As enterprise AI moves from pilot to production, Microsoft, AWS and Cognizant are among those embedding engineers directly inside client operations, as frontier AI takes on a second meaning beyond the model layer. Cognizant frames an “AI outcome gap” worth an estimated US$4.5tn as fundamentally a people and process problem rather than one of compute or infrastructure. Microsoft, AWS and Cognizant have each launched dedicated units in recent months. Judson Althoff, CEO of Microsoft Commercial Business, wrote in July that customers have “moved well beyond experimentation” and are focused on demonstrating a return on their AI investment while protecting intellectual property. Francessca Vasquez, VP of Frontier AI Engineering and Services at AWS, noted customers want to make AI core to how they operate. The model converges on forward-deployed engineering (FDE), popularised by Palantir; AWS has adopted the FDE label outright, Microsoft positions its offering as going further towards “frontier transformation,” and Cognizant argues enterprises need a permanent, accountable workforce combining deep industry expertise with AI fluency. Microsoft’s investment in Microsoft Frontier Company is US$2.5bn, and AWS’s FDE model sends pods of five to six engineers to customers for around 45 days. The shift from experimental to agentic AI has raised implementation stakes, since autonomous systems touching live business processes carry operational and security risks.