Bot Enthusiasts Series: Cristina Santamarina

This is the fourth interview of our Bot Enthusiasts Series. We interviewed with Cristina Santamarina who is currently working as a product manager at Reply.ai.

Cristina is a product manager at Reply.ai. She is now based in Spain, though in the past she lived in the CzechRepublic, Germany and Guinea. She worked in software projects for workspace management and healthcare, organized tech community events. Also, she took the stage and the class a few times to talk about innovation, new ways of work and women in technology. Her current obsession is conversational interfaces and how they are already changing the way humans interact with technology, which she follows closely thanks to her work at Reply.ai and her project bots4health.com

About Reply
Reply.ai is an enterprise level bot building and management platform that makes B2C communication suck less. Reply brings business into the ‘Messaging Age’ by scaling their 1:1 communication with customers across messaging channels, markets, and languages, all while maintaining the highest security standards. Reply.ai is based in New York and Madrid and is an alum of the R/GA Accelerator.

You’ll read some questions we asked her about chatbots and conversational flow development and her answers for Bot Enthusiasts Series.

Can you tell us a little bit about building chatbots and how the market will be in the product perspective?

Leaving aside all the considerations about tools and platforms, good chatbots are, in essence, well thought-out scripted conversations. There are two prerequisites to achieve this: First, a diverse, multidisciplinary team must collaborate when defining the flows, copy and rich content of the project as well as the metrics to measure how well the chatbot is doing. The second prerequisite is that conversations must be adaptable and be continuously improved. As with most digital products, the work is not over when the chatbot is published: a very intense editing phase usually begins once the first conversations with real users happen. That’s when the fun begins.

Looking at the chatbot market from a product perspective, I’d say the biggest challenges are educating users on what chatbots are and how they can benefit them, and for the chatbot industry as a whole–defining best practices and improving our products to optimize for discovery and engagement.

If there is an MVP concept for chatbots, how do you think bot makers can improve step by step?

The MVP concept is more relevant for chatbots than it is for other digital products like social media campaigns or mobile sites. There is still a lot to learn both in terms of how real users are using the first bots and defining the best practices that more mature industries have. As pioneers of conversational interfaces, we are still defining our very own Helvetica and Comic Sans.

A good chatbot MVP must focus on doing one thing, and one thing only–and doing that one thing right. Having a clear purpose is crucial. Since the majority of users still don’t fully understand what a chatbot is and how it works, it’s likely they’ll get confused or lost before even reaching the core of your conversation script. Avoid this by quickly introducing your bot–what a bot is–and how to get help from a human if everything breaks.

Once you have the basics down, pick your most important use case, then build a flow for *that* customer journey. And for that one only . Iterate on this introduction and on your most important use case, trying different cadences, copy, rich content. And learn from your users before you expand out of this MVP.

How can bot makers deliver value to users on chatbots?

Value is very related to expectations. One of the easiest ways conversation designers can increase the value they deliver is by adjusting the expectations we set for bot users. Most first-timers ignor e what a bot is. But when they do realise they’re interacting with a computer generated conversation, their impulse is often to “break” it, find the seams, or purposefully engage with it at a much more advanced level than what we can currently provide. We have Hal, C3PO and Roy Batty to blame for users’ inflated expectations.

Combat this by not only focusing on a single use case, but be transparent about what your bot is. The best way we can provide value right now is by creating bots that are very knowledgeable about a specific topic (just like KIA did with their new electronic car) or skilled at one or two specific, simple tasks (think checking your order status or reserving a spot in a fitness class) .

Lastly, the most important element in providing value from a bot, is the human touch. Bots can interpret intents but generally suck at casual chats, learning skills, or empathy. Humans are still much better at this so make sure to program a human take over so that you can complement and fill in the blanks of your bot’s skills. It’s all about amplifying what we can do, not replacing humans!

What do you think a functional bot should have?

I think a good bot has four main elements

1. A realistic goal and defined metrics that will help evaluate success. Bot makers should define early on in the conception phase, as different definitions of success need different flows and communication styles. Are you trying to reach a lot of people? Less people, but more qualified for your business? Long conversations? Short ones? Are you shooting for engaged users that use your bot a few times per week or is this a one-time conversation?
2. A clear navigation, supported by a persistent menu, buttons, and quick replies in the conversation AND the ability to redirects users back to the Most Important Task if they derail. I stumble upon too many bots that are bad at bringing users back to the point.
3. Rich content like images, GIFs and videos to add some branding and make the conversation more eye catching. Just because chatbots are conversational doesn’t not mean they need to be text only. Depending on your audience, add emojis.
4. A way for humans to take over–super helpful for when things get out of control, for getting feedback, quality controls… If you really don’t have the capacity to do, at the very least make a contact form easily available to users, or make sure you spend time reading full conversation logs. Pretty please?

What are your insights and suggestions for bot makers to enhance a conversational flow for maintaining a successful chatbot?

The best way to enhance or improve your bots is to listen to your users. Take the time, read full conversation scripts, and, when possible, talk to real users! Sure, looking at reports and figures is important, and thanks to sentiment analysis tools we can have a general idea of the level of satisfaction of a user. But nothing beats the human eye and the human empathy for the moment.

Reading full conversations can be tedious. If you invest a little time every week, you will better understand who your users are, their reactions to your bot (do they understand what a bot is? Are they trying to break it or do they embrace it?), and what they actually want to use your bot for. This will not only help you improve how you handle your current use cases. It will also help you with new functionality ideas.