Insights
Essays and practical guidesWhat we learned by shipping, and the methods we still use
Recent essays come from work in delivery now. Below them sits a practical archive of product, UX and sector guides retained from the previous site. Each article keeps its original URL and publication date.
Recent delivery essays
Audra Eval: How we hold our own AI work accountable
Audra Eval is our evaluation and quality-gate layer. It tests accuracy, citations and model drift before and after release.
Certified secure: Our upgrade to ISO 27001:2022
Our ISO 27001:2022 upgrade added the current control set used to govern security across client delivery.
From AI Demo to Production
A demo proves that a model can respond. Production also needs monitoring, evaluation gates, cost control and a named operating owner.
Building reliable AI document pipelines
Document pipelines fail at the handoffs between extraction, validation and review. This article explains the five-stage structure we use.
Making AI Agents Real: From Use Case to Working Sprint Prototype
The prototype includes the agents and the interface a person uses to supervise them. Four weeks on sanitised data produce a decision and a path to production.
The complexity of multi-channel patient recruitment platforms in clinical trials
Clinical-trial recruitment spans referrals, advertising and digital platforms. The difficult part is coordinating those channels without losing eligibility context.
Why our sprints run on sanitised data
The most common reason an AI project has not started is not budget. It is that nobody can get access to the data yet.
The number in the case study is never the interesting part
Every case study here has a number in it. They are true, and they are the least useful thing on the page.
The build-versus-buy question flipped. The harder question is what you want to own.
Two weeks is the number that travels. It is the least interesting thing about it.
What five agents actually do to a clinical intake file
Five agents, each with one job. Splitting the work that way was never about speed.
What "good enough to ship" means when the output is clinical
Most AI evaluation writing is about models. In health the useful question is about gates.
Running clinical AI under HIPAA and GDPR at the same time
Health projects arrive with both regimes attached. It is cheaper to treat that as one architecture question.
The KYC agent was the easy part
Automating the check is solved. Getting a compliance officer to accept the output is where the project lives.
What a design sprint for a bank has to do that a start-up sprint does not
Same four weeks, same process. What differs is what has to exist at the end.
Practical archive
41 product, UX and sector guides published before 2024 and retained from the previous site. They describe methods still in use; they are not accounts of current AI delivery work.
- What is innovation strategy? Overview, key benefits, and examples2023-12-21
- 10 UX best practices for optimal user experience in 20262023-11-03
- How to create a service blueprint2023-03-01
- Customer journey mapping: The Windmill guide to design thinking2022-08-23
- Expert interviews: The Windmill guide to design thinking2022-08-09
- All about Wealthtech: Past, present, and future2022-05-27
- Six Wealthtech apps with outstanding UX2022-05-11
- Psychology of color in financial app design2022-04-26
- The five design sprint phases for start-ups2022-02-28
- All about Fintech: History, development, and future2021-12-17
- How workflow diagrams expose weaknesses in UX design2021-12-15
- Three frameworks to understand your competitive environment2021-12-06
- Competitive analysis: The Windmill guide to design thinking2021-11-26
- Business model canvas for start-ups: The Windmill guide to design thinking2021-10-20
- 4 great feature prioritization methods for an MVP2021-06-29
- 12 questions a designer should ask before embarking on a project2022-06-30
- Enterprise Engineering: Simplifying a Connected Technology Estate2020-08-05
- Customer experience vs. user experience: Why the difference matters2023-11-24
- Figma tutorial: How to import a project from Sketch to Figma2021-08-26
- How to develop a product vision2023-01-25
- Five challenges for a self-sufficient team2019-06-24
- The dark side of UX: Deceptive patterns2023-09-29
- Desk research: What it is & why it’s essential in product development2023-04-13
- How to frame a complete and actionable product vision in 2 weeks with design sprints2020-07-31
- What is digital transformation?2023-09-01
- Effective UX research part 1: Preparing for research2021-01-21
- What happens after a design sprint workshop for start-ups?2022-03-03
- Building user empathy: The Windmill guide to design thinking2021-11-17
- How Wealthtech is democratizing investing2022-05-10
- Information architecture: Mastering the structural design process2021-05-05
- Effective UX Research Part 3: Analysis and Reports2021-06-04
- Accessibility 101: Design principles a designer must know2020-11-09
- Windmill’s Design Thinking Method: Three Tests Before Delivery2020-08-07
- Accelerating WealthTech innovation with Topaz, Windmill and AI2023-02-24
- How to build a finance app2022-03-22
- Topaz Digital: enabling Neo banks2023-02-14
- What is impact mapping & how is it used in product development?2023-07-24
- Design brief: What is it & how to make one (with examples)2023-08-16
- User interviews: The Windmill guide to design thinking2022-08-16
- Fintech vs Wealthtech2022-04-25
- Product Design: From Problem to Production2023-10-13