tim O’Reilly, a computer book publisher and tech industry figure, puts it on his blog, old dreams of artificial intelligence are “being replaced by this new model, in which we are creating more intelligent systems by using humans as components of the application.”
As Amazon says on the Mechanical Turk website “Amazon Mechanical Turk is based on the idea that there are still many things that human beings can do much more effectively than computers”.
I fully understand the value that AI technologies solve – I was granted a patent in 1995 that used pattern matching to generate graphical user interfaces on top of mini- and mainframe computer terminals – but I also know the limitations. In most cases, and in all cases where critical thinking is required, like enterprise sales or any knowledge worker domain, the problem is not the deficit or data, but the deficit of insight.
Think about this: If pattern matching alone were utilized to predict the 44th or 45th holder of the office of the President of the United States, there is no way that it would have contemplated a non-caucasian, a woman or a TV celebrity as a potential match. All prior data would have suggested a white male coming from a political background – with a near 100% degree of certainty.
I will come back to the problem of the dumb algorithm in the context of sales in just a little bit, but let me first recount the tale of Doug Lenat the founder of Lucid.ai and the creator of Cyc.
Lenat has spent the last 31 years getting Cyc to memorize an astonishing collection of general knowledge. It stores lots of knowledge that you would not find in Wikipedia or in an encyclopedia because to humans this knowledge is seemingly self-evident. It understands the evidence of Newton’s law – that if you drop an apple it will fall to the ground. But it also knows that an apple is smaller than a person, it might be a computer company that seems to defy gravity, and it also understands that because of gravity a person could not throw an apple into space.
While Cyc is built as a broad capability, Lenat points out that in each domain to which Cyc is applied, the AI faces a learning curve. “We interview subject-matter experts and also peruse documentation. We ingest that knowledge into Cyc much like you would with a human.” In my book Expert Systems Introduced I called that activity ‘knowledge engineering’, a term not used a lot today.
As an aside; in more AI projects now, companies look to employ data scientists – where in situations where critical thinking matters (like enterprise selling), their efforts would be much more productive if they hired knowledge engineers.
In a video where he introduces Cyc, Lenat goes on to say…
“The only way for a computer to learn like an expert is to have their background knowledge and the ability to reason and apply that knowledge.”
Unquestionably, there is value delivered to sales people (and other knowledge workers) from data science, analytics, and AI solutions. These technologies can uncover more meaningful signals for the seller to act upon. Using the Salesforce acquisitions as a proxy for the market I will take two examples.
- Salesforce Wave can churn through data and separate some of the meaningful signals from the noise in the data and present these signals to the user. It might show me a chart that demonstrates trends in the marketplace that I never saw before. This can help me make better decisions – if I have the critical faculties, the knowledge, experience and background to interpret the chart. It does not yet answer the “So what, what do I do about it?” question.
- Salesforce IQ will automatically capture and synch data from your inbound and outbound emails, your calendar, and your smartphone calls. It can mine the activity in email exchanges to tell you that Joe Smith never responded to your email, but it does not understand Joe’s business problems, or the right strategy to sell to Joe based on his buying decision criteria, or considering the other competitors in the deal. It does not understand the impact of Joe’s non-responsiveness on your sales strategy. It does not yet answer the “So what?” question because it does not yet have the knowledge of sales strategy in this specific context.
The best way I can explain this problem and how to solve it is to talk about Google Maps. I’m guessing that each and every reader of this blog used Google Maps (or Apple Maps) very recently. Not many of us think about how it works – and that’s how it should be.
Google did not start by tracking all of the routes every one took, storing all of the data and then turning to pattern matching to figure out the best way to get from A to B. If it did, it would have to wait until every possible route in the world was traversed at least once to come up with a set of directions.
I can’t imagine that anyone has ever walked from Cork in Ireland (where I am right now) to South Africa (where my wife is going on a walking tour later this year) but when I asked Google it came up with a good set of directions.
BTW it would take 2,398 hours to do that walk!
Think about what happened the last time you used Google Maps. To provide a foundation for its routing algorithms, Google first licensed all of the maps of the world from a company called TeleAtlas. It baked that knowledge into the system, so that when you provided the data (your desired destination), it could figure out the context (your starting location) and then analyze the best route to take you to where you want to go.
So it is with enterprise sales. You need to start with a foundation of knowledge – in our case we baked 30 years of sales methodology and over 1,200,000 of sales coaching into Altify Max – so that you have a set of maps to start from. Then the systems need the context of where you are in a deal and needs to understand your status in an account to know your starting point. Coupled with data in the CRM, and other key insight signals, you can then begin to use the power of the technology to analyze the situation and answer the “So what?” question by assessing the impact of change and prescribing a set of actions to move a deal forward.
[We call this approach Augmented Intelligence and as far as we know Altify Max is the only Augmented Intelligence Platform for Sales.]
Coincidentally, I just read Marc Benioff’s foreword to the recent McKinsey publication Sales Growth: Five Proven Strategies from the World’s Sales Leaders. As you would expect he spoke about the cloud, social, mobile, data science and IoT. Then he goes on to say this:
“This technology shift is more profound that anything we have seen, and is transforming the way we sell … but that does not mean that technology replaces sales … This is not the time to diminish sales; it’s a time to reassess and reinvent it …
By managing and motivating sales to develop trusted relationships with customers, and by using tools to make the process more transparent, collaborative, and strategic, companies can deliver sustained and consistent growth.”
For me that is almost a manifesto for Augmented Intelligence for Sales, putting the customer at the center and leveraging technology that goes beyond pattern matching and machine learning – in fact augmenting these technologies with a foundation of knowledge and an ability to reason based on insight.
When automation is applied to makes things easier, without making things better, silly things happen: I get recommendations to connect on LinkedIn with someone who is no longer with us, and Joan gets a ‘Dear Deceased’ letter.
Where critical thinking is required a foundation of knowledge is essential.
At Altify, we are excited about all of the AI, and related, capabilities that Salesforce is bringing to market. Because we deliver our solutions native on the Salesforce platform, our customers can leverage the enriched and expanded signals that Salesforce delivers and combine those with the foundation of knowledge and the Augmented Intelligence Platform that we provide with Altify Max.
Imagine having deep muscle memory from a million sales engagements, knowledge of the world’s best sales methodologies, and the insights from your own business, all coming together to guide what to do next to progress a sale, to trigger the next action, to increase the seller’s knowledge and achieve increased sales results, every day.
Now you no longer have to imagine.
Our experience with smart systems —undertaking the deep science, learning what these systems can actually do, and working every day to apply these capabilities in the world—has taught us that augmented intelligence technology does not replace, but rather enhances, human capabilities.
Instead of “artificial” intelligence, the real-world work of business is augmented intelligence – and its benefits for knowledge workers will be extraordinary.
If you want to read more of my blogs please subscribe to Think for a Living blog. Follow me on Twitter or connect on LinkedIn. I want you to agree or disagree with me, but most of all: I want you to bring passion to the conversation.
Donal Daly is Executive Chairman of Altify having founded the company in 2005. He is author of numerous books and ebooks including the Amazon #1 Best-sellers Account Planning in Salesforce and Tomorrow | Today: How AI Impacts How We Work, Live, and Think. Altify is Donal’s fifth global business enterprise.