Organisations are coming round to the benefits of AI within their systems, and adoption rates are increasing, even amongst those who were hesitant at first. However, AI adoption is no longer the challenge. Losing control of it is.
With this in mind, we brought together 15 senior Technology leaders in Munich for an open and honest discussion around what the ‘AI Operating Model’ actually looks like in practice.
From the challenges of ‘Shadow AI’ to the importance of introducing work councils early on in the transformation, some truly transformative insights were uncovered. With so much to look at, we’ve condensed the three biggest topics below:
Your Employees Are Using AI, You Just Can’t See It
One of the biggest themes during the conversation was the rise of ‘Shadow AI’.
Integrating AI into our business processes takes time and consideration, but our employees don’t like waiting. This is resulting in a spike in ‘Shadow AI’.
With incoming laws such as the EU AI Act making AI adoption an increasing challenge for organisations, remaining in control of the implementation results in longer transition times. Engineers and employees get frustrated with this and choose to skip ahead of the formal rollouts.
Whilst the innovation of this happening is good for the organisation, it also poses some risks:
- AI tools are being used outside of the approved systems, risking data leaks.
- Sensitive data is being shared without clear control or oversight.
- Outputs are being relied on without validation or top-level oversight.
These risks would usually result in the organisation taking measures to immediately stop the use of AI by employees, but our attendees actually expressed how important it is not to do that. Instead, the discussion was to embrace the use of AI by creating safe spaces and sandbox environments for employees to test within and focus on guidelines rather than overall restrictions.
This way, employees get to continue innovating within controlled spaces, and the need to move onto unauthorised platforms is removed.
Start With The Problem, Not The Tool
As the conversation progressed, another pain point became very apparent. Too many organisations are starting with AI (the tool), instead of the problem they’re trying to solve with it in the first place.
It’s a common misalignment amongst organisations because of how quickly AI has come onto the corporate scene. This dramatic demand for AI has led to organisations implementing AI for the sake of AI, not for solving real problems.
AI is not a strategy. It’s an enabler.
Our attendees highlighted that without a clearly defined and well-communicated outcome, even the most advanced tool will create more confusion than value, especially when different teams are experimenting in different directions.
In terms of what this looks like in practice, those around the table shared that it all starts with defining what success looks like upfront (both from a commercial perspective and an operational one). From there, creating clear communication lines between all business functions and pushing the ‘why’ behind the implementation is key. If you can marry those three challenges together, you’ll be on the right path for success.
One attendee in particular had great success on this by taking a stakeholder from each department and using them to build relationships across the functions and staying connected. They emphasised that this didn’t mean taking the most senior member from each to head the project up. Instead, they wanted those with the best relationships across the organisation to make alignment as effortless as possible.
Speed vs Safety
Innovating at speed, but remaining in full control, has been a challenge for organisations before the acceleration of AI. However, the technology has shown how much quicker the implementation can now be, proving to stress safety demands further than ever before.
Our attendees coined this new gap the ‘two-speed organisation’. This is where innovation and governance are no longer moving in sync. Once you add in the increasing amount of regulation around AI, it’s easy to see why this gap is only getting larger.
With legislation getting tighter, our leaders made it clear that it is now absolutely critical to involve your work councils from the get-go during AI implementation. With their support and knowledge of laws, you can bridge the gap between innovation and regulation, creating fast and safe transformation.
As we continued the conversation, the room became concerned of one thing in particular: Are we at risk of losing fundamental understanding as we scale AI?
Teams delivering work they don’t fully understand, the slow and gradual loss of core problem-solving skills, and organisations becoming operationally shallow are all concerns of scaling AI the way we currently are.
To ensure we combat these hurdles effectively, our attendees highlighted the importance of building a culture that feels confident in challenging AI outputs, not just accepting them. This goes hand-in-hand with further education and awareness around tools before passing them on to individuals.
Conclusion
The room unanimously agreed that AI isn’t actually creating new challenges; it’s simply exposing the existing ones at speed, creating the sense of new challenges.
For example, the challenge of Shadow AI. In isolation, you would jump to the assumption that AI has created this issue and that it did not exist before. However, what shadow AI has actually highlighted is that the organisation is slow to adapt, does not have safety areas for employees to test ideas in, and does not have a culture where employees feel safe saying how they work.
These are all challenges that were already existing, cracks in the wall, if you will. The addition of AI is simply the microscope that allows you to zoom in close enough to see them.
So, before you jump to add AI into your business processes, make sure you have a clear business outcome tied to it, that you communicate this well across the organisation, you involve work councils early, and finally, you create a safe culture where employees are not afraid of judgment.
