What does Einstein Bots mean for Salesforce customers and partners?
Einstein Bots will be generally available in Summer ’18 release in June.
Bots are a way for a customer to interact with a company through a chat interface without requiring a real person on the other side, and they’ll be guided through either using a menu type interface in the chat window, or just using their own natural language. As an example, I might ask the bot “What can you do for me?”, and I’m presented with “1. Check an Order, 2. Submit a support request. 3. Ask about pricing”.
I could click on these options, or I could just say “hey, can I please check an order”, and #1 would automatically have been selected.
Currently there are several bot platforms available such as Microsoft Bot Platform, Amazon Lex, Google Dialogflow, Alibaba Intelligent Service Robot, and many more. Einstein Bots are different to this as they are fully integrated into the Salesforce platform. They can access Salesforce data without code, can hand off the conversation to a real service agent in Service Cloud without any additional integration nastiness. Previously we’d our bot was configured using Salesforce, Google’s DialogFlow and a Heroku middleware layer. With Einstein Bots, this is now all on one platform, requiring no code or integration.
Einstein Bots are designed to be configured using clicks not code. It’s a wizard based interface, whilst still allowing to invoke Apex methods when appropriate. (read: you can still do complex shit even though it’s easy)
The tech is the easy part, so now what
Previously the opportunity for partners was to create the bot capability, to put all the lego pieces together. We configured the middleware and made all the systems speak to each other, “Hello World…” “Hello to you too, how can I help you!”. Our sweet spot was to target Salesforce customers who wanted an easy way to expose their existing Salesforce data to new channels, and hand off to a service agent at the right moment to improve customer experience. Now with Einstein Bots, the technology itself will be trivial to configure, we need to shift our offering focus to 2 areas:
- Bot capability — ‘who’ this bot is, what they do, what are they good at, what problems do they solve. We’ll own the process of discovering, prototyping and testing it with real users, to make sure the bot tackles real world problems and actually improves a business however they choose to measure success. We know how to manage uncertainty, and iterate aggressively until gains emerge.
- Bot knowledge — in the Matrix, Neo learns Kung-fu after the knowledge is deployed through a disc that is uploaded to his cognition through a digital cerebral connection (what a sentence). “I know Kung-fu” he says 5 seconds later. Our role will be to create these ‘discs’. To be able to teach a bot kung-fu instantly. Industry discs is an obvious start, having a bot that knows a ton about financial planning will mean that on day 1 with our client, we won’t have to train the bot as much. And we get the advantage of literally hundreds of companies into one bot, a customer simply cannot do that themselves.
Tread carefully, failure is public
Facebook messenger bots fail 70% of the time. In fact, it’s not complete hyperbole to say that ‘bots are shit’. There have been some spectacular public failures which I shan’t dare name here given I hope we can help them with the clean up.
Webchat now, which is ok, but social messaging is the opportunity
For the rest of this year Einstein Bots will only be available on web chat. That means no Facebook messenger or SMS yet. This means this won’t be a game changer until they extend it to Facebook Messenger. One of the biggest promises of chatbots is servicing customers in their existing channels.
So you won’t need them to visit a new website or download a new app. A website is not an existing channel. There’s still value in having a chatbot on your website, particularly if it’s easy to deploy and maintain, but it still requires a user visiting that website. Once I can interact with a company as simply as I do with my friends on social media, that’s when we should start seeing a big rush of Salesforce customers to Einstein Bots.
What does it mean for customer teams? Will jobs go?
I don’t think Bots will see their biggest success in replacing people. If you are having a terrible experience with a company, having to deal with a bot is only going to make it worse. I’d top up my mobile phone balance with a bot, but I won’t be buying a phone with a bot. I want to negotiate, getting the best value, chat with a real person.
But I do see bots replacing apps. If you need to order a pizza, chat with Domino’s on messenger, if you need to pay a fine, contact your council on messenger, take a photo of your fine and pay with Facebook or Apple Pay.
We naively thought it would be instant
We need to tread cautiously because the promise of bots hasn’t been realised. It looked and sounded great, and many went all in, and now many are pulling back, particularly Facebook. I think it’ll still come, but don’t expect to deploy a bot right now and reduce your call centre volumes by half or stop development on your apps and websites.
It still however is a valuable incremental gain and we can steer our Salesforce clients in that direction. Our clients have complex implementations and offerings and being able to expose that through a simple chat interface where the technology is seamless is something that should be taken on, especially when it’s just an additional feature of their existing investment.
Where to start
Where there’s lots of failure, there’s lots of lessons, and like the colleague that thinks they are smarter than they actually are, talks without thinking, and has an answer for everything, eventually they have few friends… So what’s a better colleague? The opposite of that, and that’s where we should start.