The other day I missed a call from a business contact who called my Skype for Business number. Fortunately, Microsoft transcribed it for me and emailed me the following:
“Hey James, a Jason calling you from council idix it 3 PM. Eastern Time, 2 PM. Central I believe um please. Give me a call back on my MP …”
It’s full of obvious errors, but it got the idea across, and its success is the result of AI-based research, according to Microsoft.
Since I’m more likely to miss a call than to ignore an email, I like it. Microsoft has a lot of effort invested here.
It’s the tip of an iceberg of Artificial Intelligence/Natural Language Processing in daily life — with what I’d charitably call mixed results in customer experience.
But unlike maybe Machine Learning, which still feels like it’s from the future, I’ve experienced AI/NLP-infused “neighbors and “coworkers” with increasing regularity.
Chatbots — I just moved, which required setting up phone service, cable television and other utilities. That shopping was where I encounter this technology most: I’m certain I chatted with an NLP-enabled chatbot that got me all the way to the closing stage on my cable subscription. A person came on the line — and ironically lost the sale.
Virtual Assistants — My boss Adam uses an AI-powered digital assistant called “Andrew.” “His” full name is Andrew Ingram — and his performance is hit-and-miss. While relentless, he sometimes misses cues. But here he is helping schedule a meeting with the boss:
“Hi James, I haven’t heard back from you yet about this meeting. ADAM is available on Wednesday, Dec 13 at 11:30 AM CST. Is this time convenient?
ADAM is also available on Friday, Dec 15 at 11:30 AM CST.
Andrew
Andrew Ingram – Assistant to xxxxxxxxxxxxxxxxxxx
Want me to schedule your meetings too? Sign up for your own AI assistant from x.ai.”
Contract Abstraction — On an unrelated project, we are talking with vendors who provided NLP services to extract contract information after a merger. Companies that do that in our space include some big ones, several consulting companies, and about a half-dozen startups with exciting takes on the same technology to recognize key terms in a contract and extract them into a searchable database.
Interestingly, all of those options are pretty viable in a business sense, and all of them “learn” fastest when paired with internal subject matter experts. It’s an exciting field that I’m learning more about — in the classroom and outside of it.
NLP at Google — My personal favorite, since our kids go to a French school, and our parent emails are in French. Mais ce n’est pas grâve. Google is there for me. They focus NLP research on algorithms at scale, across languages, and across domains. And that means, of course, its products are ubiquitous. Google Search, text prediction, and Translate are three examples. There’s more on how they build that kind of robust, scalable, widely distributed service here.
The NLP algorithms they use predict part-of-speech tags for each word in a given sentence, as well as features such as gender and number, and then also label relationships between words such as subjects and objects, and so on.
By the way, there are 374 publications on Natural Language Processing at Google at the link above. It’s an awe-inspiring look at the work that goes into making a ubiquitous service seem elegant and simple.
It’s not simple, but it is increasingly elegant.
James Janega works in the innovation portfolio group at Cushman & Wakefield. Previously, he led the Innovation & Insights group at Slalom Chicago and worked as an Entrepreneur-in-Residence at the University of Illinois’ EnterpriseWorks accelerator. He is a member of the Chicago Ideas Co-Op, and helps enterprise companies and startups improve the way they think about innovation strategy, customer validation, building more robust innovation capabilities and processes. You can reach him at @JamesJanega on Twitter.
James Janega 