The next customer may not be human

Picture your next customer. You might see you’re most common customer type infront of you. But the thing is your next customer may not be human at all.

It may be an AI agent with a voice number, no desk and no lunch break. It might book appointments, update systems and place orders. To do any of that, it needs identity, context and policy.

That raises a question most service providers have not had to ask yet. When software makes the call, who is the customer? And who provisions that agent, governs what it may do, and who sends the invoice?

What an agent needs before it can act

The problem is, an agent can reason, but shouldn’t be the one that verifies. It has to be told who it is, who it is talking to and what it is allowed to do. The open standards community is working on the same problem. In January 2026 the Linux Foundation's CAMARA project published a white paper on exposing network capabilities, such as device location and anti-fraud signals, to AI agents through the Model Context Protocol. The aim is for agents to consume real-time, policy-compliant network context.

Four needs stand out.

The first two are things a service provider already runs. Context and policy are where the work is. Our Enhanced Agents carry identity, context, multichannel reach and intelligent routing for exactly this reason: an agent that only answers and routes is the simple case.

The rules for AI voice are already live

Policy is not a theoretical concern. In the EU, the transparency duties in Article 50 of the AI Act became enforceable on 2 August 2026. According to Goodwin's analysis, providers of any AI system designed to interact directly with people must ensure individuals are clearly informed they are dealing with an AI, unless that is obvious from context. Fines can reach €15 million or 3% of worldwide annual turnover, whichever is higher. The AI Omnibus delayed the high-risk rules to December 2027, but it left this duty on its original timeline.

Outside Europe, regulators have reached the same place by another route. In February 2024 the US Federal Communications Commission ruled unanimously that AI-generated voices count as "artificial" voices under the TCPA, which makes them unlawful in robocalls without the required consent.

The pattern is clear. A voice agent is treated as its own kind of caller, with its own duties around disclosure and consent. Those duties need an owner, and the owner has to sit somewhere in the chain between the model and the number. Sovereignty is an operating model rather than a hosting decision: which model runs, on which infrastructure, who operates it and where the data goes. Your legal team should confirm what applies in each of your markets.

Whoever provisions the agent owns the invoice

An agent with a number, an identity and a policy needs someone to create those things, keep them in sync and meter what the agent does. That is provisioning, and it is where the invoice starts.

If agents arrive through a third party that provisions everything directly, the service provider supplies the line and someone else holds the relationship. This is a risk. The network stays essential, but the service provider's brand can vanish from the customer's view. As the CEO of Dstny, Daan De Wever, puts it: "Being essential is not the same as being chosen."

We designed Dstny Digital Agents so the service provider stays in a visable position. Agents connect to any call control as SIP endpoints, so nothing has to be migrated first. An orchestration layer provisions and governs agent capabilities across third-party environments. Our Intelligence add-ons activate through the Dstny API and bill through the service provider. And on our commercial model, service providers white-label or resell under their own brand, set their own pricing and own the billing relationship, with no revenue share.

Three questions to settle now

The agent customer is fast, but the decisions behind it are not. Three questions are worth answering this quarter:

• Who issues and governs an agent's identity and policy in your network: you, or a third party?
• Which transparency and consent duties apply in each of your markets, and who carries them?
• Does every agent you host show up in your own provisioning and billing, under your own brand?

So whose name ends up on the invoice. For an agent, the answer depends on who provisions it.

 If you want to go through these three questions for your own portfolio, we are happy to do it with you. Book a call here.