Mastercard · MDES Manager
Role: Senior Technical Writer · Consultant (Documentation and AI Content) Engagement: Contract, November 2025 to present. Full time to March 2026, then a retained cadence of release-cycle work Product: MDES Manager, the tokenisation platform issuers and processors use to configure card products for Apple Pay, Google Pay and Samsung Pay Also: MIIG documentation, and a local DITA-OT toolchain
MDES is the tokenisation layer behind Apple Pay, Google Pay and Samsung Pay. MDES Manager is how customer institutions configure it. I am a contractor on its documentation, content design and AI content work.
The Situation
A 380-page user guide, interface copy from many hands, and a support queue absorbing questions the documentation should have answered. The platform was mature. The material describing it had been maintained page by page and never governed as a system.
Scope
- User guide overhaul. Rewrote the 380-page MDES Manager user guide into a guide of about 200 pages in DITA XML. Standardised terminology, tone and formatting, to reduce onboarding support tickets. I worked section by section with product managers, product and marketing, and used LLM tooling for the rewrite
- Prioritisation. Owned prioritisation for the cross-functional overhaul. I analysed 949 support tickets to decide what to fix first, weighing user impact against engineering and design cost
- Content design across about 150 configuration screens. I adjudicated competing feedback from designers, engineers and PMs against Mastercard brand standards and W3C and COGA accessibility guidelines
- RAG-ready content repository. I designed and built the repository (14 delivery guides, in GitHub) behind Mastercard's conversational-AI proof of concept. Sections are self-contained and intent-tagged, with Q&A pairs, chunked and terminology-controlled for retrieval accuracy
- Ongoing LLM-assisted review. 30 to 40 screens per release cycle against the content standards
- AI content-review agents (in scoping). Scoping agents with the MDES design team, using Figma agent skills and Copilot. Not live yet
Precision on item 4: I built the content and the content architecture the assistant draws on. I did not build the assistant.
Decisions That Shaped It
Let the support data set the priority. 949 tickets is a better guide to what users cannot work out than any internal opinion, mine included. Fixes went in order of user impact against cost.
Write for retrieval as well as reading. A section that stands alone and says what question it answers works for a person scanning the guide and for a retrieval system pulling one chunk.
Adjudicate, with the standards as the referee. Designers, engineers and PMs disagree. The brand and accessibility standards decide the argument, so they are the reference point each time.
Stakeholders
Product managers, product and marketing, designers and engineers across MDES feature areas, and the MDES design team on the AI work. The constraints came from Mastercard's brand standards, W3C and COGA accessibility guidance, and the DITA XML toolchain.
The real work. The writing is the visible part. The engagement runs on agreeing terminology across feature areas, holding a structure while the product keeps shipping, and keeping review cycles moving.
Why This Matters for AI Enablement
This is AI work inside a regulated tier-1 payments company: an LLM-driven rewrite, content designed for retrieval, and review run against standards every release. The next step is agents that apply those standards. See AI Enablement.
→ Starknet for the blockchain infrastructure contract → AI Enablement for how I work with LLM tooling → Contact to discuss a role