Elin Gunnarsson
|2026-10-19|4 min
A prototype is not a product
Pitching an AI voice prototype is can be really easy. Everyone agrees it looks great in the demo… but then comes production.
At CASA26 in Amsterdam, our Co-Founder and Group CEO Daan de Wever described a large Nordic operator that wanted voice intelligence, and wanted it fast. The operator was ready to sign a letter of intent. Then compliance came in, then security, then the requirement to onboard everything into the OSS/BSS so it could run at scale.
Daan’s conclusion: "Fast and production in a telecom environment at scale, it doesn't exist."
That sentence is the starting point for this piece. We want to look at what sits between a signed letter of intent and the first invoice, because that is where the real work starts.
Scale brings owners
Service providers hold what a start-up never will: the customer relationship, the bill and a large amount of users. Daan pointed out the other side of the same coin. They also run several call-control platforms, long procurement cycles and many owners who all co-decide.
Each of those owners asks a different question. Compliance asks where the data goes. Security asks who operates the system. The OSS/BSS team asks how it gets ordered, provisioned and billed. Product asks how it is bundled and priced. Support asks who picks up when something breaks. None of these questions show up in a demo, and all of them have to be answered before a service provider can put its name on the product.
Sovereignty has moved to the top of that list. According to Daan, over the past year it became one of the most sensitive questions in every conversation with Europe's largest telco groups. The question is no longer only where something is hosted, but who operates it and where the data goes.
What a service provider needs before the first invoice
We use this checklist when a pilot turns into a production conversation. Each row is a question one of those owners will ask.
|
Area |
The question |
What has to be in place |
|
Sovereignty and compliance |
Where does the data go, and who operates what? |
EU data residency and an operating model that sits inside the service provider's own architecture |
|
Fit with call control |
Do we have to migrate first? |
A connection to the existing call-control platforms, with no new infrastructure |
|
Provisioning and OSS/BSS |
Can we order, activate and bill this at scale? |
Activation through an API and billing through the service provider's own systems |
|
Digital enablement |
Can end users order and consume it themselves? |
A self-service route from the service provider's own portal or app |
|
Go-to-market and bundling |
How do we sell it, and what does it sit next to? |
Packaging, pricing and bundling decided with the service provider, before launch |
|
Support |
Who answers when it breaks? |
A support model that matches the service provider's, so customers never see a stranger |
Daan was clear that the technology covers only half of this list. The other half is digital enablement, go-to-market, provisioning, bundling and support, done together. He was blunt about the alternative: the industry has a graveyard of platforms that never found users.
Build on top of the estate you have
A service provider's network roadmap will never run at the rhythm of anyone's AI roadmap. Daan also questioned something the industry has recommended for years: a big migration of the call-control estate. Something we rarely see happen. His advice is to accept the complexity. The other side of it is a huge customer base and a trusted relationship.
That is why Dstny Digital Agents are built to run inside existing call control. Our platform connects to any call-control platform as SIP endpoints, with no migration and no new infrastructure. Agents answer calls, route them, log every interaction and hand over to a human when needed. Everything runs on 100% EU infrastructure.
The overlay is only the first half. Above it sits an application layer, and around it the enablement work from the checklist. The service provider's customers meet one brand, and that brand is the service provider's.
The gap is commercialisation
Daan put it in one line: "The gap is not AI. I think the gap is commercialisation."
Operator conversations used to start with a piece of technology. Today they start with how to engage, what the new revenue stream is, and how an AI use case becomes operator revenue, with billing and provisioning at scale. Daan quoted a chief strategy officer at a large southern European operator who said: tell me the use cases, tell me how you do it, I am happy to be a copycat. Service providers want a playbook that has already worked.
The model behind our platform follows from that. Service providers resell or white-label under their own brand, set their own pricing and own the billing relationship with their customers. We do not take a revenue share. As Daan said, "the invoice means the relationship. The invoice gives you the next opportunity."
If you are moving AI voice from pilot to production, start with the checklist above and ask which rows are still empty. We are happy to go through it with you.
