Back to top

The indispensable source for professionals who create, implement and service technology solutions for entrepreneurs to enterprise.

In the Zone

Research roundup: chatbots aid customers, server sales rise, AI behaves itself

Peter Krass's picture

by Peter Krass on 06/06/2019
Blog Category: advanced-technologies

Chatbots are getting the answers. Servers are getting expensive. And AI is getting free from its biases.

That’s the latest from top IT researchers. And here’s your tech provider’s roundup:

Chatbots have the answers

How good are chatbots at answering customer inquiries? Pretty good, finds a new survey.

In the last 6 months, chatbot-using companies had 65% of their customer inquiries effectively answered by the chatbot. Only about one in five queries had to be transferred to a human agent.

So finds a new, if somewhat small survey by Aivo, a provider of chatbots powered by deep learning, a type of artificial intelligence. To get its findings, Aivo polled 30 users of its AgentBot solution.

The survey also found that most customer inquiries take 2 to 3 interactions for the chatbots to answer. Many of questions have to do with information that’s missing from the company’s website.

This information is important to the customer’s purchase decision. Fortunately, Aivo adds, it’s also pretty easy for AI-powered chatbots to dig up.

Servers: sales up, shipments down

The average selling price (ASP) of servers is rising, leading to an odd situation where revenue is up even though unit shipments are down.

So finds the latest numbers from market watcher IDC. It reports that worldwide server sales in this year’s first quarter rose 4.4% from the year-earlier quarter, to $19.8 billion.

At the same time, IDC adds, shipments of servers fell in the first quarter by 5.1%, down to 2.58 million units.

The biggest growth area was midrange servers. Sales for those systems rose 30% in Q1, IDC says, to $2.1 billion.

Low-end “volume” server sales grew by about 4%, while sales of high-end servers fell by a scary 25%.

"Demand from both enterprise buyers and hyperscale companies purchasing through ODMs was less voracious than in previous quarters,” says IDC research manager Sebastian Lagana. “As long as demand for richly configured servers supports further ASP growth, the market will offset slight declines in unit volume.”

The top server vendors by revenue were (in order) Dell, HPE, Inspur, Lenovo and Cisco, says IDC. Dell and HPE were way out in front, with Q1 market shares of 20.2% and 17.8%, respectively. No one else came even close.

AI needs to behave itself

Artificial intelligence and its machine learning (ML) offshoot have a problem with user trust, the result of privacy breaches and incidents of data misuse.

That leads research firm Gartner to make a bold prediction: Three-quarters of large organizations will hire AI behavior specialists by 2023 to improve their brands and lower their risk.

“They must make ethics and governance part of AI initiatives — and build a culture of responsible use, trust and transparency,” says Gartner researcher Jim Hare.

New tools and skills will be needed, Hare adds. Companies will need to identify potential sources of bias — such as AI programs that reflect preferences for certain races, genders, ages or locations. In this area, AI governance and risk-management tools should be able to help.

New jobs will be created, too. We’re already seeing new job titles such as ML forensic/ethics investigator and AI behavior specialist.

Consulting firms are jumping in, too, with new services to audit and certify that ML models are bias-free. Companies will then be able to use these certifications in their marketing materials, as a way to gain or regain customer trust.

One more approach involves what Gartner calls AI explanability tools. These help customers identify and correct any bias in an AI system’s algorithms. One such tool, an open-source solution called LIME (short for Local Interpretable Model-Agnostic Explanations), helps companies find hidden biases in their AI models by explaining the predictions of any ML classifier.

Getting artificial intelligence free of all biases should turn out to be one smart move.



Back to top