Automation Summit 2026: what we took away from two days in Split

Automation Summit 2026 ran on 15 and 16 October at Le Méridien Lav in Split, and this year Robotiq co-organised it alongside the Money Motion team. More than 500 people came in from roughly 20 countries, with speakers from CERN, Zscaler and Y Combinator sharing the programme with the people running automation inside banks, insurers and telecoms across the region.

The theme that held both days together was the gap between what AI can do in a demo and what it does once it meets a real enterprise.

Size Doesn’t Matter

Darko Jovišić opened with a talk arguing that bigger models are not automatically the right answer. For narrow, repetitive automation work, small language models deliver comparable results at lower cost, lower latency and with full control over where your data sits. The session covered where they win, where they genuinely lose, and how to tell the difference before committing to an architecture.

It Fits on a Laptop

The workshop straight afterwards put that argument to the test in front of a full room. A small language model handled real business tasks live, reading documents, extracting data and making routing decisions, running entirely on local hardware. No cloud dependency, nothing leaving the building, and costs that still make sense once the volume grows.

The AI Reality Check

Goran Ćudina moderated a panel on what actually happens when AI arrives inside a large organisation, joined by Mirjana Karačon from Hrvatski Telekom, Danira Mustafović from NLB Bank Sarajevo and Joško Šugar from OTP Banka. Three people who have been hands-on with real implementations, talking about what worked, what did not, and what changes once the technology meets the realities of an enterprise.

Three things worth taking home

Across both days, the same ideas kept surfacing from different stages and different industries.

AI is an organisational problem before it is a technical one. Successful implementation starts with clear processes, defined ownership, and people who have both the knowledge and the authority to act. Automation will not fix a broken process, it will only run it faster. Several speakers also pointed out that the biggest opportunities sit outside engineering entirely, in everyday operations where technical expertise rarely reaches.

Pilots are easy, production pays. AI is already delivering real results, from hundreds of working hours saved to better fraud detection and faster day-to-day operations. Getting there depends on proper integration, measurable outcomes and a clear business case. That also means setting a baseline before you automate anything, because you cannot prove the value of a change without knowing where you started.

More autonomy requires more control. Agents open up genuine possibilities, and they raise the bar on security, governance, data quality and human oversight at the same time. The pattern that kept coming up was letting AI propose while systems execute, with human review and traceable runs in between. From warehouse automation to defence, the message was consistent: AI can analyse, suggest and accelerate, and people still provide the context, challenge the output and make the final call.

Between the sessions

The Robotiq booth stayed busy across both days, mostly with people describing systems that have no API and no realistic prospect of getting one. Legacy applications, desktop software, internal portals that were never designed to connect to anything. That conversation has not changed much in eight years, which is part of why we keep having it.

Thank you to everyone who came to Split, to the speakers who travelled a long way to be there, and to the Money Motion team for running production alongside us.

See you at the next one.