Automated Accountants: Personal Assistant Technology

Digital Assistants use context and intelligence to provide a natural interface which increases engagement in enterprise cloud applications.

robot-accountantIn my house, Alexa (aka Amazon Echo) is part of the family.  My young children check the weather whilst eating their breakfast, see how their favorite sports teams are doing, get some jokes, movie times and anything that pops into their head they will just “Ask Alexa”.  They also have a lot of fun seeing how Apple’s Siri and Alexa answer the same questions and which works better for what.  Observing these interactions and interacting myself has been a very good hands on research exercise and I have been thinking for a while of enterprise applications.

We can use personal assistant technology bots to use context and intelligence to provide a natural interface and increase participation in our cloud applications.We can increase participation on two groups of users

  • Casual Users
    • They have infrequent and limited interactions with the applications and do not have the time, or the training and familiarity with the capabilities to participate effectively.  They don’t know what information they can get, never mind how to get it.
  • Power Users
    • For these users it is about reducing the time it takes to do highly repetitive or UI intensive tasks.  This is like me at home getting sports scores or weather from Amazon Echo, I could easily look it up on my phone but it is easier to just say “Alexa, weather” whilst I am pouring my coffee.

I put together a team to enter a hackothon by our UX innovation team last month and we tried to focus on the former use case.  A high level manager who is very busy, runs a team of 50-100 people and manages to budgets but does not have a secretary.  We imagined her wanting to know details of budgets and implemented three flows

  1. Inquiry on remaining budget
  2. Details of who spent a budget on what
  3. Transfer funds from one budget to another (say from travel to computer hardware)

We spent a lot of time trying to make the interactions as natural as possible, so getting the natural language trained correctly was key and we also wanted to use Amazon Echo, IM and SMS messages to interact with the live data in an ERP Cloud environment.

It was a great experience and we learned a lot technically. but probably more of a revelation was the different design strategy for these types of interaction.  The other teams also did some amazing things(read the event review here) so we were pleased to pick up third place overall and the People’s Choice Award (voted on by all the participants).

I fully expect Automated Personal Assistants to be a key interaction model for Enterprise applications going forward, just as we are seeing them start to take off in the consumer space.   Exciting times.

Little Data and The Semantic Enterprise

Small dataI have taken this term and the inspiration to get some of my thoughts in this area into a post from Jake’s recent post on the excellent AppsLab blog, I have been thinking about personalizing experiences in software for a long time but I had not ever called it little data.  It somewhat fits with thoughts I have had around The Semantic Enterprise, which is my preferred term but I think I have a different perspective on it than other people.

So my idea here is that we make use of the data we have about a person and what they do to get the small amounts of data that they really need to them, even though they didn’t know the data was there or that they needed it.  This is perfect for enterprises software because we have a lot of data about your life inside the enterprise and also a lot of data about other people similar to you and we have a whole load of data very relevant to you so we just have to pull those things together – should be simple right?

I will give an example that I noticed where this is done badly.  When I first had children, I bought new born diapers from amazon, then I started buying slightly larger diapers and then I bought a book on potty training, my buying needs progressed in a somewhat predictable way.  Amazon continued to recommend potty training books and diaper related items long after I had bought my past potty training book.  This is a case of good logic to figure out people who buy X also buy Y, but not good to be really smart and predict how my needs progress over time.  In the enterprise our needs also change over time, so we should try to predict that and help you get what you need.

So to give an example in the enterprise, I might want to recommend an online training resource on managing my employee’s expenses or doing performance reviews after I am newly promoted to a manager title, but I should not blindly push that to every manager even if they have been a manager for many years.  In finance the reports I have on my homepage would change based on my activities in the system, if last quarter I was the person approving last minute adjustments for the Slovakian Ledger, then maybe I want to see some reports on the Slovakian Ledger a little earlier in the cycle to get ahead of the process this quarter.  However if I happen to be set up as an approver for the Slovakian ledger but have never approved anything in that ledger maybe I am not going to be interested in having those reports pushed to me.

As you can imagine the possibilities are huge and very exciting, I have barely scratched the surface.  I’ll be giving this a lot of thought and bouncing ideas of smart people.  Your opinions are also very welcome, use the comments below.

Let me know your thoughts in the comments section below.