AI customer journeys need decision-ready data

Virgin Atlantic is using ChatGPT Work to connect signals across the customer journey, from research and product planning to analytics. For enterprises, the relevant point is not the chat interface. It is whether scattered data becomes a decision-ready view of a business case.

From switching dashboards to a decision packet

According to OpenAI, Virgin Atlantic employees previously had to gather information across dashboard tools, authenticated workspaces and report suites. With ChatGPT Work and ChatGPT Sites, the company is building authenticated interfaces that make different datasets accessible in one view. For competitive analysis, the digital-product team structured its research in a framework; work that had previously taken weeks could then be ready for business review within hours.

That does not prove that every AI output is correct. It does illustrate a useful pattern: AI can accelerate research, synthesis and preparation, while prioritisation and investment decisions remain owned by the business.

The customer journey is also a data product

In April, Virgin Atlantic also launched a ChatGPT app for flight search. It presents suitable options and directs customers to the website or mobile app to complete the booking. The separation is revealing: AI supports orientation and selection, while the transaction remains in the controlled booking system.

For DACH organisations, the lesson is that a service, sales or operations agent should not simply receive access to every source. It needs a defined decision packet: approved data sources, freshness, role permissions, a visible rationale for the recommendation, and a clear handoff to the system of record.

Governance starts before the response

NIST frames AI risk management as work spanning the design, development, use and evaluation of an AI system. For customer-journey use cases, this is tangible. Teams should define which customer signals may enter the view, who may see them, when a recommendation needs human review, and which decision may actually be executed.

Start with one recurring process, such as prioritising service issues or analysing abandonment reasons. Measure more than response time and adoption. Measure whether the combined view produces faster, explainable decisions.

The central question is: Which data needs to come together so that an accountable person can decide now?