The Dawn of the Context Engine. And why Clippy needs to get more interesting

[Editorial notes: Guest article contributed by Bharath Mohan, Cofounder of]

Its 3PM and you are engrossed in a discussion with a new found acquaintance, discussing the future of technology. Your phone beeps. You think its a message from a friend. It is your phone itself, “The next meeting you have may take 40 minutes from where you are – there’s a new traffic situation. You should be leaving now.”
You are about to meet a famous investor. Yes, you are prepared for the meeting. You have all the questions about your business answered. In long and short. You feel prepared and powerful. But how about knowing something other than business, that you can talk to him for the first few minutes of your meeting – just to cool the nerves? You open the last email he sent to you – hoping to get cues somewhere. Flash! He has an opinion on the Open Data Project. That’s something you know a bit about.

Its the weekend, and its 12 PM. You are considering what to do. Lunch at home, or meet a few friends at a restaurant? Your phone beeps. There’s a new restaurant two blocks away, and they serve exquisite Mediterranean – something you love.

Monday, and you are waiting in the airport, traveling to another city to make a business deal. You’ve checked in and have an hour to kill. There are a 100 other people with an hour to kill. Should you slip into a book, or make a new contact? You are now used to your phone giving you surprises. There you go! A CIO of a medium sized company is in the same plane. You also seem to have friends in common. You have just killed an hour, and knocked on opportunities.

Are these warm up scenes from a sci-fi movie? Or are they preludes to a visionary book on the future of technology? Wrong. These are actually happening to people right now. Check out some product promos that offer these experiences.

These are not just promos. People are “getting” them, and even blogging about them. Here’s Fred Wilson – a famous VC, writing about his experience with Google Now.

This is the dawn of Context Engines.

What is a context engine?

We all know about search. There’s this text box. You type some text in, and it gives you results – largely relevant to what you asked. Sometimes, they know who you are, where you are asking from, and tailor results to suit you better. Now what if you can get information without asking for anything? What if this super-smart engine guesses exactly what you need, in the context you are in, and gives you something useful? Thats a context engine. There is no mind reading here. Just plain algorithms playing on a lot of data the “cloud” knows about you – of course with your permission. We are just scratching the surface of possibilities here.

Just by being yourself, at the place where you are, knowing people you’ve known, and doing the things you are doing – a lot can be guessed about your context.

Most of your life is now digitized – you have a search history, you’ve searched from your office and home – from the Macbook Air, and the latest Android phone. Your professional network is known. Your classmates from school are on Facebook. They have their search histories, and connect to the web from their devices. Your life is digitized. Your friends’ lives are too.

Now what about the things you are doing now? Or going to do soon? Your phone knows that its lunch time now, and its a week day. Your calendar knows that you have a meeting at 3PM with this new Marketing person – who’s profile is also available. You are just reading this article on segmentation strategy. What you are doing now – is also digitized.

There’s more. What do you normally do at 6PM on a weekday? Worry about the commute back home. And how is this different from what do at 6PM on a weekend? Where should we do dinner. What do you talk with colleagues you meet everyday? News. What do you talk with someone you are meeting for the first time? Your life’s summary. History. There are patterns in human behavior.

Now connect the dots. The context engine knows you, and your taste. It knows your friends and their tastes. The context engine knows what you are doing now. It has learnt patterns in human information needs over time. It knows if you are doing something routine, novel, or extra ordinary. Putting all of this together, it guesses what would be the most useful information to you – right now!


For a context engine to be successful, it needs to crack many multi-disciplinary challenges. Smart algorithms for context understanding, design innovation to interact with users, address privacy and security considerations of users, and market an unknown latent need. To make things tractable, the products are taking one aspect of your life at a time. Google Now is going after your lifestyle. Tempo.AI and CueUp are going after your calendar and time management. is going after professional online reading. Each of these domains will have different rules to deduce context – which we’ll cover in following posts. There are some common characteristics irrespective of the aspects they cover.

Common characteristics of context engines

Context engines have to be there with you, when you are doing things. The curse of any recommender system is that a user never asks for one. The recommender system has to just be there when the user is doing something, and make a suggestion. If the user loves the suggestion, there’s wow – like in that Fred Wilson article. If he hates it, there are curses. This is why context engines have to neatly fold into the experience of something you are doing already. Google Now is slipped into that Android device. Tempo.AI is slipped into a compelling calendar app. is a browser add on.

Context engines have to be extremely interesting and precise.

Do you recall Clippy? That annoying personal assistant on MS Office, that’d popup and tell you the obvious. “I see that you are typing a letter, do you want me to help?” Context engines that say the obvious will be shunned. They have to tell something interesting every time they show up.clippy

Context engines cannot make errors. Even a right thing, if told at a wrong time is annoying. They cannot make an error in judgment and notify the user at a wrong time. Once they’ve caught his mindshare, what is said – better be precise. The user is in no mood for mistakes.

These add considerable design challenges to context engines. Google Now was not released to the whole world from day one. They chose a small section of early adopters, by restricting to Jelly Bean. Its not available on the iOS or the earlier Android phones – even though there’s nothing that prevents them to launch. They wanted to delay the user acquisition – constantly learning from user reaction. Raj Singh of Tempo.AI speaks of how they’ve been delaying a big launch – after studying user behavior. This is something we understand at With a Chrome browser plugin thats always on the browser – we had to say something useful, and never annoy users. One misbehavior would cause angry users to uninstall us.

Context engines are personal, and should offer personalized recommendations.

I may love Mediterranean food, but only during lunch time. You may love Greek, but only on weekends. A personal assistant that gathers your trust, must grow on it – on continued usage. It should offer explanations, ask for feedback and constantly learn and react.

Personal assistants are long term companions.

Its like marriage. Personal assistants have to find that sweet spot where users will continue to have them even after the honeymoon phase. Users need to feel in control.

Context engines are here to stay. You’ll see a lot more of them in the near future. Your kids will wonder how life even worked without them. So get ready to know them better in some of our following posts. We’ll cover “The rules of context”, and “Security considerations in context engines” soon.

[Follow Bharath on Twitter @bharath_mohan]

Sign Up for nextbigwhat newsletter

The smartest newsletter, partly written by AI.

Download, the short news app for busy professionals