Lila Soto: Iris, I had this thing happen at the dentist this week — they handed me a tablet, said sign here, and I just... signed. Didn't read a word. And I've been thinking about that ever since because of what I was reading for today.
Iris Holm: The Blue Planet announcement.
Lila Soto: Yeah — because that's basically the governance model they're proposing. AI flags a configuration drift problem across your network, recommends a fix, human signs off. Blue Planet and Orange Business called it 'speed with trust.'
Iris Holm: And the question is whether the sign-off is a decision or a ritual. That's the whole thing. June 17th, 2026 — Blue Planet, which is Ciena's AI division, announces CCM: Configuration and Change Management. Multi-vendor networks, AI agents, compliance checks front-loaded before changes reach the network. Sounds rigorous. Except the launch copy leads with capability claims and cites zero outcome data from actual deployments.
Lila Soto: Wait — nothing? No numbers at all?
Iris Holm: Nothing verifiable from CCM in production. That's the tell. When you have to name trust as a feature, frankly, you're admitting you haven't earned it yet.
Lila Soto: And that's the part that keeps snagging for me — not the trust framing, but what's underneath it. Like, okay, think of autocorrect. Autocorrect guesses. It doesn't *know* what you meant. And sometimes the same sentence gets corrected differently on different days. Now imagine that autocorrect is rerouting your hospital's VPN at 3 a.m.
Iris Holm: That's the actual problem. LLMs are probabilistic by architecture — same input, different output. Production telecom networks are not built for that.
Lila Soto: Techstrong.ai ran a piece on July 6th — they gave the same agent the same prompt an hour apart. Two different remediation plans for the same router config. Both plausible. Neither identical.
Iris Holm: Wait — documented, on the same device?
Lila Soto: Same prompt, same router, one hour apart. And I think that's actually what Sean Kinney at RCR Wireless was pointing at in his June 30th note — the one after DTW Ignite. He called this the real story under all the trust-and-autonomy messaging. Not 'can agents do the thing,' but 'will they do the same thing twice.' And here's the scenario that won't leave my head: engineer, 3 a.m., Blue Planet's agent has flagged drift on 47 MPLS edge devices, recommends a sync script. She approves it in 90 seconds. Six minutes later, three customer VPNs go dark. So — who decided? The agent or the person who clicked approve?
Iris Holm: The person performed a decision. That's different from making one.
Lila Soto: And the part that comes later in all of this — whether anyone has actually *published* outcome data that proves the governance isn't theater — I just can't shake that. But set that aside for a second. What does the 47-device scenario feel like from the inside for the engineer who clicked approve?
Iris Holm: She felt like she made a decision. That's the governance theater part. But — and I'll give Blue Planet this much — she *saw* the drift. Forty-seven devices, flagged, specific. That visibility didn't exist before CCM. That's not nothing.
Lila Soto: Oh, so the partial win is real.
Iris Holm: The drift detection is real. The audit trail is real. What isn't real — what nobody's published — is an outcome scorecard. Nokia and Google Cloud have an actual number: Gemini-based agents inside Nokia's Assurance Center cut fault-resolution times by up to 80%. That's a production metric on a named platform. Blue Planet has a framing. Those are not equivalent.
Lila Soto: Wait — Nokia's number is fault *resolution*, though. That's after something broke. Blue Planet is claiming drift *prevention*. Aren't they solving different moments?
Iris Holm: Fair. But Orange Business said publicly — July 14th, 2026 — that AI success should be measured by operational outcomes. Faster incident resolution, reduced manual coordination. Not agent count. Orange is a Blue Planet reference partner saying this. And they're running agentic AI across five domains: 5G security, telco cloud, RAN energy, lifecycle management, incident management. That's the scorecard Orange is building. Blue Planet still can't hand you that scorecard for CCM specifically.
Lila Soto: Huh. And at DTW Ignite in Copenhagen — June 2026 — every major vendor was converging on the same trust-and-autonomy language. Cisco, Nokia, Blue Planet. I mean, at some point that's just... synchronized marketing, right?
Iris Holm: Cisco at least said something honest — trust is *earned* iteratively, agents recommend before they execute. That's a graduated autonomy model, not a launch claim. The TM Forum's 100-plus Open APIs give the whole industry a structural edge here, Kypuros argued that in July — but that's the industry's advantage, not Blue Planet's proof. The question is which vendor is first to publish an actual outcome number. Right now that's Nokia. Everyone else is describing the ladder. Nokia climbed it.
Lila Soto: And that's what I keep thinking about — not Nokia's number, not even the 80%, but the fact that Orange Business's benchmark is just... sitting there. Publicly stated, July 14th. Operational outcomes, not agent activity. And either Blue Planet eventually publishes outcome data for CCM that meets that standard, or the silence becomes the answer.
Iris Holm: Fine. Call it governance architecture. I just hope the architecture holds the first time an agent touches a live SLA at 3 a.m. and nobody can explain which output it chose.
Lila Soto: And that's the thing — the industry spent a decade building over a hundred TM Forum Open APIs so it wouldn't have to take a vendor's press release on faith. Now would be a pretty good time to use them.
Iris Holm: Fair last word. Thanks for pushing on this one.
