AI in ITSM: Does size matter?

10.00 – Event opening
10.10 – Presentation 1: Small Enough to Move, Big Enough to Break – One organisation, one model, every mistake worth learning from (Mark Thomas, Escoute Consulting)
10.40 – Presentation 2: Prove it or lose it – Building a measurable AI programme with Rovo (Matthew Hallet and Dylan Bolton, Relex Solutions)
11.10 – Break
11.30 – Presentation 3: The ELM® Platform – A model for safe, equitable, and sustainable GenAI transformation (John Killough, University of Edinburgh)
12.00 – Presentation 4: Data to action in seconds – The rise of experience intelligence (Mark Bewick, HappySignals)
12.30 –  Q&A
13.30 – Event Close

Every organisation gets one of two advantages and never both. Small organisations can move. Large ones can absorb. The trick is knowing which you have and compensating for the one you do not. Most assume their governance comes along for the ride. It does not. Enlarged, it just becomes coarse, and everything you were quietly getting away with is suddenly there for the board to see. Governance has an aspect ratio, and most AI incidents are what happens when somebody ignored it.

The session sets out a test that holds at any size: can you explain it, test it, monitor it, and stop it. Four questions, no new committee. Then what each costs at ten people, at five hundred, and at fifty thousand.

This session sets out how RELEX built a measurable AI programme by treating AI as an operational layer in Support and Service Delivery rather than a series of technology experiments. Every initiative runs through a structured Jira-based lifecycle, from intake through build, deployment and measurement, so that adoption and time savings are tracked from day one. Drawing on 18 months of production experience, it covers how usage data becomes a management signal, how ROI evidence informs resourcing decisions, and why the operating model, not the tooling, drives adoption.

The session will cover to key stages of the project:

• The project lifecycle (idea to ROI)
• Turning usage and metrics into measurable ROI
• From ROI data to resourcing decisions
• Why the model works (repeatable at scale)

Universities are under growing pressure to respond to generative AI—balancing innovation with governance, security, equity of access, and escalating concerns about environmental impact. At the University of Edinburgh, these tensions were felt across the community: professional services worried about automation, academics faced uncertainty about how learning and teaching would change, and students were unsure what was permitted and how AI use would be viewed.

In this session, EDINA will share how we moved from that “dilemma” to practical direction through ELM®: the University of Edinburgh’s institution-wide, secure gateway to generative AI. We will outline the drivers behind ELM® (senior leadership sponsorship and ethics review), and the design principles that shaped the platform—safety, security, responsible use, inclusivity, responsiveness, and climate sensitivity.

We will also describe what ELM® provides in practice: equitable access for staff and students; a secure web platform offering both commercial models and locally hosted open-source LLMs (including a greener default option); direct API access to enable local innovation; and the deployment of custom chatbots for real university services.

We’ll cover how we are building organisational capability alongside the technology, including our training and support programme, and how we are beginning to quantify usage and environmental impact through platform defaults, awareness measures, and an impact dashboard in development.

Finally, we’ll share what’s next on the ELM® roadmap—expanding local and commercial model options, improving UX, and developing sector collaboration—along with practical takeaways for IT and service leaders looking to deliver secure, responsible GenAI at institutional scale.

Most IT organizations have a number, a CSAT, an NPS, an XLA score and not much else. We argue that a number is not enough: real improvement requires combining experience indicators, free-text feedback, operational data, and (where available) technical data, then making sense of it fast enough to act on.

In this session, Mark Bewick will show how IT teams can translate their experience data into plain language and get grounded, actionable answers in seconds. This is not a general-purpose language model bolted onto IT data, but domain intelligence built for it, running securely with no public internet exposure.

Mark will walk through how this shifts the day-to-day work of IT experience management, using real patterns found in the data:
Why tickets fail people, speed, communication, and accuracy are the three levers that consistently drive (or destroy) happiness, and where automation, better routing, and clearer closures pay off fastest.

The cost of “ticket bouncing” how reassignments quietly kill both productivity and happiness, and a practical approach (identify reasons → dashboard it → build ownership workflows → govern) to stop it.

Where the easy wins hide a small share of ticket types (password resets, access requests, device setup) account for a disproportionate share of lost time, and are ripe for automation.
Two persistent trouble spots, the self-service portal (hard to find and understand information) and laptops (the everyday comparison to personal devices that quietly erodes trust in IT).