Big Data

A smarter help-desk: Zendesk gets brainier with integration

Zendesk machine learning
Image Credit: Illustration by Eric Blattberg / VentureBeat

Zendesk, Wall Street’s latest technology darling, has soaked up some machine learning smarts courtesy of the nerds at

Zendesk, the newly public help-desk company, now integrates with, the latter’s chief technology officer Joshua Bloom announced on-stage at VentureBeat’s DataBeat conference today. is a machine-learning technology that helps businesses identify important events in their data. Equipped with’s artificial intelligence tech, Zendesk customers will be able to focus their human resources on the problems that really require human support reps, automating as much of the other stuff as possible.’s tech has a broad range of uses, including fraud and spam detection, customer targeting, and automating new business rules. Many of those applications apply to the world of customer service and support, which CEO Jeff Erhardt calls an “underserved segment of the market” for machine learning.

Response time is a key factor for happy customers, for example, which is why help-desk platforms already enable some automation and shortcuts to trigger actions specific to certain support scenarios. But those are all reliant on a master list of business rules — effectively if-then statements — which are static unless a human dives in an modifies them. That’s a problem for large businesses with dozens or even hundreds of rules, argues Erhardt.

“We can free you from the shackles of a predetermined, fixed set of business rules,” he said. connects to a customer’s Zendesk account via an open API, imports the customer’s support history, and begins to analyze the relationship between support tickets and fields. Then, as new tickets stream in, it can predict those fields, triggers, and actions based on how humans were doing them in the based.

“We’re effectively capturing that institutional knowledge and making it repeatable,” said Erhardt.

But unlike other machine learning systems, noted Erhardt, provides context that explains why it makes certain predictions — both how confident it is in a given prediction, and the reasons behind it.

“We recognize and have the humility to say we’re not always right, and we’re not always sure,” he said. has a learning framework that enables support reps to retrain its prediction model with human feedback. It’s similar to Gmail’s spam box: When you see message incorrectly marked as spam, you can hit the ‘X’ to tell Google it got the prediction wrong. works a similar way.

This all happens inside Zendesk, so the support reps (and their managers) don’t have to leave that environment to benefit from Erhardt thinks that enabling easy access to machine learning tech is essential to the adoption to the adoption of tools like

The machine learning and data analytics space is hot, with startups like Ayasdi and Skytree raking in millions from venture capitalists eager to capitalize on the newest data technologies. has raised less capital than those competitors — it raised a $2.5 million funding round in March — but it thinks it may be a better fit for businesses looking to quickly capitalize on their data.

Astrophysicists originally developed the machine learning software to search for astronomical anomalies, but subsequently realized its broader commercial value and founded The year-old startup is trying to differentiate itself from competitors with this early focus on customer support — but this Zendesk integration is just the beginning, Erhardt told us.

“There’s a lot more room for different kinds of integrations across all different kinds of systems,” he said.

In other words, expect cloud-based software to keep getting smarter.

More information:

Zendesk provides a customer service platform designed to bring organizations and their customers closer together. With more than 40,000 customer accounts, Zendesk is used by organizations in 140 countries to provide support in more tha... read more »

Powered by VBProfiles

VentureBeat’s VB Insight team is studying email marketing tools. Chime in here, and we’ll share the results.