Calix today extended the scope of a set of managed networking services it provides to third-party providers of broadband services, dubbed the Calix Agent Workforce, to include a portfolio of artificial intelligence (AI) agents, that can be used to automate marketing, customer service, operations, and field support.
At the same time, Calix also unfurled an upgraded Calix Broadband Platform, available later this quarter, that is now more tightly integrated with Gemini AI models from Google.
Finally, Calix also added a set of managed networking services aimed at small business and multi-dwelling units, dubbed SmartBiz and SmartMDU, respectively, while upgrading its existing Calix SmartHome™ and SmartTown services.
The core Calix services are already hosted on the Google Cloud Platform (GCP), so support for Gemini AI models represents a natural extension. The next goal is to keep building additional AI agents that business service providers can leverage instead of having to build and deploy them on their own, says Shane Eleniak, chief product officer at Calix.
Each of those AI agents will automate an event-driven set of tasks that should reduce the total cost of delivering managed networking services, adds Eleniak.
Arguably, the issue that many organizations are now wrestling with is to what degree to build versus simply buy AI agents. Just about every provider of an IT platform or service is building AI agents. The cost of using those AI agents will vary widely but over time more organizations are going to focus their internal efforts on building AI agents that automate workflows that are unique to their business, while relying on AI agents provided by IT vendors to automate tasks that are specific to a particular application or platform.
Naturally, the challenge then becomes determining how to orchestrate the management of custom AI agents alongside all the AI agents being made available by multiple providers of platforms and services. One way or another, each AI agent deployed will need to be instrumented to enable organizations to secure, govern and observe them.
In fact, as AI agents continue to be operationalized a hierarchy is likely to emerge, says Eleniak. The bulk of tasks will be performed by lower level AI agents that are managed via a smaller number of higher level AI agents that end users will engage with to automate workflows, he adds. “It’s a layer cake,” says Eleniak.
The most critical issue in the short term is prioritizing adoption of AI agents that actually deliver business value, he notes.
Each organization will need to assess how best to deploy AI agents in a production environment. There is undoubtedly a lot of agentic AI experimentation occurring, but as the pace of AI agents accelerates, many organizations will soon discover that many of their custom initiatives will be supplanted by an AI agent that a vendor is willing to build, deploy and maintain. The issue then becomes determining the additional cost being assessed for the privilege of invoking an AI agent provided by a vendor versus relying on one an organization might choose to build and maintain themselves.
Regardless of approach, the one thing that is clear is before too long there will soon be thousands of AI agents distributed across corporate IT environments, many of which may very well be designed to automate many of the same tasks simply because organizations lost track of which one was built by whom, for what purpose, and when.
