The 2019 bet: automation becomes a service — and it becomes autonomous.

In 2019, automation was a project business. Companies bought RPA licences, had bots built — and were then left alone with them: maintenance, exceptions, changes, operations. Robert Aksan, by then more than two decades into enterprise programmes, drew two conclusions he wanted to build a company on. First: automation rarely fails on technology, almost always on operations — it should be delivered as a service, not dropped off as a project. He called it Automation as a Service. Second: with machine learning, these systems would not just replay steps but increasingly work on their own — understanding documents, deciding cases, handling exceptions themselves.

In 2020 the thesis became BOTFORCE GmbH, founded in Vienna. The first years were deliberately down-to-earth: RPA projects as a UiPath partner, document AI inside processes, and hard delivery work in regulated environments — a credit-claims value stream at a leading Austrian bank, EMEA cutover lead of a global S/4HANA programme across five countries. That is where the craft was forged that carries every agent rollout today: owning go-lives, managing exceptions, running systems that must not fail.

From 2022, large language models supplied the missing ingredient. What the bots had always lacked — language, context, judgement — was suddenly available. BOTFORCE first built with it internally: more than twenty products of its own, from tender intelligence to planning tools, developed and operated by its own team. Not as an end in itself, but as a training ground: whoever runs software themselves learns faster than any consultancy what holds and what breaks.

Today, the 2019 bet is reality — under a new name. Agentic AI is what “autonomous work with ML” was called back then: software that owns outcomes instead of steps. BOTFORCE designs, builds and operates these agents for regulated industries in Austria and Germany — with identity, limits, an audit trail and human oversight from day one. Because the second half of the original thesis still stands: what matters is not what the technology can do, but whether you can operate it responsibly.