X Mind Solutions logoX Mind Solutions
Blog

Designing Human-Centred Customer Experiences in MCP Integrations

Technical accuracy alone is not enough in MCP integrations. Using a flight search example, we explore human-centred response design for AI.

Artificial intelligence · 2025-08-05 · 3 min de leitura

Designing Human-Centred Customer Experiences in MCP Integrations

A human-centred MCP integration retains technical data while providing the AI assistant with verified explanations and comparisons aligned with the user’s purpose. In flight searches, this means offering clear ticket conditions, flight summaries and reasoned recommendations alongside codes. The aim is to help users understand and evaluate their options without requiring the model to make assumptions.

  • 5 de agosto de 2025

An AI assistant’s ability to access flight options does not mean users can easily find the right one. This is the key distinction highlighted by a test of the Turkish Airlines MCP integration developed by Turkish Technology for an Istanbul–London flight search: a technically functional connection can still leave room for improvement in the end-user experience. At X Mind Solutions, we assess this issue not just in terms of access to data, but also in terms of how that data is presented to people.

MCP is a protocol that standardises how AI applications access tools and data in external systems. Making the retrieved response understandable to users, however, is a separate design consideration. In a flight search, fields such as option identifiers, airport codes, aircraft type codes and fare classes are meaningful for communication between systems. Passengers, by contrast, usually want to know when they will arrive, how much they will pay and what conditions apply to their ticket.

The first area for improvement is presenting technical fields alongside verified explanations. For example, an aircraft type code could be accompanied by a recognisable model name, while a fare class could be shown with the change and cancellation conditions for the relevant ticket. The critical boundary here is to avoid drawing unsupported conclusions from codes. A fare code alone should not be turned into a statement guaranteeing flexible cancellation rights, nor should an aircraft model be used to guarantee a particular level of seat comfort.

The second area is linking options to the user’s purpose. Business travel, holidays and budget priorities may call for different comparison criteria. Rather than labelling a flight as ideal for business travel, however, it is more useful to explain the reasoning behind the recommendation. Verified information such as arrival time, total fare and transport from the airport helps users decide according to their own priorities. If those priorities are unknown, asking about them is a more sound approach than making recommendations based on assumptions.

The third area is practical information that connects flight data with the onward journey. Airport transfer options, weather information and price change alerts can be considered potential enhancements to the experience; they should not be presented as verified features of the existing integration. The use of such information should depend on reliable sources and checks that it is up to date. Similarly, passenger preference or satisfaction rates should only be used when verifiable data is available; illustrative percentages should not be presented as actual customer behaviour.

Designing human-centred MCP responses does not mean removing technical data altogether. Structured fields can be retained and accompanied by user-friendly summaries, clear conditions and reasoned comparisons. This approach aims to reduce the language model’s need to interpret raw codes and to present users with clearer options. At X Mind Solutions, our core design criterion is not simply whether the system responds, but whether its response is accurate, understandable and helpful for decision-making.

Perguntas frequentes

What is the difference between MCP and user-friendly response design?
MCP standardises how AI applications access external tools and data. User-friendly response design addresses how that data is presented through clear explanations and comparisons. A working connection does not automatically make those explanations accurate or sufficient.
Should technical codes be removed from responses entirely?
No. Structured technical fields can be retained to support system operations. Adding verified, plain-language explanations to these fields supports both technical use and the end-user experience.
Does a fare class code indicate flexible cancellation rights?
Flexible cancellation rights should not be inferred from a fare class code alone. The explanation presented to the user should be based on the verified fare rules for the relevant ticket.
How should a flight recommendation for business travel be justified?
The recommendation should be explained in terms of the user’s priorities, arrival time, total fare and verified transport information. If the user’s needs are unknown, the assistant should ask clarifying questions rather than immediately label a flight as ideal.
Can transfer and satisfaction information be included in a flight response?
This information can be included if it is supported by reliable, up-to-date sources. Transfer times and fares should be verified, and satisfaction rates should only be used if they are based on real, verifiable data.

Kaynak: Orijinal kaynak

X MIND WEEKLY

What happened in AI this week?

Want practical AI news for your business? The global and Turkish AI agenda, field examples from KobiGPT and automation ideas you can apply right away: 1 email a week, ~3 minute read, no spam.

After signing up, please click the confirmation link we send to your inbox. You can unsubscribe at any time. Read previous issues →