In-person
On 3 November, we brought together senior leaders at The Ivy Birmingham for a breakfast roundtable on where AI is creating real business value, where it's falling short, and what it takes to move from individual experiments to value across an organisation.
The session was held under Chatham House rules, with leaders from financial services, legal, property, insurance, retail and technology around the table.
Read Leon's write-up of the morning: We know AI can do lots. Now let's prove it.
Five key takeaways
AI won’t fix a broken process
AI amplifies whatever it’s given.
The group repeatedly challenged the rush to find AI use cases. If you put AI on top of a poor process or fragmented data, you just get poor results faster. Map the process, understand the outcome you want and decide whether AI genuinely adds value. Sometimes plain automation, or fixing the process itself, is the better answer.
One particularly useful observation was that deciding not to use AI can be a successful outcome. Start with the problem, not the tool.
Guardrails accelerate innovation
Guardrails speed innovation up rather than hold it back.
The discussion challenged the idea that security, compliance and governance simply slow innovation. Clear guardrails give people the confidence to experiment safely and move faster.
Participants described how trusted environments, identity controls, engineering standards and clear responsibilities let experiments become reliable production capability. When guardrails are owned from the top and shared at every level, good ideas can scale.
A particularly important distinction emerged: telling an agent what it must not do is not the same as enforcing a technical boundary.
Make AI a team sport
Individual experiments need to become team capability.
Licences alone aren’t a strategy. AI lets individuals build solutions without the collaboration that used to be needed. Forty people building faster doesn’t necessarily make one organisation smarter. When everyone builds in isolation, work gets duplicated, knowledge fragments and shared understanding erodes.
The opportunity is to turn individual wins into value across the whole organisation through communities of practice, knowledge sharing and common standards.
You can’t sue an algorithm
Accountability always sits with a person.
AI output still needs critical thinking. Without it, the output becomes “slop” passed off as work. Agents need real oversight, and responsibility for what they do belongs with people, right up to the top.
This raised a related concern, particularly in legal, surveying and specialist professional services. Some of the junior work AI now removes is how people have traditionally learnt their profession. If AI does the junior work, how will people develop senior judgement? Organisations need to redesign development pathways so that efficiency gains don’t erode future expertise and professional accountability.
It works. Now prove it.
The next phase is about proving value, not possibility.
We know AI can do a lot. Now it needs clear ROI, a real grip on rising costs, and people with the judgement and critical thinking to make it stick.
Examples ranged from faster software delivery to more capacity in specialist assessment services. The more interesting question was what organisations do with the capacity this frees up. Do they simply reduce costs, or do they widen access, improve services, grow and deliver outcomes that were previously uneconomic? This revisited the productivity dividend from the London roundtable, with a sharper focus on measurable value.