Leading by design
On 21 May, OpenAI announced its acquisition of former Apple chief design officer Jony Ive’s company io in a deal worth $6.5bn. Ive, who helped create the iPod, Macbook, iPhone and Apple Watch, will “take over design for all of OpenAI, including its software”. Post-acquisition, Ive will continue working with OpenAI CEO Sam Altman to develop a family of AI-powered products. OpenAI acquired a 23% stake in io in late 2024, so this was not unexpected. As Bloomberg reported, they are not anticipating their new AI device(s) replacing the smartphone. “In the same way that the smartphone didn’t make the laptop go away, I don’t think our first thing is going to make the smartphone go away,” said Altman.
While it will be interesting to see whether OpenAI’s new hardware will be a game changer like the smartphone, or an enduring extra like the iPad, which hasn’t replaced the smartphone or the laptop, this move underlines the importance of design as a critical success factor when it comes to tech/AI adoption.
Design is at the heart of Apple’s ubiquity – the user experience, including look and feel, is why we pay more for Apple devices – and is one reason why everyone is using ChatGPT.
ChatGPT, like other popular large language models, has an intuitive and (sometimes overly) friendly user interface. These freemium models are designed to be easy to engage with, and this has surely contributed to their popularity, especially with GenZ and younger, with 92% of students (according to HEPI’s 2025 AI survey) and 80% of 13-17 year olds in the UK (according to Ofcom) using GenAI – finding their own use cases for study and leisure, rather than learning prompt engineering.
Legal design and the data paradox
This lead me to wonder whether the legal sector’s focus on prompt engineering is a temporary solution for managing how GenAI is used on the way to better designed legal AI, especially as the latest agentic AI offerings in the legal space are deliberately pointed at existing processes and proprietary data (Data is still the moat, particularly in light of the high-profile M&S and Legal Aid hacks) and avoiding hallucinations.
In a recent LinkedIn post, Michael Kennedy, head of R&D (Innovation & Legal Technology) at Addleshaw Goddard highlighted how this data paradox is holding back progress when it comes to applying GenAI to legal tasks, even though law has been in the vanguard of GenAI adoption. Because legal data is confidential/proprietary, firms cannot use foundational models for legal work, and instead are building their own workarounds – i.e. designing proprietary models for proprietary data, and focusing on prompt engineering, managing context, crafting inputs, etc – thereby breaking the feedback loop that trains and improves GenAI models. Consequently, GenAI models aren’t improving their performance in relation to legal work at the same rate as everything else, and legal still needs middle-ware and legal design to make GenAI work effectively.
While legal has always been keen on future proofing, notwithstanding the exponentially increasing number of legal GenAI experts on LinkedIn, the future is genuinely uncertain. The technology is developing so fast that even Ive and Altman who really are at the cutting edge have only said they are designing something different, and nobody knows exactly what/who it will replace. Meanwhile the next generation of lawyers and leaders – who are currently teenagers and students – are literally growing up with the new, evolving tools that are reshaping the way people use technology in their lives and work.
And as voice interface becomes more popular among GenZ and younger, whose habit of talking to ChatGPT on their phones is apparently leading to a resurgence of the teenage late-night phone call (but as well as, not instead of texting, messaging etc), it’s easy to envisage a ChatGPT device similar to Google Home or Amazon Echo. But this raises questions about data, and of course hallucinations.
Beware the AI bubble
A couple of news items in recent weeks highlighted the danger of putting too much trust in GenAI models – and AI investment hype.
A new database that tracks legal decisions derailed by GenAI hallucinated content being used in court cases (including fake citations) https://www.damiencharlotin.com/hallucinations/ has identified 136 cases (worldwide) showing the growing extent to which GenAI is disrupting court cases, not in a good way.
And while every week sees new legal AI start-ups raising massive investment, the spectacular collapse of no-code software development platform Builder.ai whose investors included Microsoft and Qatar Investment Authority (QIA) and was valued at $1.5bn in 2023 underlines the risk involved. Builder.ai claimed that it’s AI assistant ‘Natasha’ enabled users to build custom apps. It later turned out that ‘Natasha’ was not powered by AI, but by a team of human developers in India. This together with questions over the company’s revenues and accounting practices led to its speedy demise. Not all unicorns are real.
Audiobooks for early summer
I always enjoy summer reading lists. So to reflect the revival of the voice interface, I’d like to highlight a couple of audiobooks about AI/GenAI by well-known authors who are not tech engineers. The first is the audiobook version of Professor Richard Susskind’s latest bestseller, How to think about AI. I have known Richard for many years and I have read most of his books. As many reading this will know, Richard is a fantastic presenter. He has a distinctive voice and delivery style, and I enjoyed hearing him read out his own words. While all of his books are well written, this one seems more personal. It feels like he’s stepping outside his comfort zone as a pioneer of legal tech and legal AI to write more generally about his experience in the world of AI, his family and his personal beliefs. I followed this with 12 Bytes, novelist Jeanette Winterson’s well-researched and beautifully written series of essays on AI’s history, implications and potential. This was published in 2021, before ChatGPT changed everything, and again it is heavily nuanced by the author’s own perspective and beliefs, but the questions it raises remain valid.
Legal Geek are hosting four conferences this year, learn more on our events page.
Written by Joanna Goodman, tech journalist
Photo credit (Joanna): Sam Mardon
