OpenAI rents IBM's consultants, and with them the enterprise default

OpenAI said Thursday it is embedding its technology into IBM's consulting arm, with IBM's forward-deployed units trained through the OpenAI Partner Network to help enterprise clients modernize workflows ^[1]. IBM is also joining OpenAI's "elite partner tier" and launching a dedicated OpenAI Practice staffed by thousands of consultants and engineers holding OpenAI certifications ^[2].

Read that as a channel deal. Forward-deployed engineers have become the sought-after role in applied AI ^[3], and OpenAI cannot hire enough of them to sit inside every regulated bank and insurer. IBM already has those seats. Colleen Kapase, OpenAI's vice president of global partnerships and ecosystem, told The Deep View that the partnership is a "significant" opportunity because IBM works with enterprises across virtually every major industry, "including many of the complex and highly regulated environments where AI can have enormous impact" ^[4].

The three declared workstreams show where the default gets set. First, updating legacy operations with OpenAI frontier models including GPT-5.6, plus Codex and ChatGPT Work, across finance, procurement, customer operations and HR ^[5]. Second, modernizing legacy applications and software development with Codex and ChatGPT Work, aimed at faster shipping ^[6]. Third, expanding IBM's role in Project Daybreak, pairing OpenAI's cyber capabilities with IBM Autonomous Security, IBM's agentic security platform, to manage both cyber and AI model risk ^[7]. Of the four named products spanning those workstreams, three are OpenAI's ^[8].

The stated problem is real. Andy Baldwin, global senior vice president of IBM Consulting, told The Deep View that most organizations no longer struggle to get access to models, and that the challenge is putting AI to work securely at scale across large, complex organizations while facing more sophisticated AI-enabled threats ^[9]. Fragmented legacy systems poorly fitted to the pace of AI are, per Baldwin, the hurdle this is meant to clear ^[10].

The consequence is the part worth pricing. A modernization engagement is where architecture decisions get frozen for years, and the engineer writing the migration plan now carries a vendor certification. The Deep View's own read is that the deal could help clients balance adoption, security and cost pressure, or lock them into OpenAI products rather than cheaper alternatives such as open and custom models ^[11]. Note also the shape of the security workstream: the supplier of the models sits inside the practice that assesses model risk ^[7].

There is a context argument for moving fast anyway. Models from Anthropic, OpenAI and Meta have all shown cybersecurity capabilities, and accidental breaches, that surprised the researchers testing them ^[12]. Whether that argues for buying defense from one of those vendors is a separate question.

What to watch: whether the OpenAI Practice's certification path stays alongside model-agnostic delivery or displaces it; whether IBM statements of work name OpenAI products as deliverables rather than the capability being built; how the Project Daybreak work separates model supply from model risk assessment; and what an exit costs a client two years into a Codex-based application rewrite ^[6]. Also watch for the same elite-tier structure appearing at other integrators, which is how a preference becomes an industry default. Terms, revenue split and client pricing were not disclosed in the announcement as reported ^[13].