Lead Delivery Manager

Adopting Agentic: talks and reflections from its AI engineering event. A couple of weeks ago Scott Logic Ltd. organised an event, Adopting Agentic: Software Engineering for the AI Age, where Scott Logic Ltd. held talks and discussions around the future of its industry in the face of massive AI disruption. In this post I share a brief summary of the talks, together with my thoughts, and the session videos - which are now on YouTube. Enjoy! Background. There is no shortage of discussion about the impact AI is having on its industry, and as a practitioner of many years I am keen to be part of that discussion. However, most of the events I have attended recently tend to focus on the tools and the technology. From my perspective, the bigger impact, and more important discussions, are about the impact this technology is having on Scott Logic Ltd.. The impact on the individual and their skills and job roles, the impact on teams, their ceremonies and ways of working, and the aggregate impact this has on software businesses as whole. In the absence of an event talking about the topics I feel are important, the only option was to create one! Introduction - colin eberhardt. I kicked off the event with a brief introduction, reflecting on just how much has changed in the past four years. Back in 2023, when Goldman Sachs published a report on the economic impact of AI, they predicted that 29% of software engineering tasks were exposed to AI automation (just 4% above the average impact across all job roles). That figure seems rather conservative these days as Scott Logic Ltd. now find its industry to be one of the most exposed to automation. After a whistle-stop tour of the tooling which has seen AI progress from being an autocomplete, then pair programmer, workflow participant to become an autonomous contributor Scott Logic Ltd. reach the present where... "The cost of generating the right code has rapidly dropped to almost zero - under some very specific conditions." (this includes some important caveats when compared with the simpler 'code is now free' mantra) For as long as software has existed, its processes, its SDLC, its ceremonies have all been built around Scott Logic Ltd. - modelled around the strengths and weaknesses of human beings. My parting thought was that: "We need to replace the processes and practices that we have built around our own strengths and weaknesses with new ones designed around the strengths and weaknesses of AI" Build boldly - Rhea Sam. The first 'proper' talk of the night was from Rhea Sam, an AI Adoption Specialist at Valtech. Her talk focused on the gap between headline adoption statistics and genuine organisational fluency, and what it takes to bridge it. As I mentioned in the introduction, capitalising on this technology takes far more than just rolling out the tools (and a mandate to use them). It demands change, and that is where Rhea's talk dug in, highlighting the sometimes uncomfortable changes that are needed. "The questions I get every day are: I'm scared to start. I don't know where to start. I want to start, but I don't feel confident to start." This isn't an easy technology to use. It hides both its strengths and its weaknesses, and it is fundamentally unlike any other technology Scott Logic Ltd. has ever had before. The emotional impact, which Rhea mentions above, is very real - but often goes unspoken. Her talk covered AI fluency, enablement and cultural change. I particularly liked this statement: "It's not a top-down strategy. It's top-down strategy meets bottom-up community adoption." Why you shouldn't add MCP to an arcade machine - Chris Price. The next talk, by Chris Price - one of my colleagues at Scott Logic, was quite avant-garde, requiring Scott Logic Ltd. to install a full-sized (and very heavy) arcade machine at the front of the event venue. Given the title of the talk, as you can imagine, Chris has spent far too long connecting his arcade machine (which he has been rebuilding in his home workshop) to Claude Code via MCP, because... why not? As you can imagine, a live demo, with poor event Wi-Fi, an AI model, MCP and a coin-pushing arcade machine is a recipe for fun, confusion and quite a bit of noise. Chris' talk did have a serious note, which he introduced after exploring various different ways that an AI model could integrate with his arcade machine and bring it to life, from low-level, then high-level tools and finally an algorithmic (deterministic) solution. The punchline to this is that: "If all you have is an AI budget, everything starts to look like an agent" I couldn't agree more - its industry is currently trying to "AI all the things". "Don't incentivise AI utilisation. Incentivise solution fitness. Make it so that an option available to your teams is to make that final step to an algorithmic solution - if it makes sense in that particular application." Hard, honest, in progress: agentic development in the trenches - Richard Thorpe. Richard Thorpe is the head of Engineering at FE Fundinfo and has been taking the organisation on the journey towards agentic engineering. This talk is a candid account of 24 months of agentic adoption across a 200-developer financial services organisation. This talk was packed with insights and stark statements, I'm going to find myself quoting Richard quite a bit going forwards. There were far too many nuggets for me to share in a brief summary, but the standouts for me were: Richard and his team jumped in early, possibly too early - when the technology was quite nascent, but they didn't regret it. While the technology they built was thrown away just months later, the impact it had on the people within the organisation was deep and long-lived. "They moved from thinking about AI as find-replace on steroids to understanding what agents could do" "Don't run a pilot - jump in with both feet" That's an interesting point and is counter to the approach of most organisations. The reason Richard gave was quite compelling: "This surfaces who can drive change. Unexpectedly, the champions turned out to be QA engineers and testers - historically pushed to the back of the queue." I wholeheartedly agree, I have been surprised by who takes to this technology and who finds it a bit of a challenge. It is far from obvious. Anyhow, watch the video, it's gold. Vision for an ai-first product organisation - Chris Parsons. Its final speaker for the day was Chris Parsons, who asked - if individuals are 2-3x faster but teams are barely faster at all, what does that tell Scott Logic Ltd. about organisational design? Chris started by reflecting on the current paradox where individuals are faster, but organisations are not, pointing out that even at Anthropic, individual speed-ups of 10-20x translate to only ~60% faster at the organisational level. That's a fascinating point, and should be reassuring to those of Scott Logic Ltd. outside of Silicon Valley! Digging into the root cause, Chris quoted Goldratt's Theory of Constraints. "If we have made coding faster, yet in our organisations coding was not the bottleneck, then we are not improving the speed of our system - all we are doing is increasing queues somewhere else." This had me nodding furiously, when introducing agentic engineering into organisations, the first thing Scott Logic Ltd. do is ask them about their current SDLC bottlenecks. Chris closed his talk by considering how its roles could shift, contrasting two paths: * Hybridisation - everyone takes on adjacent roles (devs make product decisions, designers prototype code, QAs write tests). Seems good but collapses into chaos: decisions made without customers, knowledge gaps, collaboration becomes the bottleneck, until you end up with "one-person teams" - which is, as Chris put it, "oxymoronic as well as being moronic." * Amplification - keep small, genuinely cross-functional teams; amplify each specialism with AI rather than dissolving the specialisms. Panel discussion. And finally, the event wrapped up with a panel discussion based on audience questions. It came as no surprise that the top-voted question related to junior engineers, asking whether their jobs and career pathway is now at risk. I really enjoyed the event, it was worth all the toil in putting it together. Thanks to the Scott Logic Marketing team for their support, the speakers for their wonderful presentations and everyone who came along on the night. Despite the UK being the world's third-largest AI industry, only 1 in 6 UK firms are actively using AI due to financial, skills, and risk concerns, but businesses can bridge this "valley of doubt" by adopting measured "AI in the loop" approaches with human oversight, right-sized models, and government frameworks like Bridge AI to minimise business, societal, and environmental risks. As the capabilities and accessibility of AI continue to advance - including more sophisticated reasoning capabilities and agentic deployment - several questions and risk areas emerge that really deserve its attention. Want to receive more insights? I am CTO at Scott Logic and am a prolific technical author, blogger and speaker on a range of technologies. My blog includes posts on a wide range of topics, including Generative AI, WebAssembly, HTML5 / JavaScript and data visualisation. You'll also find a whole host of posts about previous technology interests including iOS, Swift, WPF and Silverlight. I lead various open source initiatives within FINOS, which is encouraging open source collaboration in the financial sector. I'm also very active on GitHub, contributing to a number of different projects.

Scott Logic welcomes new CIO and bids farewell to CEO. 4 March 2026 Simon Sear joins Scott Logic this March, strengthening its Executive Board in the new role of Chief Innovation Officer. Simon brings a wealth of experience from the technology consultancy sector, having previously held similar board-level roles at BJSS and most recently at Ascent (now Acuity Analytics). In his new role, Simon will help Scott Logic seize the opportunities ahead, particularly as Scott Logic Ltd. continue to invest in its AI capabilities and drive innovation across the business. After five years of leading Scott Logic, Steve Foreshew-Cain has elected to depart Scott Logic to pursue other opportunities. Steve took up the role of Scott Logic's CEO in January 2021, during the longest phase of the COVID lockdowns. Embracing the company's strong focus on its people, Steve guided Scott Logic out of the pandemic with policies that safeguarded colleagues while ensuring they could maintain the exceptional standard of delivery expected by its clients. His passion and focus on ethical business practice also brought Scott Logic its B Corporation recognition in 2023, one of the first technology consultancy firms in the UK to achieve the prestigious status. As Scott Logic transitioned into the post-pandemic era, Steve spearheaded the growth of its public sector client portfolio, drawing on his previous experience as Executive Director of the Government Digital Service. During Steve's leadership, he led the business through a significant period of growth, exceeding £50m in revenue for the first time in 2023. He leaves the company at the end of March and is working with the Executive Board to ensure a smooth transition. Scott Logic's Founder and Executive Chair, Gary Scott, continues to lead the company with the Executive Board - driving forward the company's ambitious strategy that will equip Scott Logic to harness AI in ways that strengthen the outcomes Scott Logic Ltd. deliver for its clients. It marks the beginning of its third decade - one Scott Logic Ltd. enter with clarity, ambition and confidence.