Assessing the value of agentic AI in unified communications
IT leaders should pause before rushing into agentic AI for unified communications. First, nail down the business outcomes that will actually prove its value.
IT leaders are under increasing pressure to improve business outcomes, whether it's reducing operational costs or improving customer service. Many are turning to agentic AI services that can independently reason and take action to complete multi-step tasks with minimal human oversight to free up workers within organizations and externally among customers.
More than 80% of customer service interactions will be resolved by agentic AI in 2029, according to Gartner, helping companies to reduce their operational costs by 30%.
But Gartner also served a critical reality check, indicating that more than 40% of agentic AI projects will be discontinued by the end of 2027 due to rising costs, vague business benefits and insufficient risk management.
While the goal is to boost efficiency, there are undoubtedly challenges associated with bleeding-edge technology, including concerns about data privacy, ethical decision-making and the need for sufficient integration with existing IT infrastructure.
Employees are also concerned about their futures, as exemplified by Salesforce CEO Marc Benioff's announcement that the Agentforce agentic AI platform now handles roughly half of customer interactions, resulting in a reduced head count at Salesforce.
Vendors in the enterprise communications space have also hopped on the agentic AI train, including Cisco, Dialpad, Five9, Genesys, Microsoft, Wildix and Zoom. While many of these vendors have publicly stated that employee displacement due to agentic AI is unlikely, IT leaders must cut through the noise to build a business case and plan for an agentic AI future.
Making the case for agentic AI in UC
Proving agentic AI's business value comes down to tying the technology to financial metrics, said David Smith, founder and principal of InFlow Analysis in San Mateo, Calif.
"We must tie agentic AI directly to profit and loss, along with the core operational metrics that leaders already care about," Smith said. "It's not about new, fancy AI metrics. It's about moving the needle on existing, critical business KPIs."
For customer experience , Smith said those measurements include the first contact resolution rate and increases in customer satisfaction scores.
"It's all about measuring outcomes," he said. "From an employee experience perspective, it's measuring the reduction in the employee service request lifecycle. If that changes from days to minutes, that's a major improvement in reducing frustration and helps sharpen productivity. For revenue-generating teams, we may look at sales cycle length and deal velocity."
It's not about new, fancy AI metrics. It's about moving the needle on existing, critical business KPIs.
David SmithFounder and principal analyst, InFlow Analysis
Addressing data privacy and ethical decision-making comes down to understanding the risks involved with deploying agentic AI, Smith said. Transcriptions, for example, are stored on the customer's private branch exchange. Systems must be engineered to identify and mask sensitive data before it gets processed by agentic AI models.
The human factor is another important consideration, along with defining the AI agent's role within a delegation of authority framework.
"The single biggest integration challenge is that traditional IT infrastructure is built for humans or simple point-to-point automations, not for complex reasoning by an AI," he said. For enterprises, Agent-to-Agent Protocol and a Model Context Protocol are essential to achieve richer context, especially in multi-agent scenarios, he said.
'Be your own case study'
Wildix, a multinational UC vendor based in Estonia, said users should trumpet how agentic AI in unified communications is helping their operations.
Anyone developing agentic AI should be "drinking their own Champagne," said Stuart Donner, Wildix' sales engineering manager.
"If you are at the bleeding-edge of doing something, there are not going to be too many case studies," he said. "So be your own case study -- use your own tech, solve your own problems and tell your customers how you are doing that."
For Wildix, one of its primary roles in developing agentic AI tools is to build in the safeguards necessary to ensure the technology is safe for users.
"We put every customer on its own instance, meaning we process the data but don't store customer data, nor are we trying to learn from it," Donner said.
Moshe Beauford is a writer with nearly a decade of experience covering enterprise technology, including AI, unified communications and customer experience.
