AI & intelligent services
AI that people understand, trust and keep using.
Our researchers, designers and engineers work UX-first — shaping how people ask, understand, correct and approve what AI produces, so a system earns trust on its first day and keeps it on its worst.
User research · Interaction design · Conversational & agent UX · Trust patterns · Prototyping
01 — Why experience decides
Most AI is not abandoned for being wrong.
It is abandoned for being hard to trust.
A model can be accurate and still fail. People cannot tell where an answer came from, what the system is unsure about, how to correct it or what happens when they press the button — so they check everything twice, or quietly go back to the way they worked before.
That is a design problem, and it is the one AI programmes most often discover last. Conventional software is predictable: the same input gives the same screen. AI is probabilistic, conversational and occasionally wrong, which means the interface has to carry confidence, provenance, correction and consent in ways a form never needed to.
So we design first. Researchers sit with the people who will use the system, designers prototype the interaction before the architecture is fixed, and engineers are in the room when a simpler experience demands a different data model. The result is AI that fits how people actually think and work — not a chat box bolted onto a process.
Talk to Zealogics02 — Designing with people, not for them
Research in the room where the work happens.
WATCH THE WORK BEFORE DESIGNING THE SCREEN.
03 — What the practice covers
Six disciplines that shape how people meet AI.
Where experience decides the outcome
Copilots and assistants inside existing tools
Agent review and approval queues
Search and knowledge answers
AI-assisted forms, cases and documents
Operator and field-worker interfaces
Dashboards that explain, not just display
How the team works
Human-centred research
Contextual inquiry, task analysis and journey mapping with the people who will live with the system — the sceptics included — to find where AI genuinely helps and where it would only get in the way.
AI interaction design
The right shape for each task: a conversation, a suggestion inline in a screen people already use, a form the AI pre-fills for a person to confirm, or an agent working in the background. Chat is one pattern among many, not the default.
Trust and explainability patterns
Sources people can check in a glance, confidence expressed in words rather than scores, an honest signal when the system does not know, and a record of what it did and why.
Human-in-the-loop and agent UX
Review queues, approval moments, hand-offs between agent and person, and controls to pause, correct or reverse — designed so oversight is quick enough that people actually do it.
Prototyping and usability testing
Clickable and working prototypes on realistic data, tested with real users before build is committed and again at every release — measuring task success and trust, not just satisfaction.
Design systems for AI products
Components and guidance for the states only AI has — thinking, streaming, uncertain, failed, awaiting approval — with accessibility and multilingual use built in, so every AI feature behaves the same way across your estate.
04 — Principles we design by
Six rules every AI experience we ship has to meet.
They sound obvious. Almost every abandoned AI tool breaks at least one of them.
01
Show the working
Every answer or action shows what it was based on, in a form a busy person can check in seconds.
02
Say when it is unsure
Uncertainty is visible and honest. A system that admits it does not know is believed on the days it does.
03
Keep people in charge
Anything consequential is proposed rather than imposed, with a clear way to approve, edit, reject or reverse it.
04
Fit the existing work
AI appears where the task already happens, not in another tab people have to remember to open.
05
Design the unhappy path
Wrong answers, missing data, timeouts and hand-offs are designed on purpose rather than discovered in production.
06
Learn from every correction
Feedback is one tap away, and what people correct flows back into evaluation and improvement.
05 — How we work
Design first, then build — together.
Researchers, designers and engineers work the same problem at the same time, so the experience and the architecture are decided in one conversation.
Understand
Research with real users, task analysis and the moments where AI could help or hurt.
Shape
Interaction concepts, trust patterns and the human-in-the-loop model, sketched and challenged.
Prototype
A realistic prototype on representative data, tested with the people who will use it.
Build
Design and engineering together, with usability checks inside every release.
Refine
Usage, corrections and trust signals measured, and the experience improved against them.
06 — Where design sits
Between what the model can do and what people will accept.
UX FIRST. ADOPTION BY DESIGN.
07 — Sound familiar?
What people tell us before the redesign.
“The AI works in the demo, and our people still do not use it.”
Adoption research
“Everyone double-checks every answer, so it saves no time.”
Trust & explainability design
“We added a chatbot and nobody knows what to ask it.”
AI interaction design
“Reviewing what the agent did takes longer than doing the task.”
Human-in-the-loop UX
“Every team's AI feature looks and behaves differently.”
AI design system
“It has to work for people who are not technical.”
Human-centred research
08 — The rest of the practice
Seven more ways we work on AI.
These services are rarely bought one at a time. Most programmes start with one and pull in the others as the work matures.
Advise
AI Strategy & Advisory
Work out which AI opportunities are real, what they are worth and the order to do them in — before anyone builds anything.
Explore
Transform
AI Transformation Services
Rebuild an end-to-end process around AI — the workflow, the systems, the roles and the measures — so the gain is structural.
Explore
Automate
Agentic AI & Automation
Agents that carry multi-step work across your systems under guardrails, evaluation and human approval.
Explore
Ground
Generative AI Solutions
Copilots, assistants and knowledge systems grounded in the material you approve — with citations, permissions and review.
Explore
Build
Enterprise AI Engineering
Custom AI applications, enterprise integration and deployment into the environment your security team already accepts.
Explore
Govern
AI Governance & Responsible AI
The framework, guardrails, evaluation and audit trail that let a risk committee approve an AI system.
Explore
Operate
AI Managed Services
We run what we built — and what others built — with monitoring, evaluation, upkeep, cost management and real support.
Explore
Have a problem worth solving?
Bring us the AI tool nobody is using.
Or the one you have not built yet. We will sit with the people it is for, show you where the experience breaks, and prototype what it should feel like before you commit to a build.