You Already Asked Me That!: Pre-Trained AI Answers Employee Questions – Over and Over and….
Out of the box, AI is a tabula rasa, unknowing as a babe unborn. Only after laborious and data-intensive training by data scientists can AI do anything. But what if AI came pre-loaded with specialized knowledge covering 60 or 70 percent of the training and development needed for its assigned task?
That’s the idea behind Espressive, Inc.'s natural language processing (NLP)-based virtual support agent, Barista. The enterprise service management (ESM) AI company based in Santa Clara has pre-trained Barista for ITSM (IT service management) and other business operations – such as HR/benefits, payroll, a tech upgrade, an office move – that generate high volumes of employee questions. Barista is backed by Espressive's Employee Language Cloud, which has a built-in understanding of more than 15 million phrases and 4000 workplace topics. The result, the company said, is “fast deployment without the need for expensive and rare AI talent.”
Espressive said ESM and employee self-help are ripe for pre-trained, Alexa-like AI NLP, citing findings released this week by the Pulse Report, “….62 percent of ITSM leaders are considering or actively pursuing AI projects to reduce help desk call volume so they can divert CIO budget to strategic initiatives,” said Pat Calhoun, CEO of Espressive. “AI is well positioned to enable digital transformation in IT and beyond, but it could just as easily turn into a lengthy, expensive, failed proposition. We believe that AI for employee self-help needs to advance from platforms and searchbots to pre-built consumer-like apps to gain widespread adoption.”
In a phone interview this week, Calhoun told us Barista’s democratization design imperatives are aimed at overcoming the agonies of training and scaling AI implementation – the company wants Barista to be intuitive, accessible and increasingly intelligent over time; easily deployable; extensible across the enterprise; use an architecture designed to scale with the requirements of AI versus AI bolted on.
Barista customer Okta, a $260 million “single sign-on” identity management company based in San Francisco, recently went through a headquarters move, generating a stream of employee questions into the HR, IT and other departments. When employees asked Barista how to get to the new campus, Barista asked if they would use public transportation, bike or car, and then provided instructions with interactive options, including traffic alerts, according to Espressive.
“When we decided to leverage artificial intelligence in our ITSM initiative, we expected a long and expensive deployment,” said Christopher Flynn, VP of Employee Enablement at Okta. “With Espressive, we were able to go live in just three weeks, and in the following two weeks, we automatically addressed 20 percent of our help desk requests submitted via Espressive — ultimately increasing team efficiency, and freeing us up to focus on strategic initiatives. The added bonus was finding immediate value outside of IT by leveraging Barista to respond to employees’ questions about our upcoming headquarters move.”
A core piece of Barista is Espressive’s Employee Language Cloud, which the company said incorporates AI technologies and enables Barista to understand what employees are saying in workplace language, provides a personalized response and, when a question can’t be immediately answered, uses machine learning to identify the correct team or department where the answer can be gotten.
Espressive said the language cloud is constantly growing – as employees ask questions, Barista learns new phrases, phrase structures, synonyms and topics, and is available via a crowd sourcing model to Espressive customers (Expressive said anything specific to a customer remains their IP).
“We believe tool kits and platforms and third party tools, like Watson, they’re great tools, but the problem is most companies don’t know what to do with them,” Calhoun said. “And if you look at employee self help, it turns out 80-85 percent of questions, irrespective of the organization, are the same. People are asking about things like 401Ks, vacation benefits, what to do about a broken laptop, how to do something in Office 365 – there’s a lot of repetition in what people are looking for in every organization, which means the Employee Language Cloud is validated, because nobody wants to recreate the wheel over and over again."
Let's add that no help desk staffer wants to answer the same question over and over again.
Another way Barista gets smarter is by “listening” while agents respond over the phone to tickets. “That means Barista learns while help desk agents do their regular jobs versus requiring agents to find time to continuously write and update knowledge base articles,” the company said. “Barista gets smarter every day, so employee experiences are continuously improved, increasing adoption.”
Calhoun said Barista does more than answer questions, it has workflow capabilities to take employees through a procedure. When issues such as employee onboarding or laptop refresh involve more than a basic answer, Barista delivers step-by-step guidance.
Barista delivers services across Android, iOS, Microsoft Windows, and MacOS, as well as through a browser. It also can intercept employee questions through multiple channels, including email, Slack, Microsoft teams, QR codes and calls to the help desk.
The end result, according to the company: customers report employee adoption of 50 to 60 percent and reduced help desk call volume of 30 to 50 percent.