Leaders Connected: What Practitioners Actually Said About AI

There is a format that works well for tech community events and is notably harder to get right than it looks: two fireside chats, a room of senior practitioners, enough space in the conversation for people to say things they wouldn't put in a conference talk. No vendor pitches. No slides trying to earn a consultancy engagement. Just two people who have done difficult things talking honestly about what that was like.

The Leaders Connected relaunch did this. Jody Marks and Grant Spencer from MRJ Recruitment organised and hosted the evening, and by all accounts the guest list was genuinely curated: people who work at the intersection of data, technology, and commercial operations rather than a generic networking crowd.

Anna Barsby on Building Data Systems That Actually Work

Anna Barsby, Founder and Managing Partner at Tessiant, has held senior technology leadership roles across Asda, Halfords, and Morrisons. That means she has seen the data infrastructure challenge from the inside of three significant UK retail organisations at very different stages of their digital journeys. The kind of pattern recognition that comes from that is hard to replicate.

Her fireside chat was ostensibly about data strategy but kept gravitating back to the human and organisational dimensions. Which is probably the honest version of where most data programmes actually succeed or fail.

The specific framing she used landed with most of the room in the slightly uncomfortable way that accurate observations tend to: organisations often try to build AI capability on top of data foundations that were never designed to support it, then express surprise when this does not work.

Tessiant's positioning as practitioners-turned-consultants (rather than consultants who have read about practice) matters here. The texture of experience from someone who has sat in the meeting where the finance director asks why the ERP data and the EPOS data do not match is not something you can fake, and it shows up in how advice lands.

The observation that landed hardest: data quality problems tend to get worse when you add AI, not better. Not because AI is uniquely bad, but because it scales whatever it finds. If your product data is inconsistent in the underlying systems, an AI tool that surfaces product data at speed will surface that inconsistency faster and more visibly than a human analyst ever could. This makes data quality the prerequisite, not the subsequent task. And the unglamorous piece of work that keeps getting deprioritised until the AI project exposes why it mattered.

Kevin Evans on Spend Intelligence and AI at Scale

Kevin Evans, then CTO at Rosslyn, came at the AI conversation from a different angle. Rosslyn's platform is in the business of making messy enterprise data usable: connecting, cleaning, and contextualising spend data from thousands of source systems into something a business can actually act on. Less glamorous than AI-powered personalisation, and considerably more foundational to whether any of those things work.

The conversation covered where AI is genuinely adding value in data operations versus where it is being bolted onto processes that were already working reasonably well without it. Evans' view, grounded in the reality of building and shipping production systems, was that the most durable AI applications solve problems that were previously unsolvable rather than automating tasks that were already being done acceptably.

Spend data anomaly detection is a reasonable example. The volume and complexity of cross-system spend data makes human analysis at scale genuinely impossible; it is a tractable machine learning problem once the data is clean enough to work with.

What These Events Are Actually For

A pattern ran through both talks that is worth naming: a clear-eyed acknowledgement that the hard work is usually not the AI itself. It is the data infrastructure, the organisational readiness, and the unglamorous plumbing that makes the system trustworthy enough to act on.

Events like Leaders Connected do not generate flashy takeaways. Nobody is going to post a quotable statistic from an evening like this. What they generate is calibrated understanding: the kind you get from spending a couple of hours listening to people who have navigated real problems at real scale, in real organisations, in the North of England.

There is a version of the AI in commerce conversation that happens entirely at convention-centre scale, with global platform companies talking about global platform numbers. There is the version that happens in rooms like this, which is more immediately actionable for the organisations actually making decisions. Both are worth following. The second one tends to be harder to find.

Good work from Jody and Grant for putting this back on the calendar.