Level AI, which provides call center automation software, raises $20M

The shift to digital during the pandemic put pressure on customer service operations. A collection of recent reports — from HubSpot, Salesforce and Microsoft — prove this out. Ninety-three percent of customer service teams believe that customer expectations are higher than before. On the customer side, 69% of people expect more personalized service from companies while getting products and apps. The stakes are high. Nearly 58% of consumers say that they’ll end ties with a brand or business due to poor customer service.

Automation technologies are one potential way to deliver improved customer experiences. In 2020, Salesforce found that 37% of customer service departments are seeing a higher return on investment from automation than any other department. But challenges stand in the way of adoption, including technical issues, a lack of vision or strategy, resistance to change, and implementation costs.

Ashish Nagar, formerly a product manager at Amazon Alexa’s conversational AI team, set out to solve these challenges by founding Level AI. Based in Mountain View, California, 60-employee Level AI provides real-time conversation monitoring and assistance for contact center agents as well as search tools that enable agents to find information across their organization.

Automating service tasks

At Amazon, Nagar worked on speech recognition and natural language understanding technologies across search-related Alexa services, with a focus on conversational speech understanding. With the learnings from his two years there, Nagar and a small team of engineers built Level AI’s platform, which aims to blend human and machine intelligence to make customer call center service more efficient.

“The idea first started with the use of voice assistants to help frontline workers like nurses, contact center workers, and technicians. We observed a pain point: for these frontline workers, phones and laptops are not a great way to access computing, as they work with their hands,” Nagar told VentureBeat via email. “We focused on the customer service category after seeing the huge market, customer pain, and opportunity to impact the lives of 4 million to 5 million Americans.”

Nagar claims that Level AI can understand what customers are actually saying during a conversation and scour through different knowledge sources (e.g., Google Drive and Zendesk) to surface proactive hints. Agents can refine the platform’s suggestions, he adds, by flagging hints that don’t help with a customer’s issue.

Level AI uses AI to monitor all support conversations for scenarios set by a company’s quality assurance team, tagging key moments in conversations. When the system tags a conversation, the team — or a manager — can look over it provide agent feedback. With the platform’s rubric builder, management can also set up their own grading systems to evaluate support conversations. Additionally, they can use Level AI’s InstaReview feature to automatically pick out conversations with “rich behavior” or that exhibit “negative behavior” and tag them with a color-coded label for later review.

Level AI offers other analytics tools that show agent performance, team efficiency, and related metrics, allowing managers to build custom reports and automatically send status update emails with key stakeholders. Another recently-introduced capability monitors agents’ screens, allowing managers to record desktops for “performance, compliance, and overall activity.”

Possible pitfalls

Level AI, while potentially helpful on its face, is susceptible to the same biases that plague other conversational AI tools. Flaws in these tools often arise from biased data; the millions to billions of examples from which the systems learn can be tainted with text from toxic websites that associate certain genders, races, ethnicities, and religions with hurtful concepts. For example, a team at Penn State recently found that posts on social media about people with disabilities could be flagged as more negative or toxic by commonly used, public conversational AI systems.

Level AI’s screen monitoring feature might give some companies — and agents — pause, as well, considering the ways in which it might be misused. Managers cite the need for protection against time theft — according to one source, employers lose about 4.5 hours per week per employee to time theft. But workers feel differently. A recent survey by ExpressVPN found 59% of remote and hybrid workers feel stress or anxiety as a result of their employer monitoring them. Another 43% said that the surveillance felt like a violation of trust, and more than half said they’d quit their job if their manager implemented surveillance measures.

We’ve reached out to Level AI about its AI development processes and privacy policy and will update this article once we hear back.

An expanding segment

Platform-specific concerns aside, organizations often face blockers in deploying call center automation technologies. One 2021 report found that tech integration — and broken processes — are major (if obvious) barriers in general to running a better contact center.

Still, the customer service automation market is expanding rapidly — particularly in the call center. Research from The Harris Poll indicates that 46% of customer interactions are already automated, with the number expected to reach 59% by 2023. Level AI — whose customers include Toast, OpenTable, and Carta — today announced a $20 million financing round that brings the company’s total raised to $35 million at a valuation four times higher than the last round.

Driving the enthusiasm, in part, is a shortage of talent. During the pandemic, companies saw hold times increase by 34% and escalations rise by 68%, according to the Harvard Business Review. Turnover in the call center industry, meanwhile, is averaging between 30% to 45%. The ambitious promise of platforms like Level AI — as well as competitors such as Balto, Talkdesk, Asapp, and Talkmap — is that they can enable agents to do more with less without compromising on customer service quality.

“The pandemic has been a major accelerant for the business. It led to most customer service agents going remote, which led to a big need for Level AI-like products. Customer service needs to manage, train, and monitor performance in a remote environment, which can only be enabled by technologies like Level AI,” Nagar continued. “Movement to the cloud is still in the early stages in the contact center space. This is both a challenge and opportunity … there is a lot of untapped potential there.”