Novo
Summary
- Novo is a financial platform providing checking and debit products to small businesses and independent contractors around the $100,000 annual revenue mark.
- As the CX team scaled from a handful of people to over 50 agents, they were looking for a solution to automate ticket triage/routing and accurately tag conversations according to customer sentiment and intent.
- Lang has given Novo the accurate, real-time tagging they need to route tickets based on agent skill sets, understand VOC and reduce resolution time significantly. Now, they’re making data-driven predictions on the ROI impact of service and product changes, increasing retention and revenue.
Key results
Savings in agents' time: A 2-minute reduction in handling time per ticket, saving 400 agent hours monthly and allowing Novo’s team to focus on customer relationships.
Streamlined routing: Around 60% of inbound tickets are now automatically triaged and routed without requiring a human agent to intervene.
Tagging of fraudulent tickets: Lang identifies and tags bot-submitted tickets allowing Novo to eliminate 95%+ of fraud attempts automatically.
Reduction in hiring needs: The real-time automated routing Lang enabled meant Novo could hold off adding a full-time data analyst to their team for 12 months.
Data-driven decisions: Data provided by Lang is shared with other teams across the organization to predict churn, retention, and revenue — and direct product roadmaps.
Background
Brian Kale is Director of Customer Success at Novo. A key aim of his team is to communicate as effectively as possible with their customers — many of whom have been discouraged by opaque policies and regulations at traditional banks. “We wanted to bring a progressive philosophy to this market in terms of how customers are treated,” Brian explains, “Using human language, providing clear knowledge, explaining as much as we can in as simple terms as possible, and empowering customers through knowledge, tools, and guidance.”
The challenge
“I joined in January 2019 as the head of customer success. We had an intern helping with tickets, but everyone there was using their personal Gmails and sending out text messages and stuff from their own phones,” Brian recalls, “I was hired intentionally by the founder Michael, who wanted to prioritize customer centricity. It’s one of our founding principles and values.”
At the start, with only a couple of people on the CX team, tracking customer sentiment and intent was easy, but as the team grew, that became impossible. As Brian explains, “We went from five people to 30 people in like three months on our Zendesk instance. And we basically lost all control of the quality of data tracking or tagging what the contact reason was.”
The solution
Brian was introduced to Lang through an industry contact, who gave him a rundown on what was possible. “I was interested in how we could streamline intent tracking — that can help take seconds off of tickets,” he recalls, “And when I started talking with Lang — learning about how we can set up business rules and routing to predict what tickets are about without a human making a guess — that was huge. That’s the initial reason why we went with Lang over other tools that could get us partway there but nowhere near this capacity.”
Prior to his role at Novo, Brian had worked at an artificial intelligence company for three and a half years. He’d been exposed to natural language processing, reading emails, and making guesses and assumptions, and thought he understood the challenges. But Lang’s accuracy still came as a surprise, as he recalls, “We gave you (Lang) about six months of data, and you pretty accurately returned back a lot of the same tagging we’d done manually. That was kind of shocking — it was the real, ‘ah-ha’ moment.”
Realizing the value
A couple of months after implementing Lang, Brian remembers a stand-out moment that cemented the platform's value for executives outside of the CX team. As he explains, “In the financial space, there's always risk — threats of fraud where the institution or your customers could be targeted. In one case, bots had been set up to create inbound tickets, and they had similar patterns that, you know, spread out across thousands of tickets.”
“It was hard for an individual human to see, but Lang spotted it pretty quickly and flagged it. And so we then set up business rules in Lang that routed these tickets and closed them out automatically because they were junk. It would have been multiple agents' work without Lang. That was like the first time the whole company kind of went, wow, good job!”
Then and now - how Lang has impacted CX at Novo
“The fact that Lang can automate the ‘what’ means that we no longer have to work on the ‘what’ — we can work on the ‘why,’” Brian explains. We asked him to share his thoughts on the four main benefits Lang has brought to his CX team since implementation.
- CX agents can specialize based on their expertise.
Prior to Lang:
As a small team, Novo’s CX agents had to be on top of the whole breadth of inbound inquiries; as Brian says, “When we were five people, everyone had to know everything. We have around 200 kinds of ticket resolutions, which can be very stressful.” That put pressure on agents, switching between topics and having to refresh their knowledge on the fly.
After implementing Lang:
Now, Novo’s agents can focus on their areas of expertise, developing deep domain knowledge on certain types of queries, and allowing them to resolve those issues more efficiently.
“I think the hardest part of any agent experience is the uncertainty. You want agents to be focusing on about 20 to 30 types of contact reasons because that's a lot of information as is. So Lang really helps them — they don't have to stress out as much,” Brian says.
- Routing tickets is mostly automated — and far more efficient.
Prior to Lang:
Pretty much every routing decision would require human input; as Brian recalls, “Agents would often have to click into tickets, read it, think about it, and then route it.” That was a major time sink for the CX team, which had to be completed before communicating with the customer and solving the tickets.
After implementing Lang:
Now, much of that work is done automatically — and near instantly, freeing up agents to deal directly with customer issues. “Lang is able to route 50 plus percent of tickets — based off the intent it predicts,” Brian says, “So those are tickets that my agents don’t have to view to route. We have seen our handling times come down by a couple of minutes since implementing Lang.”
- Fraudulent tickets are easily identified and dealt with.
Prior to Lang:
Looking back at Novo’s data history, Brian identifies a significant increase in bot-submitted inquiries, explaining, “There’s a huge spike — mostly around passwords and logins. It went from like 50 a week to like a thousand in two days, and that's really when the panic alarm went off.
After implementing Lang:
Lang spotted the emerging pattern and flagged the fraudulent inquiries as they came in. Once alerted by Lang, Brian’s team identified the source as being a bot network and were able to set rules to deal with any future instances, closing them off automatically — meaning agents didn’t waste any time dealing with them manually. “When we implemented the Lang rule, (the issue) just went away — back down to 50 or so a week,” he recalls.
- Product development and QA are based on customer feedback.
Prior to Lang:
Tracking contact reasons accurately was pretty impossible, given the size of the datasets Brian’s team had to deal with. “We do 20,000 tickets a month,” he points out, “You know, 5% are about X, 10% are about Y, etc. If we did not have any tagging, these things would just be going by us — our product team would be making guesses or having to actually read individual tickets themselves.”
After implementing Lang:
As Novo became more confident in their contact reason tracking, because they no longer relied on humans doing it, they were able to prioritize better and identify key customer experience breaks. As Brian explains, “Any change, from check processing to ordering timelines, to small improvements like cosmetic changes, we've been able to justify it.
“Prioritizing which pieces of knowledge to improve or AB test is based on tags. If something has no volume — we don't work on it. If something has hundreds of mentions a month, well, then that's what we should focus on.” he continues, “And the tags also help us with QA. We link our QA scores and our CSATs to tags. We know which tags have higher or lower CSA and higher or lower QA scores. So our VOC QA reports are very tag driven.”
Looking to the future
For Brian, the next step in Novo’s CX journey is real-time sentiment analysis. “It’s about finding the nuance in previous tickets and conversations and saying, well, they've had this experience before, we predict their sentiment level, can we maybe prioritize and escalate this more?” he asks, continuing, “We want to get to a place where we can dynamically change our business needs and rules — prioritizing certain types of interactions or types of customers. And Lang will be the bedrock of that.”