Open sesame

The core problem: Incumbent BSS vendors built systems for humans to operate through proprietary interfaces. Now, AI agents are becoming the primary users of software that interact through application programming interfaces (APIs) and command line interfaces (CLIs), not graphical user interfaces (GUIs). If your vendor’s system can’t be accessed by an AI agent through open interfaces, your entire AI strategy stalls at the vendor boundary. You need systems that are open to any AI, not just the vendor’s own copilot.

Aaron Levie, CEO at Box, recently said something that telco needs to hear: AI agents, not humans, are becoming the primary users of software. And agents don’t click interfaces. They call APIs and CLIs. To an AI agent, if a capability only exists behind a GUI, it doesn’t exist at all.

Now look at your BSS stack. How many of those 400+ systems can AI agents actually talk to? I don’t mean through your vendor’s copilot, which it’ll happily sell you as an add-on. I’m talking about working through open APIs and CLIs that any agent, from any provider, can call independently.

For most operators, the answer is almost none.

Vendors couldn’t keep up with Open APIs.
Now the bar is higher.

This problem didn’t start with AI agents. The industry has been pushing for open interfaces for years, and vendors have been dragging their feet the entire time.

TM Forum has built a suite of 100+ standardized Open APIs designed for exactly this: plug-and-play interoperability across BSS/OSS systems. A solid 60% of carrier RFPs now mandate compliance. More than 40 operators have publicly committed to preferring compliant vendors. The demand is clear.

As far back as 2019, a TM Forum survey showed vendor Open API maturity running 26% behind what operators needed. Years later, the pattern holds: on TM Forum’s own certification leaderboard, the biggest vendors in the industry (Amdocs, Ericsson, Nokia, Netcracker) are still outranked by smaller companies that took open standards seriously from the start.

That was the old bar, and the large incumbent vendors couldn’t clear it. Now the bar is even higher. In an agentic world, agents don’t just need access to the functions TM Forum defined. They need access to everything: configuration, administration, troubleshooting, custom business logic, data export, the full surface area of the system. They need CLIs for automation and scripting, and chaining commands together the way a power user would. They need model context protocol (MCP) servers so they can discover and use capabilities without needing custom coding for each vendor’s proprietary interface.

TM Forum Open APIs are the bare minimum. And most vendors aren’t passing the test.

Telcos’ particular bind

Levie’s assertion that agents are becoming software’s primary users applies to every industry. But telco is more exposed than any of them, for three big reasons.

1. No industry has more systems for agents to reach. A typical enterprise runs a few dozen core applications. A Tier-1 telco runs 400+ BSS/OSS applications. Every one of those systems is a potential dead end for an AI agent. The more systems you have, the higher the probability that any given workflow hits one that’s closed off. With 400 systems, it’s practically guaranteed.

2. Telco workflows are inherently cross-system. In most industries, a business process might touch two or three applications. In telco, a single subscriber interaction touches at least four, and more likely six. For example, a retention offer requires billing, provisioning, CRM, care, and often product catalog and network policy. An agent executing that workflow calls each system in sequence. If any one of them doesn’t have an open interface, the agent stops dead.

3. Telco vendor contracts are measured in decades. Once AI agents start routing around a closed system, in most organizations, it’s relatively easy to switch to an alternative within months. In telco, you’re locked into vendor contracts that run five to ten years. A BSS migration takes two to three years even when it goes well. These closed systems become anchors that drag your entire AI strategy backward for the duration of the contract.

And here’s what makes all of this compound: every advance in AI, from OpenAI, Anthropic, Google, whoever, benefits you only if your systems are reachable. When your systems are open, a better model instantly makes your retention agents smarter, your charging logic more adaptive, your network policy more responsive, all without changing a single BSS component. When your systems are closed, none of that progress reaches you. You’re locked into whatever AI your vendors ship, on their timeline, at their price. Open systems compound with AI progress. Closed systems are frozen in place.

This is why telco lags behind other industries that have already made this shift. Financial services rebuilt around open APIs. Retail moved to composable commerce. Those industries can plug AI into any system because the systems were built to be plugged into. Their AI strategies compound with every model generation. Telco is the outlier: still running on vendor architectures designed decades ago. The speed gap between telco and every other industry isn’t about talent or ambition. It’s about access. And it will keep widening until the architecture changes.

The incumbents are caught

Legacy BSS systems were built for a world where humans were the users: GUIs, workflows, support tickets. That world is ending. And every incumbent showed up at MWC 2026 with the same roadmap: their agents, running on their platform, accessing their systems through their interfaces. 

Next time you hear their pitch, ask this follow-up question: can my agents reach your system, or only yours?

A vendor that builds agentic AI on top of a closed system hasn’t embraced the agentic era. They’ve slapped a chatbot on the same old walled garden.

Don’t wait for your vendors

You can ask your vendors to open up. You can mandate TM Forum Open APIs in every RFP. And they’ll nod, promise a roadmap, and deliver at the same glacial pace they’ve been delivering for the last decade. My guess is they will ship openness the day after never.

But using systems like the Totogi Ontology means you’re not stuck on their timeline anymore. An ontology gives you the access your agents need. It sits on top of your existing systems (all of them) and creates the open, agent-accessible interface your vendors won’t build. Your AI agents talk to the ontology. The ontology talks to your vendors’ systems. You’re not stuck waiting for your vendors to change anything, open anything, or certify anything. The ontology handles it all.

No other platform in telco gives you this kind of flexibility. Hyperscalers give you connectivity through MCP, but only to systems that are already open. Legacy vendors give you their own agents on their own closed systems. Data lakes give you copies of what you need, but no way to make anything happen in your systems. The ontology is the only architecture that gives your independent AI agents open access to every system in your stack, regardless of which vendor built it.

Fast forward your AI strategy

Here’s my challenge. Go to every vendor in your stack and ask this question:

Can my AI agents access your system through an open API or CLI?

Tell them you’re not talking about their copilot, or their agentic platform. You’re talking about your independent AI agents that you control, that work for you, that you can point at any system you choose.

If the answer is no, you’ve found your AI bottleneck. But now you know what to do: call me. Because, in a world where agents are the primary users of software, a closed system isn’t just inconvenient. It’s irrelevant.

Recent Posts

Get my FREE insider newsletter, delivered every two weeks, with curated content to help telco execs across the globe move to the public cloud.