This week we’re talking about chatbots, artificial intelligence and the rise of the machines.
But first, breakfast:
- Dave: BLT.
- Conor: PB + Banana on toast and a of coffee.
- Sarah: Coffee + Waffles.
- Lisa: Instant (cheating) porridge + tea.
Conor bought a car for the first time! Conor and his wife, Alana, went to visit the folks at Motorpoint Birmingham, who made the process of buying their first car a breeze. They picked up a new car in Urano Grey. Not as cool as Grey Dreams.
We followed up from Gwera question about how to navigate remote roles during the interview process. Conor weighed in on his experience being the first UK hire for Expensify and how he negotiated his recent move to Birmingham.
We received a question about sponsorship for Jitbit and we entertained the idea of allowing sponsorship to cover some of our operational costs. We’d like to hear your thoughts! How annoying are ads? Would you still listen if we had sponsored content?
We received a tweet from Wiliam (single letters for the win) asking how we share positive feedback with our developer and product teams.
Dave explained a recent change at Buffer that allowed customers on the free plan to also Pause their feeds. They received fantastic feedback from customers and shared this with the team via Slack.
At Timely, the team share #feedback and #celebrate-good-times opportunities in specific channels. This allows them to pass on customer’s general feedback about the product, the support they have received and their overall Timely experience. This is particularly important when new features are released. Includes a terrible metaphor for babies.
At Expensify, they track customer churn and engage with those at risk customers who look like their engagement is winding down. The whole company is kept in the loop with the feedback and suggestions that come out of that process.
Geckoboard also use Slack to keep teams in the loop with customer feedback. However, when Lisa worked at Moz, Slack was too noisy a channel to meaningfully share that information. Instead, they brought both positive feedback and constructive criticism (including feature requests) to their weekly meeting with the product team.
On to the actual topic! Chatbots!
What are chatbots? An automated process that engages with a customer. They are usually based on machine learning models (that the technology has acquired from analysing huge amounts of data). The biggest area of advancement is around natural language processing, which allows chatbots/AI to interpret conversations based on keywords and the context in which they are asked.
What are our experiences with chatbots and AI?
At Expensify, they leverage Intercom’s “Operator” feature. This works with their Educate product to suggest related help documentation that might answers their questions. The customer is asked if the guide answered their question and if not, is connected to an agent and given an expected answer time.
It also helps agents to respond to customers by providing related information and suggestions.
Lisa chips in to mention that Zendesk and their “Answerbot” also provides a similar service, where three related documents are offered to the customer/searcher.
Expensify are also developing their own Concierge bot, which automates functions within the app (running reports, submitting reports etc), as well as some basic interactions with customers. It’s still early days for Concierge, but we look forward to hearing more about it in the future!
Heather from Geckoboard wrote a great article about how they use their own chatbot. You can find that here: https://www.geckoboard.com/blog/customer-success-team-bot/
They analysed billing conversations to identify the common questions that customers ask and taught the bot the different ways in which it can be asked. Customers know they are talking to a bot and are given the option to talk to a human. Of course there are limitations!
Geckoboard also uses Intercom’s operator to greet people that engage with them on their website, as well as collect information from potential customers. They aren’t able to leverage its other functionality as they don’t have their docs stored in Educate.
Sarah comments that movies like “Her” and “Ex-Machina” have brought attention to AI in the mainstream media, but we’re a long way away from reaching anywhere near that level of intelligence from machines. This is something to bear in mind when thinking about how you can implement chatbots and AI in your company.
Lisa talked to Jenny Yang of Aida at SupConf, who said that chatbots are only coming for the boring repetitive parts of your jobs. For more complex cases, human touch will always results in better outcomes.
Remember how terrifying the concept of self-checkouts were?
Conor makes a great point about recurring, repetitive tasks and using these as a herald for where your product needs to improve. Then you won’t have to answer those questions at all!
If you don’t have the room or resources to remove those hurdles, then chatbots could be a way to remove the strain from your support team and move that human energy to other areas that need it more.
Sarah refers back to our last episode in tagging conversations and how machine learning can help you to get more qualitative insights into your conversations.
You can’t underestimate the power of conversation to build rapport and relationship with customers. Automating those conversations may remove opportunities to build those relationships 1:1.
Expensify have recently shared their community.expensify.com forum, where customers can interact with other customers, alongside Expensify employees. This 1 to many situation gives them the opportunity to nurture those relationships, without the “commitment” of a 1:1 customer interaction.
In Support-land, it feels like we’re moving away from CSAT (Customer Satisfaction) being a marker for customer loyalty and towards CES (Customer Effort Score). Could Chatbots help us to get out of the way and let our customers get on with what they want to do?
Dave’s Siri must have heard their ears burning and decided to join the conversation. Siri hilarity then ensued. P.S. Want to sponsor us on Ko-Fi?
HelpScout have some great articles about Chatbots and if you should use them, here:
IMPORTANT NEWS! Save the date for the next London Support Lab: February 1st 2018.