The king is dead, long live the kings

Just a year ago, a single major language model was at the centre of the professional discussion, but within a very short space of time, the topic has shifted to agents.

Gerhard Kürner
3 min readNov 16, 2024
Gerhard Kuerner CEO 506.ai

A year ago, a secure chat window was the ultimate in day-to-day business, but today it is no longer enough to lure a dog, er user, out from behind the stove. The rapid pace of development is once again painfully brought home to many companies when they have just introduced a chat-only window and perhaps even won an innovation award with it. At the same time, they then realise that employees don’t use it that intensively after all and certainly don’t achieve any measurable employee satisfaction or increase in productivity. Questions and answers in the chat process are a thing of the past.

Even today, and even more so in the future, only agents will support people in completing tasks or do them completely autonomously instead of humans. Even if agent is a loaded word in our language, there is currently no better word. With these agents, the question of the next best AI model will no longer arise, as we unfortunately still see in the media every day. It will be a thing of the past because a good agent system or interconnected processes, so-called ‘agentic-chains’, will automatically select and use the most suitable model or models.

But what do we actually mean by agents today? Agents are specialised programs or systems that are able to carry out tasks independently in order to achieve certain goals. They differ from conventional large language models (LLMs) in that they not only respond to requests, but can also act proactively, make decisions and carry out actions independently.

As with ‘AI’, we are still at the beginning here too, but 2025 will be the year of the introduction of agents in everyday corporate life. In terms of content, this next evolutionary step into the AI era is relatively easy to explain: AI systems must be able to access company data, they must be given a detailed description of what is expected of them and they need defined task templates.

This description of how agents work is important because it will change companies much more than we can imagine today. Not where employees no longer have a job, no, it will slowly but profoundly change the organisation and design of company processes. Imagine if a third of all administrative tasks could be automated. Who will be responsible for accompanying this development and adapting it sustainably? This is not an IT activity that affects the infrastructure, but a function that decides which processes are carried out by the company agents and which are not — this is an activity that we have not yet integrated into the company in this way.

And no, agents are not a spectre of the future that we can wait for. And no, introducing a software tool for all employees and holding a few workshops on it is something that everyone should have learnt in the last 12 months — that’s definitely not enough.

It’s time to rethink and take new steps towards this future.
Think of ‘M’, the head of MI6, who trains agents, like James Bond. Yes, the comparison falls flat because it’s not about hostile activity, so take the peaceful version of MI6, but start now to build expertise and understanding, even if the impact doesn’t turn your entire business model upside down today, but think of all the ways you can combat staff shortages and ever-growing bureaucracy on a day-to-day basis and stop wasting your employees’ valuable time.

--

--

Gerhard Kürner
Gerhard Kürner

Written by Gerhard Kürner

CEO 506.ai - European AI platform for integration and automation - Vice Chairman AI Upperaustria

No responses yet