Customer Service Just Got Smarter

Announcing Service Desk Integrations in Watson Assistant

Good customer service isn’t really a differentiator anymore — it’s a requirement to stay competitive in most industries. So much so that 62% of consumers say they’d be willing to move to a competitor’s product after only 1 or 2 poor customer service experiences. This is why businesses are moving from traditional communication methods (i.e. phone, email) to more convenient and innovative ways of interacting with customers and solving their problems (i.e. self-service functionality, web-chat, proactive outbound messages, Facebook Messenger, WhatsApp, Twitter, etc…).

However, this improvement in convenience for customers has led to new problems — the proliferation of customer communication channels and a huge increase in conversation volumes within the contact center. Before businesses can truly embrace “omnichannel” customer communication, they have to…

a) centralize their customer communications into a consistent experience on any channel

b) introduce intelligent automation

The good news here is that Watson Assistant can help you with both problems—just think of it as a centralized problem-solving engine for your customers. And although we’d love to think that you could eventually build the *perfect* virtual assistant (that handles every possible sales or support question to completion), the reality is that some issues are nuanced, complicated, and/or emotional. This means that a human agent will undoubtedly need to jump in to finish the job when an assistant cannot.

Announcing Service Desk Integrations in Watson Assistant

The even better news is that we want to make it way easier to launch an assistant that your users will love using — faster. That’s why we’re announcing fully managed service desk integrations with Intercom that…

• respond to questions from your end users only when confident
• handle the contextual handoff process between your assistant and your human agents
• can suggest responses to your agents (in-line) to help you build trust in how your assistant handles questions

Watson Assistant’s goal is to save your human agents precious time by handling common or monotonous scenarios. This lets your agents focus on the things that matter most: complex, nuanced, and/or high-value customer issues.

Our very first integration (that launches today) is with a product called Intercom — a customer messaging platform that helps drive business growth through better relationships throughout the customer lifecycle. We love Intercom because of their super personalized and messaging-first approach to managing customer conversations.

Here’s a small taste of what the integration can let you do with Watson:

Watson can gather information (left) before transferring to the correct agent (right)

Excited yet?! The rest of this post explains all of the cool features within the Intercom integration in more detail!

Quick Setup

Watson Assistant acts like another agent on your team within Intercom. This means that it uses a separate seat within your Intercom Inbox subscription.

Once you’ve added the new seat representing Watson, it’s super simple to connect (in just a few clicks).

Quick Intercom connection process

Responding When Assigned

The fundamental behavior of Watson within your Intercom workspace is that it responds only when it is assigned to a conversation. This allows you to accomplish a few key things:

Scenario 1: Intercom assigns new conversations to Watson

This scenario is the ultimate goal of most of our customers — Watson acting as the initial handler for every new conversation. And it’s super easy to set up in Intercom too. Just configure routing rules within Intercom’s “Automation” settings to route some or all of your new conversations to Watson.

Route conversations to Watson using Intercom’s automation rules

Scenario 2: Agents manually reassign conversations to Watson

Another key scenario involves Watson at the end of a conversation.

Let’s say an agent has been talking to a user about a sales issue — and then resolves the issue and closes the conversation. Often times, users will effectively “reopen” the same conversation with a brand new topic. The “reopen” in intercom happens wherever the message was left when it was closed (i.e. in an individual or team inbox), but the new topic may be better suited by a different team. Here’s an example:

Thanks for sending that transfer over to my friend!

No worries!

[closes conversation]

[2 weeks later] Ugh I don’t have enough money to pay my bills this month. Can you help me with that too?

Oh no! I can’t help with that, but our Support team can. Let me transfer you.

With Watson, agents can now reassign a message to Watson (at the logical end of a conversation or topic) and then close that message. In the event that the user sends in a message within the same thread down the road, Watson will triage the message like it would any other. It can either respond directly if it’s trained to handle the message or it can route the message to the correct team in Intercom based on the topic.

Agent reassigning a conversation to Watson (and then closing it)

Topic-Based Escalation Routing

Lots of our customer service operations use very simple, rule or keyword-based agent routing policies. But this often leads to “misroutes,” wherein the end user experiences a delay in response as the initial team reads the issue, realizes the incorrect routing, and then re-routes the conversation. Here’s a common example:

Hey I’m having trouble making a money transfer to a friend

(Wait 5 Mins)

Whoops, you were routed to the billing team — let me get you over to the transfer support team!

Ok

(Wait 5 Mins)

Hey, let me help you with that transfer!

Look familiar? With Watson, you can not only detect the user’s general “intent” (i.e. Transfer Money) but also gather additional clarification information (i.e. How much are you trying to pay?) before deciding where to route the conversation.

You can also summarize and pass the context about the information gathered to the agent so that they’re fully prepared to handle the task when the conversation is transferred to them.

Watson can gather information before transferring to the right agent

Additionally, you can configure Watson to handle “digressions” during a flow so that a user is always one step away from an agent if they get frustrated and/or ask directly for an agent.

Digression scenario (left) and digression configuration (right)

Inbox Monitoring

When you’re just getting started with Watson, you may not want to route every new conversation through it.

Here’s why: let’s say you only have 5 “dialog nodes” (effectively different actions) built into Watson. If you were to route every new message to a Watson Assistant (with only 5 things built in), your assistant would respond to most of those messages with something like “I didn’t understand you, did you mean…?” This isn’t a great user experience.

Example assistant dialog that covers only 5 topics (left) doesn’t handle off-topic situations quite well (right)

That’s precisely why we built Inbox Monitoring — a feature that silently monitors an Inbox in Intercom and takes action only when confident. To make it work, just create a new team in Intercom that’s only meant for Watson. Then link your assistant to the inbox like so:

Enable your assistant to monitor a Watson-specific inbox in Intercom

And make sure you update your automation settings in Intercom to route to this inbox instead of to Watson directly:

Change initial conversation routing to the Watson-monitored inbox instead of to Watson directly

Inbox Monitoring lets you accomplish a few really cool things:

Respond Only When Confident

By default, Watson will respond to any message it is trained to handle when the intent confidence is above 75%.

This means that whenever an off-topic message, like the “I can’t pay my bills” example from above, reaches Watson’s Inbox, it’ll silently get escalated to a general team inbox monitored by humans.

What’s even cooler here is that the off-topic message will still make its way into the analytics section of your skill. This lets you review it and re-classify it so that Watson can handle it properly in the future!

Silent Team Routing

Some topics, again like the payment example from above, are always better suited for human agents. Inbox monitoring lets you silently route conversations to the right team, all without Watson saying anything to your end users.

To configure payment problems to go straight to the Support team, you would want to do two things:

1 | Set the “action” for payment problems within Inbox Monitoring

2 | Set the team to handle payment problems in the escalation settings

The resulting experience (for the agent) would be the following:

Suggest Responses to Agents

You can also have Watson suggest the proper response to your agents. This helps you build trust in the validity of your responses (by testing them on agents) before letting Watson take over the topic entirely.

Watson providing response suggestions privately to agents

Ignore Pleasantries

The last thing you may want to configure is to have Watson ignore pleasantries at the beginning of a conversation. Often times, users will send something like “hey” as an initial message before getting to their real issue. Here’s an example:

Hey

I can’t pay my bills!

You don’t want Watson to act on the “hey” — but rather, on the second message. To make this work, simply configure a “hello” node to “do nothing” in the Inbox Monitoring settings.

The resulting experience (for the agent) would be the following:

Instead of this:

That’s the integration in a nutshell! We’re super excited to see what kinds of scenarios you’ll automate within Intercom.

To get access to test out the integration:

• If you’re already a Watson Assistant Plus or Premium customer, the integration is available now!
• If you’re not a plus customer yet, our sales team can set you up to try out the plus plan for 30 days — shoot them a message requesting access here
• If you aren’t a Watson Assistant user in any sense yet, sign up for a free plan, then go ask our sales team to get you a trial of the Plus plan :)