The money for artificial intelligence companies is with the companies, but the challenge isn’t building models. Now it works within a company that wasn’t designed that way.
Makers of AI models are racing to address the gap between idea generation and integration. Microsoft launched Frontier Company with $2.5 billion and approximately 6,000 engineers, technology consultants, and industry experts. Their job is to build AI systems that reside within enterprise customer organizations and produce measurable results. Amazon has committed $1 billion to a similar effort.
According to PYMNTS Intelligence’s Enterprise AI Benchmark Report, 71% of executives at companies with more than $1 billion in annual revenue identify organizational readiness as the primary barrier to AI performance. Only 11% cited technology itself.
Forward-deployed engineers become the default enterprise AI model
Monthly job openings for forward-deployed engineers, or technical specialists, who are embedded directly into customer organizations to customize and integrate AI systems with existing operations, increased by more than 800% between January and September 2025. A wave of vendor deals followed.
OpenAI Chief Revenue Officer Dennis Dresser said in a May 11 press release that the challenge is no longer model functionality. We help companies integrate AI systems into the infrastructure and workflows that drive their business.
“Forward-deployed engineers can help you sit with your organization, sit with your users, understand your workflows, pull that functionality from your back-office applications, connect it to your models, and actually build intelligence around each workflow,” Dresser said in a May 11 CNBC report.
OpenAI and Anthropic have both launched enterprise deployment ventures.
OpenAI’s Deployment Company, founded on May 11 as a majority-owned subsidiary, has raised more than $4 billion from 19 investors including TPG, Bain Capital, and Goldman Sachs, acquired AI consulting and engineering firm Tomor, and added 150 deployment engineers.
On May 4, Anthropic formed a parallel $1.5 billion joint venture with support from Blackstone, Hellman & Friedman, and Goldman Sachs to focus on midmarket companies that lack internal resources to execute frontier expansions.
Both ventures leverage private equity backing to reach portfolio companies directly and create sales channels that bypass traditional enterprise software procurement cycles.
Social media giant Meta is also creating a new division called Enterprise Solutions. This division aims to place engineers and product managers directly within large enterprise customers to deploy AI tools. Naomi Gleit, Head of Product Meta, detailed the structure in an internal memo. Product managers will lead client engagements, data engineers will prepare enterprise data for Meta’s AI systems, and software engineers will integrate Meta’s products into existing client operations.
In announcing Microsoft Frontier Company, Microsoft Commercial Business CEO Judson Althoff said in a company blog post on Thursday (July 2) that customers are moving beyond experimentation.
“They are now focused on delivering tangible business results and demonstrating the return on their AI investments…” Althoff said in the post.
Early Frontier Company customers included the London Stock Exchange Group, Unilever, and Land O’Lakes.
Even if the implementation is successful, the fundamental data problem will not be resolved.
The simultaneous announcements will put the largest AI labs in direct competition with companies that have traditionally owned enterprise technology implementations. Accenture, Deloitte, TCS, and Infosys have each built substantial AI services practices. OpenAI and Anthropic will gain priority access to backers’ portfolio companies by deploying unique models through embedded engineering teams backed by private equity capital. The deal creates a parallel channel that can bypass traditional integrators for these accounts, Forbes reported on May 28.
For businesses, consideration goes both ways. Deployments built on a single vendor’s tools become more dependent on the infrastructure over time, even if the contract allows for the use of competing systems. Microsoft said customer data will not be used to train its models and customers can continue to run competing AI systems, GeekWire reported Thursday.
The PYMNTS Intelligence report, “Enterprise AI Readiness Gap: Enterprise Data Reveals Real Barriers to Scale,” found that data quality, governance processes, budget constraints, and opaque process ownership are each key barriers for 46% to 63% of executives at large companies. The report found that 85% of these companies report that their data remains fragmented or only moderately integrated.
Vendor-embedded engineering teams work on workflow and integration layers. The underlying data and governance gaps require a different kind of work.
For all of our coverage of PYMNTS AI, subscribe to our daily subscription AI Newsletter.
