By: Eoin Coughlan

In a recent study with GSMA Intelligence earlier this year, we asked SMEs around the world a simple question: “Who do you trust most to help you build the AI capability your business needs?” The answer surprised some people. SMEs said they trust their communications service provider more than they trust the software vendors selling them tools, and more than they trust the hyperscalers running their cloud workloads.

That finding matters. SMEs make up about 90% of businesses if you count them all up. And right now they are looking for a partner. They do not have a CIO office the size of a telco department or a large systems integrator. What they have is a business to run and a sense that AI is moving past them. They want someone they trust to bring the platform, the toolset, the management of those assets, and the agentic framework that lets them get to work without having to become AI specialists themselves.

For telecoms, that trust is the most underused asset on the balance sheet. The question is whether the industry will move quickly enough to convert it.

A different kind of starting point 

I have been through enough hype cycles to be cautious about the next one. What is different this time is that we have clear evidence from the AI-first leaders that the gains are real. The organizations furthest along, including our own teams, have turned internal AI adoption into billions of dollars in productivity over the last two and a half years. Around 100,000 of our people are now using AI-assisted development every day and seeing average productivity gains near 45%.

Numbers like those create two camps. The companies that are concerned about missing out on AI, and the companies that are concerned about messing it up. Both concerns are legitimate. The work for partners like IBM and for the telcos who serve enterprises and SMEs is to bring AI to those customers in a way that is consumable, usable, and that they can trust.

Trust is built on a few things. The right data at the right time, at the right quality. Governance and security that are not bolted on. A platform that customers can audit and explain to their regulator. The industry has been chasing those three for a long time. The difference now is that the technology has caught up to the requirement. That is why the serious investment going into this space is concentrated in a few places. Real-time data platforms that can be trusted. Tooling that brings observability and automation into one place. And sovereign platforms that build in the security, compliance, and auditability that regulated industries require.

What the offering looks like 

One approach now taking shape is a sovereign AI platform that runs on whichever vendor's hardware the operator chooses for its AI factory. It is built on open source, air gapped, and protected. It carries 162 international compliance checks running constantly, so that if a regulator audits the platform tomorrow, the required reports are available at the touch of a button. That removes one of the slowest conversations in the enterprise sales cycle. It provides the environment to address operational sovereignty, data sovereignty, technical sovereignty, and AI sovereignty in one place.

What sits on top of that platform is where the revenue lives. The CSP can be a tenant itself and instantiate its enterprise customers as tenants. In turn, those tenants can manage sub-tenants for their internal business. The platform can offer a curated catalog of software for each tenant to select and consume safely. It can offer agents that handle HR, finance, and marketing, the workloads where SMEs are already buying point solutions from three or four vendors and stitching them together themselves. We have a name for what that stitching turns into inside large enterprises. Agent sprawl. Most SMEs are heading toward the same problem on a smaller scale, and they would much rather avoid it entirely.

The agentic framework matters here. SMEs do not want to buy software, integrate it, manage it, and then wonder which agent goes where. They want their CSP to be the CIO office assistant they cannot afford to hire. While SMEs are looking to partner, CSPs need to provide the sovereign environment that lets their customers accelerate the innovation that compliance and technology challenges have held back. That is a different product from connectivity. It is a different conversation than the one CSPs have been having with their enterprise customers for the last decade. And it is most likely a much higher margin one.