Why AI Use Cases Will Fail With Poor Data Inputs

AI has, perhaps inevitably, been a running theme of this year’s Enterprise Connect, dominating both the announcements from vendors and the discussions among attendees. RingCentral launched its RingSense AI platform, Zoom revealed its new OpenAI features, and NICE launched its own generative AI solution. 

There is some debate over how valuable the available AI-driven insights and use cases might be, given that, for most companies using multiple UC applications, data is not unified across those applications. 

Zeus Kerravala, Founder and Principal Analyst at ZK Research, told UC Today that the theory behind this wave of AI solutions is sound, in that combining communications information with other information should inform smarter decision-making. However, Kerravala isn’t sure whether “most businesses have good enough data in order to integrate with communications to do that”.  

“Look at the RingSense announcement,” Kerravala said.

“I’m going to combine my Ring info with Zoho and Salesforce and ServiceNow and all these great applications. I’m going to provide some great insight for my sales team and my customer success team, but who loves their Salesforce data?

“In a company I used to work for, a couple of our sales guys just put in there, ‘met with customer’.” 

Kerravala said AI’s most practical initial value in the UC space is as an input mechanism, in which AI can listen to a call, summarize the meaning, and record that call and summary in the CRM software, creating a more useful data set for a future AI to provide more instructive insights. “In theory, good AI leads to better AI down the road(…) Right now, I just don’t know if businesses have the data or data discipline inside” to develop valuable insights immediately.”

Kerravala, after citing Adobe as an example of a vendor using generative AI to help content creators in the creative industries, said there are further practical applications of generative AI in business environments too. For example, planning the talking points for an upcoming meeting by analysing the data of past meetings with the same attendees. 

“AI can be used to improve the pre- and post-meeting experience,” Kerravala said. “Most of the vendors do a pretty good job mid-meeting. We have virtual backgrounds, transcription and translation capabilities. But what happens when the meeting is over? Help me get the meeting minutes out to people and help me prep for meetings. That’s where I see this stuff having a pretty big impact.” 

Kerravala also sees AI as having practical applications in contact centres, as featured in announcements by Webex and Microsoft at this year’s Enterprise Connect.

“I think a lot of the AI capabilities we could have for the contact centre are chargeable items,” he added. 

He argues that companies have started noticing the superior long-term economic value of contact centres after the period of post-pandemic growth resulted in a competitive UC marketplace: “Contact centres, you deal with customers, you deal with sales, so the impact you could have on an organisation is much bigger … I think you are seeing a pivot that way to more companies being contact centre-led.”