Cognizant and Cognition are expanding their enterprise AI engineering collaboration after deploying Cognition’s Devin AI software engineer at Odyssey Logistics, where the companies report a 37% net cost saving on a legacy modernization project. The deployment illustrates how autonomous AI coding agents are moving beyond code assistance into governed software engineering workflows involving migration, testing, documentation and deployment.
AI agents move into enterprise modernization
Enterprise software modernization has long been constrained by the amount of engineering work required to convert legacy applications without losing the business logic embedded in them. Cognizant and Cognition are targeting that problem with Devin, an AI software engineer designed to handle multi-step development tasks alongside human engineers.
The companies began working together in 2024 and entered an AI engineering collaboration in January 2026. Cognizant says it has trained more than 3,000 associates on the Devin platform and has also used the technology internally.
The latest deployment provides a more concrete example of the model. Odyssey Logistics was modernizing the technology supporting its transportation operations, including legacy Microsoft Access and Visual Basic for Applications forms whose underlying business rules needed to survive migration.
Rather than conducting a conventional rewrite, a seven-person Cognizant engineering team used Devin to assist with the conversion into cloud-native applications. The agent was also used to generate tests, deployment pipelines and documentation. Cognizant says every change was reviewed and approved by a technical lead or architect before being merged.
The result, according to the companies, was roughly one-third higher delivery throughput and a 37% net cost saving compared with the conventional approach.
Those figures are company-reported results from the Odyssey deployment rather than an independently audited benchmark.
From coding assistance to autonomous engineering
The significance of the deployment lies in the scope of work assigned to the AI system.
Traditional AI coding assistants generally focus on generating or completing code in response to developer instructions. Agentic coding systems are designed to operate across a larger sequence of software-development tasks, including planning, implementation, testing and iteration.
Gartner’s 2026 research describes enterprise AI coding agents as moving from AI-assisted development toward agentic software development spanning the software development life cycle. The research identifies coding, testing, debugging, code translation, documentation and requirements management among the capabilities emerging in this market.
Cognition’s Devin sits within that broader category. The company’s positioning is not simply about producing code faster, but about assigning an agent a multi-step engineering objective and allowing it to work through the associated tasks while human engineers retain control over approvals and production changes.
That distinction becomes particularly relevant for legacy modernization, where the challenge is often understanding and preserving existing behavior rather than writing new code from scratch.
Odyssey provides a modernization test case
Odyssey Logistics is developing its OdysseyONE platform to consolidate order intake, shipment execution, visibility and lifecycle management across its transportation network.
The modernization effort involved legacy Microsoft Access and Visual Basic for Applications forms containing business logic that needed to be preserved. According to Cognizant and Odyssey, the companies first tested the AI engineering approach on a separate standalone legacy application. That project was completed 21 days after kickoff, after which Odyssey applied the approach to a larger transportation-management modernization effort.
The current phase is now in production, according to the companies.
Cognizant says Devin handled portions of the legacy conversion while generating associated testing, deployment and documentation artifacts. Human technical leads and architects remained responsible for reviewing and approving changes.
That governance layer is important because autonomous software engineering introduces a different risk profile from conventional code assistance. An agent capable of modifying multiple files, running tests and iterating on its own can increase throughput, but errors can also propagate across a larger workflow if validation and permissions are weak.
The enterprise coding-agent market expands
The deployment comes as enterprise AI coding agents become a distinct software category.
Gartner estimates the enterprise AI coding-agent market at approximately $9.8 billion to $11 billion in annualized value as of April 2026. Gartner also reports a net average productivity gain of 19.3% among engineering leaders surveyed, although the figure represents the broader enterprise AI coding-agent market rather than Cognizant’s or Cognition’s specific technology.
Gartner’s research also identifies Cognition among the vendors evaluated in its 2026 Critical Capabilities research for enterprise AI coding agents, alongside companies including Anthropic, Amazon Web Services, GitHub, Google Cloud and OpenAI.
The competitive landscape is consequently shifting from autocomplete and developer copilots toward systems capable of orchestrating work across the software development life cycle.
McKinsey’s research points in a similar direction. Its 2026 analysis says software engineering and IT are among the functions where scaled AI-agent use is most advanced, while its research on AI-enabled software development emphasizes that larger gains depend on redesigning development processes rather than simply adding AI tools to existing workflows.
Cognizant adds enterprise governance and delivery
The partnership also illustrates a second development in enterprise AI: specialized AI agents increasingly being packaged with technology-services expertise.
Cognizant is positioning itself as the accountable delivery partner while Cognition supplies the Devin technology. The companies have expanded their joint offering beyond modernization to include quality engineering, where Devin can identify test-coverage gaps and assist with test generation, as well as a Frontier AI Cyber Defense offering.
They also plan to explore applications in security remediation, AI assurance, observability-driven application maintenance, legacy modernization and greenfield development.
This model places autonomous AI inside existing enterprise delivery structures rather than treating it as a standalone developer tool. For large organizations, that can matter as much as the underlying model capability because modernization projects require security controls, architecture review, testing, change management and accountability.
From AI-assisted coding to AI-managed workflows
The Odyssey project points toward a broader change in enterprise software engineering. AI agents are increasingly being evaluated not only on how much code they can generate, but on how much of the development workflow they can coordinate and execute under human supervision.
Gartner expects agentic coding to continue expanding across the software development life cycle, while its 2026 research emphasizes governance, observability and AI-ready architecture as important foundations for production deployment.
For Cognizant and Cognition, Odyssey provides a production example of that approach applied to legacy modernization. The reported 37% cost saving is specific to that project and approach, but the larger technology story is the integration of autonomous engineering into a governed enterprise delivery model.
If these systems can reliably handle more of the repetitive work surrounding migration, testing and maintenance while engineers retain control over consequential changes, software modernization could increasingly become an agent-orchestrated workflow rather than a predominantly manual redevelopment project.
Market Landscape
Enterprise AI coding agents are expanding from code completion into planning, code translation, testing, debugging, documentation and broader software delivery. Gartner estimates the enterprise AI coding-agent market at $9.8 billion to $11 billion annualized as of April 2026 and reports a 19.3% net average productivity gain among engineering leaders surveyed.
The market is also moving toward agentic workflows that can execute multiple development tasks in sequence or in parallel. Gartner says more than 65% of engineering teams using agentic coding are expected to treat IDEs as optional by 2027, reflecting a potential shift toward automated development platforms.
At the same time, production adoption requires governance. Gartner identifies AI-ready architecture, observability and AI engineering capabilities as foundations for scaling enterprise AI agents, while McKinsey’s research indicates that meaningful software-development gains are associated with organizations redesigning workflows around AI rather than simply deploying coding assistants.
Top Insights
- Cognizant and Cognition are applying Devin to enterprise modernization, extending AI coding from assistance toward multi-step autonomous engineering workflows.
- Odyssey Logistics reports a 37% net cost saving and roughly one-third higher delivery throughput on its modernization project.
- Human technical leads and architects reviewed and approved Devin-generated changes before production, maintaining governance over autonomous engineering work.
- Gartner estimates the enterprise AI coding-agent market reached $9.8 billion to $11 billion annualized by April 2026.
- The collaboration is expanding into testing, cybersecurity, application maintenance, security remediation and greenfield software development.
Power Tomorrow’s Intelligence — Build It with TechEdgeAI
