Enterprise

Companies are looking to a bot called Spoke to answer employees' questions

Key Points
  • Spoke has raised $28 million from Accel, Greylock Partners and other venture capital firms.
  • A people-operations staffer at a start-up that has adopted Spoke spends much less time answering her colleagues' questions.
Spoke's founders. From left, Pratyus Patnaik, Jay Srinivasan and David Kaneda.
Source: Spoke

At an e-commerce start-up called Tophatter, employees used to contact its people operations chief, Sana Hafeez, with questions, because she's been there for years. She would do her best to answer those questions, but that took time out of her day.

Now, all around the start-up's San Francisco headquarters there are red posters advertising the technology of another start-up, called Spoke, whose chatbot in the team communication app Slack can answer people's questions.

"I just point to the sign and say, 'Have you asked Spoke?'" Hafeez told CNBC in an interview. "Need to know how to use the printer? Ask Spoke." She said she spends much less of her time at work answering the same mundane questions. She says she can do more important things, like meeting with colleagues individually.

Word about Spoke has gotten around, and on Thursday the company said it has raised $28 million in funding.

First, when founders David Kaneda, Pratyus Patnaik and Jay Srinivasan left Google last year to work on the start-up, they took on an $8 million round led by Accel. Earlier this year they raised an additional $20 million led by Greylock Partners. Other investors include Felicis Ventures, Index Ventures, Red Dog Capital, Spider Capital and Webb Investment Network.

Part of the allure of a start-up like Spoke is that its product can be used inside Slack, as opposed to being just another stand-alone app. Messaging platforms like Slack are increasingly being used in place of email for work communication, and information is becoming more accessible to everyone on the team as a result. But in addition to Slack, Spoke is accessible as a web app, over email or even in a text message conversation.

It also helps that Spoke's technology uses artificial intelligence to figure out which existing answers in a knowledge base apply to an employee's inquiry, and then provide better answers over time as it learns from whether people tell Spoke if its answers adequately respond to their questions.

"It's basically human-curated machine learning," Srinivasan, the start-up's CEO, told CNBC. He and Patnaik arrived at Google in 2014, when the tech company acquired their previous start-up, Appurify, whose technology tested apps on real mobile devices to make sure they were working as they should. (But Srinivasan isn't the Spoke employee who's most skilled in AI. That would be chief data scientist Rajhans Samdani, who was a senior research scientist at Google before leaving to join Spoke last year. Samdani has built Spoke's learning system atop, among other things, TensorFlow, the open-source AI software that Google released under an open-source license in 2015.)

If an employee feels an answer from Spoke is insufficient, the software reaches out to admins in the right part of the company to come up with something better. The system gets better at determining the right people to field questions based on its understanding of who has answered certain kinds of questions in the past. Consumers have come to expect that sort of intelligence in apps from web companies like Facebook or Google, but in the world of enterprise software, it's not always there.

Many companies are in the habit of spending money on internal service desk software, so the company will have a way to sell revenue. Publicly traded Atlassian and ServiceNow are among the companies that sell software in the so-called service desk category.

ServiceNow has been focused on automating parts of its software with AI thanks to an acquisition announced earlier this year. But as a rule, many existing ticketing and knowledge management systems "are clunky, slow and hard to understand," as Greylock investor Jerry Chen put it in a blog post on the funding news.

And while ServiceNow typically rolls out its software inside big companies, Spoke is interested in gaining adoption among smaller companies, like Tophatter, which has 86 employees. Other early access customers include DoorDash, Neura and Turo, Srinivasan said.

Spoke, which has 20 employees — mostly in engineering — is only taking on a few early customers each week, before letting any company start using it early next year.

Hafeez doubts that Tophatter would stop using Spoke even as the e-commerce start-up keeps growing. Because if it were to give up Spoke, "we will not be disseminating information in a way that is easy, quick and accessible," she said.