New design patterns for AI

Sarah Gold
Writing by IF
Published in
4 min readJan 31, 2024


16 new design patterns that empower people

Each of the patterns focuses on the core trust, collaboration and robustness challenges in products and services that use AI, including generative AI.

The new patterns, graphics designed by the goat David Marques.

There are patterns from model provenance, showing confidence to digital proofs for AI.

Each of these patterns are things that we’ve helped clients research or implement, or that we see in the AI systems we use. As always, our emphasis is on patterns that actually empower people.

We need this work, now

This is more important than ever.

We are at a constellation of inflection points about what kind of AI enabled products and services we will use, and at the same time it is harder for us to make good trust choices for ourselves, the organisations we work for, and the people we care for.

AI makes trust more complicated

That’s partly because technology, like AI, makes trust more complicated, in several ways:

  • AI makes it nearly impossible to distinguish whether what we are talking to is a human or a bot. That’s contributing to big societal and business problems from cyber security to individual harms like fraud.
  • Fast changes in UI and more personalisation make it harder for people to understand how a service works, and so seek redress when something goes wrong. AI accelerates this fast moving, personalised service paradigm.
  • Many of us lack useful mental models of how systems work, and in the absence of useful mental models we will each rely on folk stories. So we don’t know how to access our rights or take control of an AI system, increasing fear and mistrust.
  • Generally, we are overly trusting of technology, believing that AI is more correct than a human and is unbiased (neither of which is true). Combined with the way that many AI systems are designed to feel friendly or delightful, it’s likely that our over-trust of AI systems will continue with bigger and bigger consequences.

…Ultimately, that lack of trust makes it harder for teams to innovate successfully or responsibly.

Patterns can help create trustworthy services

Design patterns are a useful way to start solving these problems.

A design pattern is what you get when someone solves a common problem and they do it in a way that’s easy for other people to copy.

We use patterns across everything we design, from cars to public services. That’s why you see the same things pop-up again and again, like how to log in or to add credit card details. Which is another benefit, design patterns help build familiarity with new experience paradigms.

But we don’t only need patterns that help make the user experience easier. We need patterns that also add just enough friction to:

  • Demonstrate just enough information about the underlying system and gubbins so people are empowered to make better choices.
  • Humanise AI’s complex systems to reduce the frustration people can feel when using automated tools.
  • Help people understand why and how a service has changed, so they feel safe and respected in their experience.

… and more!

Patterns are a win for policy

Supporting the development of design patterns that demonstrate trustworthiness, is a powerful way to inform more effective policy and compliance.

Prototyping patterns lets teams communicate a complex policy by showing how it could work to industry and users. Testing patterns also helps to quickly uncover the trade offs and riskiest assumptions, so that industries establish what good looks like, faster. That means we can create better policy and services, that create better outcomes for people and organisations.

For example in the pattern of “watermarking” (which is part of China’s Cyberspace Administration regulatory requirements, and the USA and EU are set to follow) the team at IF exposed some of what is hard about making watermarking work, for users and businesses. I wonder what would have happened had policymakers prototyped watermarking ahead of pushing it. I suspect we would have watermarking that is more fit for purpose, and easier to implement…

And a win for business teams too

These problems of trust, collaboration and robustness are only going to get harder to solve as we use more AI systems, everywhere.

People have high expectations of trustworthy organisations, and they expect responsible solutions across the private and public sector. For example, we know that Gen Z’ers expect technology to understand them, and anticipate their needs, but that it cannot be at the price of privacy. So understanding how to best deliver user value whilst demonstrating trustworthiness is more and more important.

The organisations that understand this, and start investing in how to differentiate their products and services with trustworthiness, have the advantage.

Get in touch

We work with private and public sector clients around the world on some of the most challenging implementations of AI systems, and have done since 2016. Our speciality is making trust actionable, transforming principles into execution that makes impact.

Do get in touch if you want to understand more about our work, how we can help, or want to pilot one of these patterns. Email me here:

Thanks to everyone at IF for making the next iteration of the design patterns catalogue happen, and special thanks to Valeria Adani, Mark Dunbavan, David Marques and Peter Wells. Team work makes the dream work!



Sarah Gold
Writing by IF

Designing for trust. Founding partner and CEO @projectsbyif