Akhil wants to buy a phone which does not heat up fast, as his phone is to be used in his Uber – always engaged and in sunlight. His favourite ecommerce website can’t understand what he is looking for, yet there are lots of people who want to buy exact same phone. Welcome to the bot world where user will be able to freely tell the exact problem/ use case and not just be thrown up with some filters on category which 2-3 product people decide should exist. Where organisations will learn from one failed reply to user to make sure next time replies will be done around that problem.
The biggest self-created enemy of commerce apps as we will see will be imposing filters on users and not let them tell exactly what they are looking for. Its no surprise now to know that maximum books sold on Amazon when it was launched was books which were not available over major Book Stores. When users are given power to express their exact problems and those are fulfilled, we have always seen a huge business building out in that particular space. How that will be done? – well technology has always enabled organisation achieve what was seen impossible at one point of time and will be easily understood by ones who have first hand experience working in Neural Network or been part of process where humans provide training to machines.
Problem right now for any bot-tech product is getting rich data. And data is very private to each bot or website. Users buying phone over your bot will not be talking or asking in a similar way over Quora. Research and transactions have different behaviour and we already know how user behaviour is different over booking.com and tripadvisor.com. The only way to collect data is to let users type in whatever they want to and express what they are looking for while training algorithms to use this trained data to find exact goal next time. Fast failures is going to be biggest win in bot tech because more the failed data, more is training, better the bot.
Bot tech right now is enabled by available tools like wit.ai, api.ai, luis.ai etc. , but these are going to be intermediary solutions until bots organisations build data enough to self-train and write their own algorithms for ways they want to solve this problem. Then will be a bot-tech war as we had already seen previously in Search Engines.
What we know today is – it’s not going to be about just algorithms, it will be about data (more). And we should care more about user providing us whatever free text solves their problem, while providing replies productive enough to do goal fulfilment as well as narrating bot capabilities.
In one of the re-training data exercise, we discovered a user asking ASL from MagicX after 3 successful transactions, what do you think? It’s a win or not for us?