What I work in, by function.

Research and user-centred design

Every engagement starts with people, not screens. Interviews and contextual inquiry to see the work as it is done, personas built from data rather than assumption, journeys and service blueprints to find where the manual steps and the waiting sit, then testing on prototypes before anything is built.

The University of Antwerp programme is the fullest written example: thirteen to fifteen users recruited by segment, ten task scenarios, moderated and recorded sessions, card sorting for the information architecture, and repeat measurement once live.

UX and UI design

From information architecture and user flows through low, medium and high fidelity wireframes to final UI and production handoff. I do the wireframes and the journeys myself. Fidelity rises only as questions get answered: a low fidelity frame settles the structure, a high fidelity one carries the client's brand so a review is of something that looks like the product.

The Betfred templates from November 2010 on the site were drawn in Visio; the 2017 TMHCC screens in Axure and Sketch; the 2026 work in Figma and directly in code.

Design systems and accessibility

A design system is how seventy broker sites, three hundred and fifty credit unions or twenty-eight thousand pages stay consistent without a designer looking at each one. The style guide is written when wireframing begins and kept current through delivery; the component library carries every state, and accessibility is a property of the components rather than an audit at the end.

Product management

Discovery to definition to a backlog a team can estimate, with the number each initiative has to move written down before it starts. Build, measure, learn is the operating rhythm: a hypothesis, an experiment, a reading, a decision. Personalisation and experimentation platforms are product tools here, not marketing ones.

Betfred ran more than 250 build, measure, learn initiatives a year on this cadence. The three horizons model and the innovation pipeline are on the Approach page.

Project and delivery management

The plan, the budget, the RAID log, change control and the status report to the sponsor, owned by me on every engagement rather than handed to someone else. Agile squads by product line where the organisation can support them, sprint and release planning with UAT and QA gates before each release, and offshore and third-party partners managed to estimate and acceptance.

The Avios digital target operating model on the site shows the squad structure, the fast track and the release model as they were designed for a client.

Development and engineering

I build with AI tooling rather than by hand, and the products are real: SoloBusinessAI, FloraLoop, Cura Mirai, the ShelterPoint demonstration and the DXP Auditor all run on the stack below. Knowing the stack is what lets a design conversation with engineers be about trade-offs rather than translation.

Platforms: CMS, DXP, commerce and search

Sitecore is the deep one: I led the practice across fifteen enterprise clients, built the architect team and ran the partner relationship. The others range from platform selection and architecture to hands-on implementation, and the list is what a multi-site estate in a regulated business tends to run on.

AI tools and conversational platforms

Two eras. Before large language models, conversational systems on intent-model platforms where every dialogue path was authored by hand, deployed for Channel 4, Adecco, Ofcom and others. Since 2024, products built with model APIs directly, with grounded retrieval, output guards, deterministic rules in front of the model, and evaluation corpora with regression tests.

SoloBusinessAI runs twenty-four coordinated agents; ShelterPoint's assistant is trained in five knowledge layers with a compliance boundary; the DXP Auditor is eighteen audit agents scoring a site.