IBM's Watson and DeepQA Software
In the early days of our technology’s development, machines helped us to overcome a human limitation. Tools made our fingers as strong as steel, or enabled our eyes to see in the dark, for example. But today the machines we make push this much further: they replace the human’s involvement in the task entirely.
IBM’s DeepQA software, for example, the brains behind its Watson products, beat Garry Kasparov in 1997 to become the new Grand Master of chess. In 2011, Watson went on to beat its human competitors and win Jeopardy. And since then, the folks at IBM have not been sitting around idly chatting to Watson. Their latest DeepQA-based product is Watson the Super Sales Rep. They exchanged the Jeopardy database in DeepQA for one with data on Watson, DeepQA and IBM Hardware. Watson can now sell itself to prospects wishing to program it to sell their own solution to their own prospects.
It makes sense. Watson is just too complex for one person to fully comprehend. The team setting-up Watson to sell a company’s solution, requires database administrators, and hardware, software, domain, and knowledge-based system experts. No need to fear for your job just yet: the hardware for the Super Sales Rep is reputed to cost more than $32 million. After you add in the no-doubt expensive DeepQA software and pay that aforementioned team to do its thing, you’ll have spent more on your sales force than most companies spend on theirs, ever.
But the trend in our Brave New World is clear. The path of progress is inexorable – it flattens objections and ruthlessly obsoletes older technology. We’re on a journey to human obsolescence and it’s too late to change direction now.
So, dear reader, what do you think?
Does this conjure visions of Terminator, or perhaps a world like Rollberball where people play instead of work?
I ask because this is the dilemma we face in Gossamar every day. While we haven’t yet reached the point of automating a person, we constantly face the question of which parts of your sales and marketing process lend themselves to automation and which parts should we keep for the human team members?
Seems to me you can divide the activities you do to market and sell your solutions into three categories concerning this problem:
1) Obvious tasks to automate
2) Obvious tasks to assign to people
3) Tasks that are more difficult to decide.
Obvious tasks to automate are of course routine ones. Tasks which are done the same way each time using the same resources and the same techniques. Fielding a request for information, for example. Sending out an email blast is another good one.
Obvious tasks for people rather than machines are those which involve a conversation. Yes, despite the capabilities of Watson the Sales Rep, not many companies can afford to implement such a system and for now we’re all going to keep person-to-person interactions going and not transform them into person-to-machine conversations.
Other obvious candidates for people are activities which cannot be anticipated in advance. For example, a prospect (or client to elevate the situation a little) complains about your company or its solution on Twitter. You need to respond. While automation can help you know about the complaint and locate it for you, you will need to read it and act on it in a way which you sincerely hope is never going to become routine!
As always, though, the difficulty lies in that 3rd group of activities which are not that easy to determine. I suggest that these activities will most frequently be both ones which lend themselves to being automated and to being handled by a person.
For example, qualifying prospects. With automatic scoring of prospects based on their activities on your website, and with the automatic grading of prospects based on their answers to your Profile questions, the machine does this in an objective and uniform way. But when a rep gets a lead which the Inbound Marketing Automation system thinks is “sales ready”, he or she is going to call the prospect and try to qualify him or her even further. At some point the rep may decide that while the prospect does indeed have a great score and grade, they are not in fact a good prospect. Ideally, if this happens, the rep will have a word with the marketing team so that the grading and scoring setup can be adjusted to downgrade this type of prospect.
How about developing more insight into your prospect’s buying behavior? Again, your marketing automation software will gather the data for you automatically. Digital footprints to show you where the prospect went on site, what she read as opposed to merely glanced at, and what he downloaded. But the analysis of the data is your domain. Interpreting it all takes a 50 thousand foot view with the ability to drill into any detail which catches your eye. One of the things which humans excel at, given the right tools and data.
After reading what I had written here, I suddenly realized that there is a 4th category of automation versus people: automated routines which can be taken over by a human at will. We try to design an Inbound Marketing Automation system to do things on its own entirely automatically. But there are always going to be situations where a person believes he or she can do better than the machine just did. And unless you have been teleported into a bleak future where machines have total control and humans are mere “drones”, any IMA system worth its name is going to allow a person to leap right in and take over.
In the end, the best approach is to make a good first cut at assigning tasks to people versus machines. And then to use the concept of Continuous Process Improvements to Think, Plan, Do, Measure and Repeat your way to a better and better solution.
Bit-by-Bit #59 from Eric.