A couple of weeks ago IBM released the first set of trial accounts for IBM Verse, its new Watson-powered email system for enterprises and individuals. The IBM marketing around Verse position it as a transformative approach to email, and a re-imagination of what email is like. IBM’s people have put a lot of effort into talking up the product, even though many IBM employees haven’t yet tried the offering (even those that are trying to sell it to customers). You can see the initial comments I made when Verse was announced in November 2014.
I have some reactions to IBM Verse that I would like to share in this blog post.
Every vendor is allowed to cast a vision of the future. It’s what empires are built on.
Every vendor is allowed a first version. The real test is what happens after the feedback rolls in and what is incorporated into the second version.
The Vision for IBM Verse
There are a few vision statements for IBM Verse. Here’s but three of what I’ve taken notice of:
” … analytics that intelligently and automatically surface users’ most important people and critical actions to focus on for the day. By learning unique user preferences and priorities over time, IBM Verse provides instant context on people and teams.” (see IBM press release)
” … IBM Verse learns your behaviours and adapts to the way you work.” (see IBM Web Site)
” … to unify the services people regularly invoke to get their work done, allowing individuals to apply analytics to prioritize activities; this should help organizations derive actual business insight from all the services used to collaborate.” (see IT Business Edge on IBM Verse)
In light of what IBM says its vision is, do we believe in the vision? Has email become so unmanageable by humans acting unaided, are we thus in need of machines to save us from the beast we have created for ourselves? My answer is yes.
The Reality of IBM Verse (April 2015 Release)
The first version of IBM Verse was released to public testers in early April. It was difficult to know whether the reaction from early Verse account holders was a train wreak in slow motion, or just the initial groaning pains of being forced to adapt to a new way of doing email. One early tester (and an IBM business partner) went within an hour from “it’s nice and will make a nice iNotes update” to “don’t bother; the more I play with it, the more confusing and worse overall it gets.” Another early user said “The more you use it, the “verse” it gets.” Ouch. There were a few other testers who posted comments on their early experiences, and a couple who shared their experiences with me privately. No one said an unqualified “this is amazing.”
For those that make their livelihood from the IBM suite of products, I’m sure they will stick around and try Verse again and again. For real customers, not so much.
I had access to a test account. It could send and receive email. The other promised capabilities were … less functional.
Why I’m Particularly Interested In IBM Verse
In 2006 I gave away my research and consulting firm called Shared Spaces and went to work for an American vendor creating a transformative / re-imagined approach to email (it was not IBM). At the same time I started working on a PhD at the University of Canterbury – which I never finished so don’t call me Dr Michael – which was originally aligned with the work of the vendor but morphed when I resigned from the vendor into a study of how to build intelligence into email in order to develop collaborative ways of working among employees. The theoretical area I started to explore was commonly known as speech acts, and the idea I started pursuing was whether it would be possible to automatically determine the characteristics of how people were communicating through email. In particular, I wanted to look at the back-and-forth interaction between people on a discussion list, analyze the linguistic structure of the messages, and thus see how the interaction was happening. This was the proposed stage one of my research: to write an analytics engine that automatically classified text-based discourse in order to ascertain the speech acts (the underlying structure of the interaction).
For example, consider the following pattern in text-based discourse:
Person 1 – Do you know anything about … [a request]
Person 2 – Here is what I know about … [a disclosure or potential answer]
Person 1 – Thank you. [the closure / end of communication]
The above is an analysis of the interaction at the level of a complete message, but you can go deeper and analyze the sentences and paragraphs inside a message to do a deeper analysis. I won’t get into that here.
Let’s run the same pattern above, but make Person 2 exhibit a different collaborative flavour:
Person 1 – Do you know anything about … [a request]
Person 2 – Why do you want to know about … [a push-back]
Person 1 – I have a valid need to know about … [a justification]
Person 2 – In that case, here is what I know about … [a disclosure or potential answer]
Person 1 – Thank you. [the closure]
What’s interesting about the second pattern is that Person 2’s first response could be based on security privilege, an unwillingness to be helpful, or fear of not knowing the answer. In order to figure out what made that person tick, you’d have to look at a much larger set of their messages, and ascertain whether this was a standard pattern for Person 2 (the way they worked), or just an infrequent pattern.
The above – albeit at the level of sentences and paragraphs in a message – was going to be stage one of the research. Stage two was going to look at how to systematically introduce changes into patterns of human communication through text-based discourse. It would give someone in an organization the ability to see the dominant patterns in use (a pattern dashboard), as well as the ability to “turn the dials” to introduce change. For example, if the dominant pattern among a group was to pushback constantly (which in some cases can be a demonstration of anti-collaborative behaviour), you could turn the dial to decrease the ability of people to pushback. When Person 2 pushes back in the example above, before the message was sent the system would encourage (not force) through a coaching dialogue that Person 2 is exhibiting an anti-collaborative behaviour and that they should change the way they worded the message or the type of message they were sending. Clearly you would have to identify patterns carefully, and set up the ability to turn the dials carefully, because there are moral dilemmas in taking a system-brokered approach to changing communication patterns.
As I said, I had to give up my PhD research to focus on building a new business, but the ideas of intelligent systems, conversation analysis, and reasoned recommendations absolutely grabbed my attention for some time.
Back to IBM Verse
The idea of applying analytics to what is going on in email is a great vision. There is too much email sent and received, too little clarity on what to focus on, and too much overwhelm in dealing with the torrent. A lot of my work deals with reducing the email problem by shifting people and teams into other systems that are more fit-to-purpose for communication and collaboration – Microsoft SharePoint, IBM Connections, Jive, and similar tools. And yet there is still a valid role for email to play, and anything that vendors can do to improve the experience is welcomed.
The idea of “Needs Action” is one example of a good technological improvement in IBM Verse, albeit with provisos around effective use. The Needs Action capability allows an email sender to mark an email for explicit attention by the recipient, and to set a timeframe for the response. While that is a one-sided action which needs to be interpreted and accepted by the recipient in light of his or her priorities, being able to make the request clear in both a human-readable and machine-understandable form is a good move. IBM Verse is hardly unique in being able to be used in this way for sender-driven clarity on requested action, but it is a good capability nonetheless. With respect to effective use, Needs Action will only work if the recipient pays attention to the structured request and either deals with the request in the timeframe noted or renegotiates with the sender.
Recommendation: Focus More on the Sender
I think IBM Verse – or any email system with in-built intelligence – needs to focus more on changing sender behaviour. We see elementary forms of this with the “encouragement” in Outlook Web Access to send links to files in OneDrive rather than the attachment directly. I would like to see the analytics power of Watson applied to encouraging senders to be more effective in their use of email and other tools. For example, the following prompts / directives could be shown to a sender:
“Don’t bother sending that email to Bob. He’s not interested in that topic anymore; it’s not part of his focus areas.”
“You are wasting your time sending emails on this thread. Everyone you are writing to has put the thread on mute.” (this requires recursive interrogation of past action and current intent)
“Your email is poorly written. Please clarify what you are trying to say. I don’t understand it, and I’m Watson.”
“If you want to make progress with this issue, reach out to Sally, Jerry, and Chuck. They have much more interest and political willpower for this.”
“You and Roman have already swapped 10 emails on this. Pick. Up. The. Phone. And. Call. Him.”
“Are you listening to me? I told you to call Roman, not send an email requesting a meeting.”
“Email is not going to work for this topic. You need to build a coalition instead of pushing your agenda. Start with one-to-one coffee chats.”
“Sorry buddy, but you can’t send any new email at the moment. You haven’t dealt with the ten critical emails in your inbox. Deal with those first, and then you can start into new things.”
If IBM could apply analytics, machine learning, conversation analysis and similar techniques to improve sender-side behaviour and thus organizational productivity / communication / capability with IBM Verse – and if it actually worked – who wouldn’t be jumping onboard?
The Moral Dilemma of Verse
While the moral dilemma of analytics in IBM Verse will be less than that faced by Alan Turing and company after breaking Enigma, there are still moral dilemmas to be addressed within organizations. Assuming that IBM Watson can figure out what is happening within organizational life and can thus surface that intelligence within human-readable dashboards, what would be the impact of:
– executives being able to see which employees are working on top priority issues and which are not?
– employees being told by Watson that the issues they are working on are not aligned with corporate priorities?
– executives being able to stack rank employees based on their Watson-interpreted effectiveness?
– the ability to see who is most aligned with new issues, and who will be most likely to support new initiatives?
– the silencing of voices of discontent (which can also be innovative ideas in formative stages)?
– the filtering out via de-prioritization of non-core messages and thus a move towards mono culture?
(The above are “what-ifs,” not statements of current capability as far as I have seen disclosed.)
Each of the above can be used in good ways for good outcomes within organizational life under certain conditions. Each can also be used in very bad ways in the wrong hands.
There are many things that remain unclear about IBM Verse. I can only assume that the free / beta / test version of Verse that has been given out is quite different to what the early customers of the paid version are getting. I fail to see how the reference customers in the original press release can state with such certainty that they will proceed with IBM Verse in 2015 if there is no difference between the two.
IBM makes the statement that IBM Verse is a direct result of the $100 million it has spent on Design Thinking. Let’s be clear that IBM hasn’t spent $100 million on IBM Verse. To have delivered the product we got as a first version after spending $100 million would have signalled gross incompetence; that is not what has happened. Second, in light of comments from an early tester that ” … the Design Team for IBM Verse have done nothing more than introduce new flat colours and a few stock pictures; they have no clue of real life,” we must hope that there is a lot more under the covers than on the covers themselves. For this some patience and more time will be required. IBM has the opportunity to prove that design thinking offers a fantastic new approach to designing software for humans; opportunities of this magnitude should not be wasted.