Last week, Loomery gathered Product and Technology leaders from a diverse set of organisations to hear their perspectives on where AI is actually heading.
Is Vibe Coding going to take over all engineering? Will AI spawn a new, more human interface ? Can AI fundamentally transform organisations and how work is done... or just pump up some efficiencies?
We had some shared perspectives, some differences of opinion, and incredible stories of making progress in the real world with the help of AI. And what’s stopping that from happening too.
Here are five things we’ve been reflecting on since.
1. Adoption is broad; full transformation is rare Most are now using AI tools like Copilot, ChatGPT, or other internal copilots daily. The general sense from the group was that productivity gains are real, particularly in writing, research, and quality assurance. But in most cases, teams are simply automating what they already do rather than reimagining how the work should get done. AI generated meeting notes are now routine, yet few teams stop to question whether those meetings were necessary in the first place.
Having said that, unlike our previous event, we did hear about a Vibe Coded product in production at a major organisation. While it was clearly in a specific context (internal users, bounded complexity, specific data), it feels like the prototype-product rubicon is blurring for many now.
2. AI outputs are butting up against human bottlenecks that can’t be automated away “Just use AI” is quickly becoming a default suggestion, alongside increasing volume of outputs and compressed timelines. Our group described a growing cognitive load and time pressure on their teams as they generate more content and even more to review. In these new cycles, humans aren’t unnecessary but end up becoming “the squishy bit” in the loop, responsible for judgment, redirecting effort and ultimately accountability.
At worst, this all adds up to a squeeze against human limitations and the hard limit of working hours in the day.
3. Baseline communication quality is going up; below the surface, experts are still critical LLM’s core strengths mean communication quality overall is improving: drafts are clearer, typos are fewer, and the baseline standard is higher. Yet complex work in areas like technical architecture, security, and regulated domains still requires expert oversight.
The “70% problem” - identified by Addy Osmani at Google - appears far beyond code: AI can take you a long way, but the last mile remains firmly human-led.
Third era of personal computing, or more of the same smartphone dominance?
4. Emerging user behaviours are giving us clues about what a third era of computing could look like Users now expect many interactions to behave like ChatGPT: conversational, context-aware, and forgiving. If they use web-chat and it doesn’t work in this way, frustration can grow sharply. ChatGPT has effectively laid down a new expectations marker for conversational computing.
Some of our attendees are starting to see AI agents acting as “users” of their websites, and are actively planning ways to reshape products in response to this shift. We also heard about how younger audiences are defaulting to voice-first interaction, a behaviour that is increasingly influencing how teams communicate internally as well.
Loomery 's thinking on what the next era of computing might look like feels directionally right, but there’s so much to-be-defined, and clearly lots of pockets of different activity rather than a single dominant approach locking-in yet.
5. Rapid change is bounded by risk, compliance, and data reality Highly regulated sectors are adopting AI cautiously, using it primarily for triage, prioritisation, and assistance rather than for automated decision-making.
Widespread enterprise adoption might be following a pathway reminiscent of cloud: inevitable but slower than the hype suggests, constrained by security, governance, and data readiness.
As Matt Clifford has put it - "there are no AI shaped holes lying around" - and the largest untapped value is likely to come not from swapping tools but from redesigning workflows altogether.
What next? Fuelled by a fine breakfast and lively discussion, all of the group agreed this is an area that isn’t standing still. As ever, in a highly uncertain environment, our best bet is to keep experimenting and invent the future we want rather than letting it happen to us.
If you are thinking about these sorts of challenges, prototyping into the face of them, and want to meet other Product and Technology leaders who are too, we’ll be running another AI Bulls & Bears breakfast on 29th January next year.
Get in touch with us if you’d like to join us in the crypt next time 🧛
