Future Unicorn #227: Level AI

The Quild Future Unicorn is a weekly product-focused note highlighting one early-stage startup with statistically significant signals of becoming a unicorn.

Level AI is an omnichannel customer intelligence platform to automate quality assurance programs and provides advanced contact center analytics, personalized agent coaching, and real-time agent assistance. Level AI's semantic engine monitors and analyzes all of your contact center conversations across all channels.

Founders: Ashish Nagar (CEO)

Signals:

Pain point and persona

Digitization of customer interactions

The voice of a company is no longer the local store associate that customers used to see in person, but rather the contact center agent we may never talk to again. Younger generations prefer to not call companies, instead preferring to communicate via live chat, texting, email, and social media. With are a half-trillion words exchanged between companies and customers in contact centers every day, the latter has become the new digital storefront. While transcribing all these interactions is an opportunity, making it valuable and actionable is a challenge.

Its hard being a contact center agent

Contact center agents are among the most stressed and lowest paid professionals. Everything about their work is monitored and evaluated. They're expected to know everything about a product and problem - when even the people who built those product themselves don't. This has led to high turnover rates of up to 44% in large contact centers.

Difficult to train and retain contact center talent

For managers, the high turnover rate of contact center agents and the shift to remote work during the pandemic have made training these agents more difficult. According to Gartner, 80% of contact center leaders expect an increase in remote work post-pandemic, which will further complicate the training and monitoring of agents.

Product

Level plays in the contact center quality assurance, performance management, and - to some extent - knowledge base categories.

Ingesting, transcribing, and organizing data

The first part is getting three types of data organized into Level. The first is customer data: CRMs/CDPs for basic customer info and ERPs for their purchase/order data (think Salesforce and NetSuite). The second is interaction data: this spans several channels from email to transcribed phone calls (think Twilio, Five9, Gmail, Zendesk). The last data are customer and product policies: knowledge base/wikis (think Guru, Zendesk).

Quality assurance (for contact center managers)

The QA module analyzes each interaction, including key moments, intent, and verbal behaviors. These are fed into a scoring model that scores each interaction which then rolls up to an agent score as well. By scoring all interactions, QA managers can quickly identify and resolve issues.

Real-time assistant (for contact center agents)

Agents have to triage and resolve customer requests and issues in real-time. That is what happens when agents put customers (you) on hold. Level listens to conversations and surfaces suggestions to the agents. There's a lot happening here in real-time: ingest voice data, transfer it to a transcription model, transcription happens, then another model (or set of models) figures out customer intent, then another model searches through knowledge bases to surface the most relevant information.

Customer experience analytics (for CX leaders)

And of course, reporting. CX leaders need to know how the team is doing, how the customers are feeling, and so on. For example, if the number of unpleasant conversations have increased, managers can drill down to understand is it because a product has defects or some other issue.