Billions to put engineers inside client companies: what the forward-deployed engineering wave means for Swiss businesses

In the space of a few weeks, the biggest names in tech have announced large-scale investment in the same approach, which is to send their engineers to work inside client companies. AWS is committing USD 1 billion to an organisation of forward-deployed engineers (FDEs), CNBC reported on 30 June 2026. Microsoft has created ‘Microsoft Frontier Company’, with USD 2.5 billion and 6,000 embedded experts (Microsoft announcement, 2 July 2026). OpenAI has launched its own deployment entity, the ‘OpenAI Deployment Company’ (OpenAI, 11 May 2026). The message is the same everywhere. The AI models are good enough and getting them into production is the bottleneck, so the engineers are moving to the front line. For Swiss businesses, Djtal draws three conclusions from this wave: the model has been validated at the very top, the billions will stop at enterprise clients, and the practice itself can be written into a contract by a company in French-speaking Switzerland.

What the wave admits without saying so

These announcements are an admission. For three years the dominant story was that AI would be adopted like self-service software, with a subscription, some training and then the gains. In 2025, MIT measured where that model leads, and found that 95% of generative AI pilots produce no measurable return. The giants have drawn the conclusion Palantir drew twenty years ago. Value appears when an engineer builds inside the client’s real systems, working with its data, constraints and teams. On its own, the product stalls before production.

Three readings for a Swiss company

First reading: the model has been validated at the very top. When AWS, Microsoft and OpenAI all put billions into the same approach, the market has its answer to whether an AI project needs an embedded engineer to succeed. A company offered ‘licence, training and good luck’ now has a benchmark to measure that offer against.

Second reading: those billions will stop short of you. The giants’ FDE organisations target enterprise clients such as banks, global industrial groups and public sector bodies. A company with 20 to 250 employees in French-speaking Switzerland will never see an OpenAI engineer move into its offices. The New Stack noted on 28 May 2026 that ‘the FDE model used by leading AI companies is typically inaccessible to smaller organizations due to the cost of sustaining it’.

Third reading: the practice itself travels. The model works without a giant’s budget. It requires an engineer who writes code inside your systems, a measured operational result and autonomy handed over at the end of the engagement. A company in French-speaking Switzerland can make those three requirements terms of its contract, provided it puts them in black and white before signing.

The question to put to your next AI supplier

“Do your engineers work inside our systems, and what stays with us when they leave?” The answer distinguishes the three kinds of work on offer (consulting, integration and embedded engineering), which we compare in our guide to FDE vs consultant vs integrator. We describe how we apply the model ourselves, on the ERP and CRM systems that companies in French-speaking Switzerland already run, in Forward-deployed engineering at Djtal.

At Djtal, nine AI agents have been running our own processes since May 2026.