From buying products to customer service or managing contracts — most interfaces provided by companies and organisations still feel clunky, inefficient and annoying. And that’s usually not because they’re badly designed, but because we don’t use them regularly or they’re too hard to reach.
Imagine you want to start saving for your retirement.
Which of these two options would you prefer for buying a pension plan from your bank?
Both options are interfaces are provided by your bank. Most people would probably prefer option 2, as pensions can be quite tricky to understand and the website can feel overwhelming with a lot of open questions. But option 2 comes with a few drawbacks too, as you have to book an appointment, need to go to your local branch and cannot just deal with it whenever you want.
This is not only true for buying a pension but for just about every situation where you have to interact with a company or organisation. Just think of the customer service for your telco, the website of your local government or the app of an airline. Ideally, we’d always have a dedicated person that is available immediately 24/7 via phone or messaging, speaks our language and can help us with any problem wherever we are. This is very hard and expensive to deliver with humans...and that’s where conversational software comes in.
We believe that natural, automated conversations are often the simplest interface available. We’re also convinced that companies and organisations who are able to leverage such conversational software early on will have an unfair competitive advantage, using AI to build “new moats” with a system of intelligence at their heart.
Despite all the hype around conversational interfaces and bots, we’re still in the early days of conversational software. That’s because human conversations get complex very quickly, usually going beyond just “one question / one answer”.
In the example below, there are endless different ways to buy a pension:
This conversation seems simple for humans, but it is not for computers. In most cases, you have a multi-turn dialogue that requires the system to understand context and asks questions back if needed.
When we were prototyping our first bots in late 2015, there really weren’t any good tools out there for more sophisticated conversations. Talking to hundreds of other developers through our Meetup, at bot conferences and online confirmed that we don’t know how to build conversational software yet.
And so, Rasa was born.
Our mission is to simplify the world’s interfaces with conversational software.
Rasa is based on two simple principles:
We’ve heard over and over from bot developers all around the world that they don’t want to be dependent on 3rd party SaaS tools like API.AI to provide the intelligence layer of their applications. By using open source instead, developers can tweak the machine learning models for their data set and also to integrate it into a larger IT architecture. Additionally, enterprises want to own their data, deploy their systems in their private cloud or on premise and are afraid of vendor lock-in.
At Rasa, we believe that great conversational software will be built by developers and that open source is key to making this technology as widely accessible as possible. So, at the end of 2016, we open sourced our NLU engine which is now used by thousands of developers and companies worldwide. On top of that, we provide enterprises with the right tools to develop and run Rasa within a large organisation combined with world-class support.
We’ve read a huge amount of academic literature on dialogue systems and NLU, and spoken at length with top research groups in the field. We feel that we can make a big impact by filling the gap between what researchers publish and what developers and enterprises need, and shipping production-grade software.
At Rasa, we do applied AI research and ship tools to developers all over the world.We do collaborate closely with leading universities worldwide to push the boundaries and also do research ourselves.