- Kenmei is developing an intent-based agentic AI solution for network operations
- The emerging player is backed by operators Telefónica, MasOrange and Andorra Telecom and partners with Microsoft
- Nokia, Ericsson and Ciena's Blue Planet push rival agentic AI platforms as the race toward autonomous networks intensifies
I first heard about Kenmei in a LinkedIn post by Microsoft Telco Industry CTO Rick Lievano. The top result on a Google search for the company name was a "cross-site manga tracker." Huh?
Then I scrolled down a bit on the search page and saw the fourth search result, for a company with the tagline, "Building the Autonomous Telco Networks of the Future."
Ah. That makes more sense.
Kenmei is an emerging vendor based in Valencia, Spain, joining the field of providers developing tools for network autonomy.
"Our mission is to move from today’s rule-based, human-in-the-loop operations to eventually achieving what TM Forum defines as Level 4 autonomy,” Ali Wansa, Kenmei VP of sales, told Fierce. “That means getting to autonomous operations, where the AI itself will decide and act, and humans supervise by exception. But for us, governance is critical — nothing happens in the network without an engineer or operator reviewing the plan of action and deciding whether to execute it.”
Kenmei, which is backed by three of its operator customers — Telefónica, MasOrange and Andorra Telecom — is betting the obstacle to autonomous networks isn't the AI agents. It's the data underneath them.
The vendor launched its Network Performance Data Product on Microsoft Fabric, sold through Microsoft Marketplace, at TM Forum's DTW event in Copenhagen in June. That followed a Mobile World Congress demo at Microsoft's booth of an intent-based agentic AI solution: an orchestrator agent interprets a stated intent, selects a specialist agent — troubleshooting, root cause analysis, anomaly detection, uplink interference — pulls data from Fabric and returns a plan of action for an engineer to approve or reject.
Months to days
Kenmei's founders sold their previous company, Ingenia Telecom, to probe vendor Astellia. The team's specialty, then and now, is cell traces — records of every signaling event in the radio access network (RAN), which can reach hundreds of gigabytes per hour, Wansa said. Kenmei's business has been turning that raw data into a unified, query-ready layer, rebuilt from scratch at each operator: it consolidated RAN data sources on Azure and Databricks at Telefónica and correlated 21 data sources at Swisscom. The Telefónica deployment's heavy Azure consumption caught Microsoft's attention and grew into a strategic collaboration announced in March.
Level 4 autonomy is the same bar Microsoft applies to its own network operations.
The Network Performance Data Product packages Kenmei's work — parsed, correlated data plus an ontology layer agents can reason over — inside the operator's own Fabric tenant. The product should cut the time needed to onboard a new operator's network data, a process of integration, parsing and security approvals, from months to days, Wansa said.
Agents are relatively easy, Wansa said. "All of that can be built by many players. What's important is the data foundation."
The product is still in early stages. "I want to be very clear and transparent, we haven't deployed it. This is new," Wansa said.
A crowded road to autonomy
At the same DTW show, Nokia rolled out an Autonomous Networks Agent Library, Ericsson pitched agentic AI operations built with Mistral AI and Ciena's Blue Planet began rolling out AI agents across Lumen's OSS.
Kenmei's challenge is scale and dependence. It raised a $4.3 million (€4 million) Series A in 2024, while its rivals are the industry's largest vendors — though it has joined Ericsson's EIAP rApp ecosystem, hedging its multivendor bets.
Kenmei's next test is converting launches into deployments. The company plans to extend the data product beyond the RAN to core and transport networks. Kenmei describes itself as multicloud, but for now, the company is betting on Microsoft. The bet is that as operators get serious about agentic AI, they'll find the hard part is the data — and that its small team of cell-trace specialists will be the solution.
The name, by the way, comes from the Japanese word for wisdom. It's pronounced "ken may."
And also by the way, here's everything I know about manga: They are Japanese comics. I had to double-check Wikipedia to be sure I was right about even that. I don't know what a "cross-site manga tracker is." When it comes to comics, Calvin & Hobbes is more my speed.
Read more on Fierce Network about network autonomy:
Microsoft telco CTO details where AI is paying off — and where it hasn’t