OpenAI just admitted what every CIO already knows: buying AI is easy. Deploying it at scale is the hard part.
The company's new partnership with Infosys and six other global systems integrators (Accenture, Capgemini, CGI, Cognizant, PwC, and TCS) through its "Codex Labs" program reveals the real enterprise AI playbook — and it's not what the marketing brochures promised. While Codex now has 4 million weekly active users (adding 1 million in just two weeks), moving from pilot programs to production-ready deployments requires something AI can't automate: human expertise.
The Numbers Tell the Real Story
Infosys reported $267 million in AI-related revenue in Q4 2025, representing 5.5% of total revenue. That's real money from real deployments across 4,600 active AI projects and 200 autonomous AI agents. But here's the catch: Infosys stock is down 22% this year, along with the broader Indian IT services sector, which lost $56 billion in combined market value after AI-driven automation fears sent investors running.
The irony? The same companies being threatened by AI automation are now the distribution channel for AI vendors like OpenAI. It's classic "if you can't beat them, join them" — but with a strategic twist.
Why IT Services Firms Still Matter in the AI Era
TCS has $2.3 billion in annualized AI revenue. HCLTech hit $620 million. These aren't small pilots anymore. But the market still doesn't know how to price AI's impact on traditional IT services margins. AI coding assistants deliver 15-30% productivity gains, which sounds great until you realize clients expect to pocket those savings through reduced headcount and compressed timelines.
Here's what changed with the OpenAI-Infosys partnership: Instead of fighting AI deflation, IT services firms are betting they can capture the deployment layer. While AI can write code, it can't navigate enterprise change management, legacy system integration, security compliance, or the political reality of getting 5,000 developers to adopt new tooling.
Nandan Nilekani, Infosys co-founder and chairman, framed it clearly: "While AI agents can automate tasks and enhance productivity, enterprises still require deep systems integration, governance, trust frameworks, and large-scale transformation capabilities to fundamentally re-engineer their businesses."
Translation: The technology is a commodity. The deployment expertise is not.
The Codex Labs Distribution Model
OpenAI's Codex Labs program is essentially a channel strategy wrapped in enterprise support. Partner firms get workflow-specific demonstrations, hands-on implementation guidance, and adoption support. In return, OpenAI gets access to the client relationships and delivery capabilities of firms operating in 60+ countries.
Early enterprise use cases show where the real value is:
- Virgin Atlantic: Increased test coverage and accelerated code reviews
- Ramp: Building new features faster with AI-assisted development
- Cisco: Analyzing large code repositories and aiding incident response
- Rakuten: Leveraging Codex beyond coding for knowledge work (briefs, plans, follow-ups)
Notice what's missing from that list? ROI numbers. We know Codex users are growing rapidly (4 million weekly actives), but we don't know what enterprises are actually saving or how long deployments take to reach production.
The Enterprise Deployment Gap
Infosys is combining OpenAI's Codex with its Topaz Fabric platform — a composable, agentic services suite designed to move enterprises "from AI experimentation to practical, responsible deployment and measurable business outcomes." That's vendor-speak for: We've seen too many enterprises get stuck in pilot purgatory.
The partnership focuses on three deployment areas:
- Software engineering and legacy modernization — where technical debt meets AI automation
- DevOps workflows — automating the deployment pipeline itself
- Multi-step business tasks — claims processing, code generation, document analysis
Here's the real question CIOs should ask: If AI delivers 15-30% productivity gains, why do you need a $300-400 billion AI services market to deploy it? (That's the total addressable market Infosys cited for AI-powered transformation.)
The answer: Deployment costs dwarf licensing costs. Integration with proprietary codebases, change management for developer teams, security and compliance frameworks, and ongoing optimization aren't one-time expenses. They're permanent overhead.
What This Means for Enterprise Buyers
If you're evaluating AI coding assistants, the OpenAI-Infosys partnership reveals three critical insights:
1. Vendor Lock-In Is Real
Integrating Codex with your internal systems isn't a weekend project. Once you've customized it for proprietary codebases and trained teams on specific workflows, switching costs are high. Plan accordingly.
2. Services Revenue Will Exceed Software Licensing
If Infosys is making $267 million in AI services revenue (5.5% of total), and OpenAI is building a partner network to scale that model, expect deployment costs to be 3-5x your annual licensing fees. Budget for multi-year engagements, not SaaS subscriptions.
3. The "Pilot to Production" Gap Is Where Most Enterprises Fail
Accenture's Chief AI Officer Lan Guan said Codex enables "moving from static requirements to working solutions in hours." That's true in demos. In production, you're still dealing with legacy systems, security reviews, compliance audits, and developer adoption resistance. The 7-partner Codex Labs network exists because that gap is where deals die.
The Competitive Landscape: Who Else Is Betting on IT Services Distribution?
Infosys isn't the only IT firm hedging its AI bets through partnerships:
- Infosys + Anthropic (announced February 2026) — building enterprise-grade AI agents
- HCLTech + OpenAI (prior partnership) — focused on enterprise AI deployment
- TCS (standalone strategy) — investing heavily in AI transformation, betting on bundled deals
The pattern is clear: AI vendors need distribution. IT services firms need to stay relevant in an AI-automated world. The partnership model solves both problems — at the enterprise's expense.
The Unanswered Questions
Despite all the announcements and revenue numbers, key enterprise questions remain:
- What's the actual ROI from Codex deployments? We have productivity metrics (15-30% gains) but no total cost of ownership data.
- How long does it take to reach production at scale? Moving from 100 developers to 5,000 developers isn't linear.
- What happens when AI deflation catches up to services margins? If clients can do more with fewer people, IT services revenue compresses — even with AI-powered offerings.
- Who owns the customization IP? When Infosys integrates Codex with your proprietary codebase, who controls the resulting workflows and configurations?
Until vendors and their partners publish real case studies with financial outcomes, enterprise buyers are making deployment bets on incomplete data.
Bottom Line: AI Deployment Is the New Consulting Goldmine
OpenAI's Codex partnership strategy with Infosys and six other systems integrators reveals the real enterprise AI business model: Software vendors provide the technology. IT services firms provide the deployment expertise. Enterprises pay both — repeatedly.
For CIOs and CTOs: The 4 million weekly active Codex users prove developer demand is real. But rapid user growth doesn't mean painless enterprise deployment. Budget for services costs that dwarf licensing fees, plan for 12-24 month rollouts (not 90-day pilots), and demand ROI transparency from both your AI vendor and your systems integrator.
For CFOs: The $267 million Infosys made from AI services in one quarter shows this isn't vaporware. But the 22% stock decline also shows the market hasn't figured out how to price AI's long-term impact on IT services margins. Don't assume deployment costs will come down as the technology matures — they might actually increase as complexity grows.
The real winner in this partnership? Not OpenAI (they still need distribution). Not Infosys (they're still fighting AI deflation fears). It's whichever enterprise figures out how to deploy AI at scale without multi-year consulting engagements eating their productivity gains.
Good luck finding that playbook. Nobody's published it yet.
Want to calculate your own AI ROI? Try our AI ROI Calculator — takes 60 seconds and shows projected savings, payback period, and 3-year ROI.
Continue Reading
- Why Enterprise AI Projects Fail: Lessons from the Field — Understanding the deployment gap between pilots and production
- The Hidden Costs of AI Adoption: TCO Beyond Licensing — What enterprises actually spend on AI transformation
- IT Services vs. AI Automation: Can Both Win? — How firms like Infosys are navigating the existential threat
Source: TechCrunch
