Managing Support Knowledge With AI: Talla Helps Toast

One of the great challenges in knowledge management has always been getting the right knowledge to front line workers in real time. Old-style knowledge repositories are simply too difficult to search through when a customer is waiting for an answer. I’ve been poking around the area of managing customer support knowledge for over two decades, and it’s always been challenging—not only to get the knowledge out to the front lines, but also to get it into a system in a relatively straightforward fashion. If AI could solve this problem, it could help a lot of companies.

So I was excited when I started hearing about Talla a couple of years ago—first from Rudina Seseri at Glasswing Ventures, who has funded Talla and where I’m an advisor—and then from Rob May himself, the CEO of Talla. Rob is an interesting guy who not only runs an AI company, but also invests in them, writes a great AI newsletter, oversees a podcast (here’s one I did with him), and occasionally hosts dinners with smart and fun people in the Boston area (I have been invited but haven’t been able to make it, so perhaps I am not smart or fun enough).

Talla has taken on that customer support knowledge issue, and does so through a variety of AI technologies. May recently wrote a blog post that describes the various ways that companies attempt to embody knowledge into chatbots, and it’s clear that Talla’s multi-technology approach is the most desirable if you can pull it off.

Answering customer questions well requires a number of attributes. First, you have to be able to “ingest” structured knowledge content on products and services. Two, you need to add new content easily when a new question comes up. Three, you need to be able to grasp the customer’s intent and figure out which words or phrases are critical in the question. Finally, you need to deliver answers in any medium or format the customer prefers, either through customer service rep intermediaries (“rep assist”), or directly (“customer assist”). Talla seems to do all those things well.

I’d call it a “service knowledge automation platform” more than a specific tool. I you are serious about managing service knowledge, for example, outdated content will have to be identified and updated. Content requires classification with tags and other metadata types. Duplicated knowledge content needs to be de-duped. You will of course need to identify when new content needs to be created to address a customer issue, and assign people to create it—even multiple authors if necessary. Talla handles all that for you.

Of course, most organizations employ transactional systems—generally CRM—for customer support, and any chatbot or knowledge automation system has to interface with them. Talla tries to accommodate all CRM and office productivity platforms, including Salesforce, Zendesk, Hubspot, Gmail, MS Office email, Salesforce, Slack, etc.

Most Talla customers start with the software by ingesting a large amount of text. A Talla bot reads through manuals and online documentation to identify question/answer pairs. That content corpus typically addresses about half of customer questions. The other half comes through usage by customers and service reps, who identify missing content and either add it immediately or forward it to subject matter experts for later creation. Talla’s analytics provide details on the quality of the knowledge base, average call resolution times, and the number and percentage of questions the system can answer on its own.

Talla at Toast

One customer that is using Talla successfully is Toast, a fast-growing Boston startup that provides software for restaurant point of sale and management. Toast’s customers are U.S. restaurants that range from food trucks to national chains. If you’re trying to run a restaurant, you don’t have a lot of time to fix your technology, so Toast needed rapid, 24/7 answers to customer questions. The company did an evaluation of AI tools for customer service, and selected Talla in early 2019.

Dave Snow, Toast’s support enablement manager, said that Toast learned the needed knowledge quickly—“like a smart toddler.” After the software was rolled out to reps at the end of February, ramp-up was very fast. “Talla doesn’t forget anything,” Snow commented, “and it learned what it needed to be effective in 3 or 4 months.” He said that after content ingestion the system could answer 45% of questions, and is now at over 90%. The system interfaces well with the Toast’s Salesforce CRM system.

Emmanuelle Skala, Toast’s Senior VP of Customer Success, said that her company’s goal with Talla was to increase agent productivity, but the software had also filled many gaps in the service knowledge base. She said that Toast would eventually let customers use the knowledge base directly, deflecting some support cases altogether.

Lauren Mecca, Talla’s Senior Director for Customer Success, said that one challenge with any customer support tool is to get reps to use the system. But Toast did that really well, she said. Toast’s Snow said that there was some initial apprehension about how Talla would fit into the service workflow, but it quickly went away. He commented:

We had a very sociable culture historically in the support organization. With no more leaning over the shoulder of the rep next to you, people weren’t sure they would get right answer. But once they began to get more efficient and correct answers in a short time, people became very positive. New reps, in particular, love it—they say they’d be “lost without it.” They probably would be, because we have more knowledge that you can store in one person’s head.

Toast’s service reps were never worried about losing their jobs to Talla, Emmanuelle Skala said:

It was important to socialize the idea and get buy-in from the teams that Talla was a tool to make them more successful. They saw it as investing in their training and knowledge, and making them more productive. We’ve achieved a 10% reduction in average call handle time. Our service tickets have gone up by 25%, so a reduction in time makes a big difference. We still have to hire more people, but meeting the customers’ needs is critical to Toast’s success. We’ll continue to hire reps. Maybe in the future Talla can help slow the growth of new hires, but that wasn’t our goal. We were already an award-winning service organization before Talla, and it will just make us more successful in the future.

Talla

As products and services become more complex, more organizations will need to capture and manage support knowledge and make it easily available at the front lines of support. AI tools like Talla can provide a big boost to companies’ efforts to make their customers successful.

 

I'm the President's Distinguished Professor of IT and Management of Babson College, a Digital Fellow at the MIT Initiative on the Digital Economy, and a Senior Advisor t...