Google Cloud and Accenture have formed a new joint operation, the Accenture Gemini Enterprise Business Group, that will place engineers directly inside client companies to get Gemini-based tools actually running in daily workflows. The unit sits organizationally under Accenture, not Google, according to a Google spokesperson.
That detail is the real story. OpenAI, Anthropic, Microsoft, and Amazon have each stood up their own forward-deployed engineering (FDE) units this year, betting that implementation, not model weights, is where the durable revenue sits. Google chose to rent that capability from a consultancy instead of building it internally, which suggests the company sees distribution muscle, not model quality, as its current bottleneck.
Forward-deployed engineers, the term for staff embedded with customers to translate raw model access into working software, have become the mechanism enterprises now depend on to get any return from AI spending at all. Enterprises with the budget for frontier models have not consistently had the staff to wire those models into their own systems, and that gap is what every major lab is now racing to close through headcount rather than new releases.
Google’s position going into this deal was weak by one measure. Ramp’s August data on enterprise card spending puts Google at roughly 6% of U.S. AI spend among its customers, against 43.5% for Anthropic and 39.7% for OpenAI. A Google spokesperson pushed back on that framing, noting Ramp’s sample excludes large strategic cloud deals with Oracle, Meta, Anthropic, and ServiceNow that fall outside model API billing. Both things can be true: Google’s per-seat API traction trails badly while its enterprise contract book still grows.
Accenture is not new to this arrangement. It has already launched comparable forward-deployed programs with Microsoft in March, ServiceNow in May, and SAP in June, making it the common contractor across four different hyperscaler and platform bets rather than an exclusive partner to any one of them. Google itself layered this deal on top of an existing $750 million commitment that placed its own engineers inside Capgemini, Cognizant, and Deloitte, plus a separate arrangement to embed staff at CVC Capital Partners’ portfolio companies.
The Wall Street Journal first reported the tie-up, and the arrangement calls for training up to 1,000 Accenture staff on the Gemini Enterprise platform. That is a training pipeline, not a headcount Google directly manages or controls.
The structural risk for Accenture is that it is now the delivery layer for at least four AI vendors at once, including specialized rivals built for nothing else, such as Anthropic’s Ode and OpenAI’s The Deployment Co. Those firms compete for the same implementation work without the overhead of a diversified consultancy, and neither answers to shareholders expecting margin from software licensing rather than staffing. If forward-deployed engineers are becoming the actual distribution channel for enterprise AI, then Accenture, not Google, currently owns more of that channel across the industry, and Google has just paid to rent a larger share of it rather than compete for the underlying relationship directly.
Buyers evaluating Gemini Enterprise in the next quarter should ask whether the FDE team they get is Accenture’s general AI practice or a unit trained specifically for their contract, since that distinction determines whether Google’s platform gets a dedicated advocate or a shared one.
According to TechCrunch reporting by Rebecca Bellan, published September 8, 2026.
