- Blue Stream Fiber is already seeing tangible AI gains, said SVP Josh Turiano
- Pre-agentic AI, the operator spent about $11,000/month just to process work orders
- Turiano thinks smaller operators like Blue Stream have an edge with handling AI data
As operators seek a solid return on investment (ROI) for AI, Blue Stream Fiber is using agentic AI to clear “more than 70% of every single work order” without a human checking it, Josh Turiano, Blue Stream SVP of AI Strategy and Deployment, told Fierce at Broadband Nation Expo.
Before Blue Stream built an in-house agent, it typically took about seven employees and “50% of their time” to evaluate each work order and then route it to the field technicians – a process he said cost the company about $11,000/month.
Now, Blue Stream’s agent – dubbed Metis – looks at each work order “three days in advance, four times a day,” Turiano explained, and it takes roughly two employees to verify that the AI is doing what it’s supposed to do.
“You have maybe two people handling the exceptions, and the equivalent for the tokens that we spent was less than $2 the first month we launched it,” he said.
By tokens, Turiano refers to the building blocks of text that OpenAI models process and how OpenAI charges for ChatGPT usage. Think of them like “minutes on a cellphone plan, back when cellphone plans used to have minutes,” he explained.
But it’s no secret AI-enabled tasks naturally equate to fewer humans involved, and concerns are mounting about AI’s impact on the workforce and how workers can adapt.
Blue Stream, like other operators, stressed that the goal is to replace tasks, not jobs.
“Don’t come in and just say, we’re going to replace all of customer service with AI. It doesn’t work that way,” Turiano said. Instead, focus on an individual discipline in the organization and find ways to use AI to streamline the 10-12 tasks that discipline comprises.
That way, the AI can take care of the tasks with “easy, simple decision making” while the human employees are “strategically focused on the really important parts of your business,” he said.
The advantage of small ISPs with AI data
The data deluge is frequently cited as the top hurdle operators must overcome to reach full network automation. On that front, Turiano thinks the smaller providers could have a leg up on the Tier One guys.
“The bigger operators are going to have a harder time because they have so many disparate sources of data,” he said. “One region versus another might be stored differently, have a different IT stack.”
There’s also just a ton of old data in the mix for large ISPs. AT&T, for example, has records that are 10-20 years old, “because it’s old copper lines and it’s really hard to expose that to an AI,” Turiano noted.
Whereas smaller operators, “as long as they’re smart about how they grow, then they can keep everything in a data lake” and build up the AI from there.
In Blue Stream’s case, he said it took under 90 days to implement its Metis system and get a sizable ROI, and the company can use the AI for a number of queries.
For instance, the company can input a customer’s account number and the AI can summarize the health of the network equipment. The employee taking care of the customer service request can see whether a device has, say, a weak Wi-Fi signal or out-of-range optical receive levels, “while the customer is still describing the problem,” Turiano said.
Further, the AI can identify how many times a customer called, the duration of those calls as well as the reasons they put in the call.
“So we segment all of our data and we have it clearly defined,” he said.