Cognizant will help Travelport, a major travel technology provider, upgrade its systems using Anthropic’s AI models.

Travelport operates a cloud-based database of more than three million vacation properties, airlines and car rentals. Travel agencies use the platform to plan trips for clients. Additionally, the software speeds up related tasks such as rescheduling flights.

The UK-based company will use Claude to automate certain aspects of its internal software development workflow. Cognizant will provide the technical expertise needed to turn the idea into practice. The consultancy is no stranger to Claude: it partnered with Anthropic last November to make the model series available to its workforce.

“AI-generated code and AI-driven quality evaluations are powerful accelerants, particularly in the travel industry where real-time systems, high transaction volumes and customer-facing workflows demand consistent accuracy and reliability,” Anup Prasad, the senior vice president and head of Cognizant’s consumer business unit, told Boardroom Insight.

Cognizant and Travelport will use Claude to generate software code. They also plan to apply the technology to two other aspects of the application development workflow: pull request reviews and test creation.

A pull request review is a process through which members of a development team check one another’s code for bugs. It’s usually carried out immediately before an update is rolled out to production. If a large number of developers submit code changes on the same day, pull request reviews can pile up and slow down the release cadence. Claude can help software teams work through such bottlenecks.

Test creation, the other task that Travelport is looking to speed up, is the process of writing scripts to speed up bug detection. Software teams sometimes create upwards of dozens of custom scripts for a project. Travelport and Cognizant will use Claude to reduce the amount of manual work involved in the task.

Using AI to automate test creation can lead to situations where a script designed to find software bugs itself contains errors. Cognizant says that it has guardrails in place to catch and resolve such issues.

“At Cognizant, these systems operate within a framework of human accountability, and this is a non-negotiable design principle for us,” Prasad said. “We don’t rely on LLM outputs in isolation. Our AI quality engineering model embeds human review, automated testing, validation, and adversarial red teaming across every stage of the development lifecycle, underpinned by our Responsible AI framework which mandates transparency, traceability, and auditability at every step.”

Travelport will use Claude to automate manual work for not only its software developers but also the travel agents who use its software. According to the company, the goal is to speed up tasks such as buying new airline tickets after a flight cancellation.

Embedding AI in a customer-facing software product comes with different challenges than using the technology to speed up internal workflows. One common issue is drift, a phenomenon whereby the LLM model that powers a service becomes less reliable over time. Cognizant says that it can address such challenges. 

“In production, outputs are continuously monitored for anomalies and drift,” Prasad explained. “When inconsistencies are detected, systems are designed to flag, isolate, and correct them before they create downstream impact with clearly defined escalation paths to human reviewers who validate that AI-generated evaluations align with business intent and regulatory requirements.”

Cognizant will use a technology called MCP to power some of the AI workflows that it’s building for Travelport. It’s an open-source tool that enables AI models to communicate with external systems such as hotel databases. MCP encrypts LLMs’ traffic and enables developers to implement mitigations against a range of common cyberattacks. 

“Our Travelport engagement will leverage the Model Context Protocol (MCP), an open standard pioneered by Anthropic which governs how models interact with enterprise systems and data, adding a further layer of safety, reliability, and controllability,” Prasad said. “The result is a ‘build-time and run-time’ assurance model that shifts AI from assumed accuracy to demonstrable reliability.”

Travelport expects to release the first customer-facing features developed through the partnership later this year.

Photo courtesy of Cognizant