Now with AI, Guru brings the knowledgebase to service agents

Don't you just hate it when you contact customer service and the agent can't help you resolve your issue from the get-go? It's a common problem because of three big challenges businesses face with putting the right knowledge in front of agents — collecting it, keeping it fresh, and making it available when it's needed. Guru is a startup that has a novel approach to solving these problems, to which it has just added three new capabilities, including artificial intelligence (AI).

We first caught up with Guru at the turn of the year to find out how it tackles these problems by presenting how-to knowledge directly in the agent's workflow — usually as a browser extension or in a Slack message thread — while keeping the information up-to-date by prompting subject matter experts for regular verification. Now Guru has taken a big step further towards speeding up delivery of the right information to agents by using AI to proactively suggest knowledge that may help solve the issue.

Previously, an agent would have to actively search for help, but with the addition of the new AI Suggest feature, Guru automatically serves up pertinent knowledge based on the content and context of the conversation. The machine learning is trained using previous interactions in the Guru system, looking at what information agents use or decide not to use in each interaction, and whether they're using it in a service context or a sales context. The training is iterative, so that it continues to improve over time, explains Guru co-founder and CEO Rick Nucci:

It's because we're in that workflow that we're able to create these really high-value training datasets ...

[Agents] are using Guru to solve tickets and chats. Every interaction they do is training Guru, and specifically the browser extension begins training itself from what you do every day.

It's completely implicit training. Both the usage and non-usage are signals for making the engine smarter all the time.

AI making humans better

The interplay between the human agents and the machine learning is a crucial element, Nucci adds. While some in the customer service industry may see AI as a way to automate humans out of the process altogether, Guru sees it as a way of helping them be more effective at their jobs:

The risk is, we're going to rush as an industry and say, 'Let's put algorithms in front of our customers and get rid of the customer service agents and sales people.' The reality is that the technology is not going to understand critical things like empathy that is required to have a great customer experience.

This type of technology makes humans better at their jobs and this is key in how we think this evolves.

The second new feature, called Sync, introduces the ability to integrate information into Guru from an existing knowledgebase or other source. This gives Guru more appeal to established organizations that have already built up a significant cohort of knowledge which remains relevant for solving customer issues. It's a very common request, says Nucci.

There's just always going to be other sources of knowledge. We're more interested in making the knowledge accessible to you than worrying about where it's stored.

The third feature, called Impact Analytics, collects metrics to show the impact Guru is having, for example measuring how frequently it's being used when interacting with customers, or analyzing which content is mentioned most often. This is an evolution of existing analytics capabilities in the product, Nucci explains:

Our customers were figuring how to do this on their own and we realized this was something we could productize and offer to all our customers.

Customer service as a revenue generator

Guru aims to align itself with the growing trend towards seeing customer service as a source of potential revenue rather than merely a cost center. It therefore wants to be known as a "revenue empowerment network," says Nucci:

We're really trying to more intentionally make it clear that we're talking about revenue teams. That includes sales teams, but we want to acknowledge service and success teams as part of the revenue team. Those service teams play a huge role in customer loyalty and customer propensity to renew or upgrade.

The value of good service leads to the propensity to buy more. We see more and more customers organize themselves around this concept of revenue teams.

The term also acknowledges the empowerment and confidence that team members feel when they have easier access to knowledge, while the network element references the potential of resource-sharing, he adds:

'Network' because the more people and companies that use Guru, the more powerful it gets. It does become a network, not siloed to a department. It's relationships across a company that make it work.

AI and the knowledgebase

The way that Guru is using AI makes extensive use of that network effect. Although it's possible to get started with AI Suggest just based on finding content matches in the knowledgebase, it comes into its own when it starts to have enough historic behavioral data to make suggestions based on the workflow context in which the 'cards' of information are used. Nucci explains:

What knowledge is being used in the day-to-day workflow and the insight that comes from that is the real interesting takeaway.

That correlation between the chat and the Guru cards used is all part of how Guru is training itself, by being able to understand the specific topic context and the corresponding content that was used by Guru.

The regular verification in the Guru platform also plays a crucial role in ensuring that the knowledge being referenced is sound, he adds:

We knew that, if we didn't get that verification engine in place in an ongoing basis, this wouldn't work. You would change your product and the answers that used to be right would be wrong. The underlying platform that makes sure that the knowledge is accruate is totally critical to making this all work.

The final ingredient is to deploy the right combination of machine learning models at the right time, he says.

It's supercritical that there's an intentional narrow focus in how these things get built in order to make them good and useful ...

When you start using Guru on day 1 and you don't have historical training data, it's going to start suggesting things right away using more NLP-based techniques of comparing words and phrases in the chat and finding correlations in the Guru knowledge.

Over time, we move to a deep learning model when the volume of data warrants that. That combination of techniques we're pretty keen on. What we've learned is the techniques you use evolve as your usage of the product evolves over time.

My take

This new evolution of Guru's offering brings it firmly into line with the philosophy of coaching networks advanced by one of its investors — Gordon Ritter of VC firm Emergence Capital. Guru now becomes a prime example of the class:

Coaching networks are enterprise applications that collect and analyze human behavior, and then guide their users toward behaviors that achieve better outcomes.

It's a strong use case for AI, applying machine learning in a well constrained business context where the technology can start to add value immediately. That's refreshing compared to many AI-infused product announcements that lack practical business relevance.

Guru is also one of the few platforms we've come across that directly harnesses the virtuous cycle of engage, monitor, improve that we see at the heart of the XaaS model, helping companies collect feedback from customer interactions that help them improve the customer experience. Therefore it should be no surprise that most of Guru's customers are digital businesses and SaaS vendors, the two types of business that are most aligned with the digitally connected XaaS model.

All of this means we'll continue to keep a close eye on Guru and look forward to reporting back on how customers are making use of the platform.