Doing what life expects from me

This year has been a turning point in my life. For most of last decade, I pursued what I expected from life. From grad school, to building a startup from 3 to 50+ employees, to joining Facebook.

The combination of the amount of reading I’ve done this year & the time to reflect over the past year (thanks to 80% gross margins on $B revenue, I’m not waking up every other night to an outage or a customer call from Australia anymore!) on the privilege that surrounds me (I learnt phrases like “fuck you money” ಠ_ಠ), I decided that I need to do what what life expects from me*.

The New Jim Crow, combined with a few other books, made a lasting impression on me. After reading them, I understood that the cards are stacked against prisoners trying to re-enter society. I realized I understood very little about the specific challenges that a prisoner faces. I remember discussing this with our book club, and the common sentiment was frustration at not knowing what could be done. There were no obvious points of leverage, no silver bullets. For me, it was clear that the best way to learn what was going on & to have an impact was to roll up my sleeves.

After a few months of research, today I start my volunteer orientation with the California Reentry Program that works out of San Quentin state prison to help prisoners re-enter society. To the best of my understanding, most of the work involves conducting research on behalf of prisoners, and providing them guidance based on this research. It also includes teaching them some skills such as resume building, interviewing – things we take for granted.

I’ll be lying if I said I’m not nervous. I’m not nervous about going to prison & working with prisoners (even though the handbook has pages after pages of rules & restrictions, which is a stark contrast after spending most of my life trying to break rules). I’m nervous about what impact this will have on me. Will I be able to handle coming face to face with people who are trying to take a second shot at life? Will I be able to continue doing this over next several months or years, or will I stop? As with all things I’ve done, we’ll find out. One thing is for sure – I haven’t felt this motivated, yet powerless about anything in life. Here’s to doing what I think life expects from me!

* – A phrase I borrow from Victor Frankl.

Questions machines should answer

Most of what I do is build products that help people make better decisions & allow them to do what people are truly great at – building relationships. There’s a host of machine learning techniques that can be used to improve decision making. However, it is important characterize types of decisions made in a typical business. At their core, these fall into 5 categories – Who, What, When, Where, Why.

Who?

This is a seminal question for businesses. Who should we sell to? Who should we hire? Who should we promote? For majority of these, today we rely on search, heuristics, or expert systems. Search and keyword engineering is how most recruiters try to find prospective candidates when they’re building their pipelines. Heuristics is how most sales & marketing teams identify & qualify prospects. Expert systems are often what most analytics platform rely on to provide insight, e.g. which customers show poor engagement.

The problem with existing approaches is primarily that it still relies on people to make judgements or rules, and doesn’t integrate lifecycle learning with decisions in real time.

What?

Once you’ve successfully identified who to engage, the next question is what do you engage them with. For a sales prospect, what is the best collateral for you to share with them? For a potential candidate, what message will resonate most with them & convert them? Again, today we rely on a rep or a recruiter’s ability to search for & mine information to determine what the message & collateral should be.

Many people believe this differentiates good reps from average reps. I fundamentally think this is incorrect. People should do what they are good at, which is building relationships, negotiating, etc. People should not be judged by how good they are at something a machine will be much  more effective & efficient at.

When?

Getting attention at the right time matters. Most existing notions about right timing range are rules based, heavy on conjecture, & “one size fits all”. An example of this is “best time to engage a candidate is at the X year mark”, or “best time to engage a prospect is at the beginning of a new quarter”. At best, these provide guard rails around actions, & at worst, these are old wives tales at best, created based on intuition of “experts”.

Here again, machine learning provides us an opportunity to determine the best time to engage a prospect, based on features of the prospect as well as matching pattern against successful engagement in the past.

Where?

Where we engage prospects matters today more than it ever has. Between all the means of communication (email, phone, ad, in app notification), form factors of consumption (desktop, mobile, watch, etc) as well as  awareness of location, we have the ability to understand the most suitable place & means who which we should communicate. Email has & continues to be the lowest common denominator of means of communication, but this is going to change rapidly.

Why?

What stands in the way of advances in machine learning is how people interpret recommendation, & act on them. We are conclusion forming machines, and absent of conclusions, we have a hard time acting. When a machine answers any of the 4 questions above, a human will need to understand why that’s a good answer. A successful implementation of machine learning for these should be able to inform the user, and help them develop an intuition for a certain decision, or else – people will simply ignore the recommendations.

To summarize, the future of enterprise productivity belongs to companies that enable their customers to answer the who, what, when, & where better than anyone else.