You Can’t Bolt AI Onto a 1990s Org Chart

Play Audio Version

I recently talked to the founder of a new startup. His focus is advising companies on how to achieve success using artificial intelligence (AI). It is no surprise to see that there are more and more AI consultants out there, but I thought his approach was interesting.

He insisted that AI is not just a tool. It is a fundamentally different way of a designing a business. His argument was that if you are considering how AI can help your business then you should start out with the idea of transformation at the beginning.

The concept of agentic AI makes it clear. The way that many companies try using AI today is as a productivity tool. The aim is to use the tool to make things faster or better or cheaper. The company is not doing anything differently other than changing the tools that are being deployed. The tools get better and things get faster.

But think about how most companies are structured. There is an operations team, a marketing team, a sales team, HR, IT, and so on. Often each department has their own information system and data – it is usually a challenge getting them to share information.

Now, imagine having one single enterprise-wide source of company information so your virtual assistants can run around handling different tasks. It’s an inflection point. A completely different approach that will trigger the future of how companies are structured. Using AI effectively demands that we think about how companies are structured – it’s not just a bolt-on tool.

MIT published research indicating that 95% of generative AI projects are not creating any measurable return on investment. Companies are really struggling to find any value from AI. This research has been endlessly shared – usually by those who doubt AI, with this used as proof that it doesn’t work.

But the real story is that many companies have tried AI pilots and small projects. They work well. The failure occurs when they try scaling up to the enterprise level. The reason is that the underlying data that needs to be shared and accessible across all areas of the business is not easy to bring together.

I think that this really is one of the most important challenges to the use of AI. It will be both an opportunity and challenge for those executives who really want to utilize AI in 2026 – and beyond.

Let’s explore some of the biggest bottlenecks that are causing trouble:

• Infrastructure and security. Here we have issues such as data protection, compute power, and compliance to industry regulations. It is relatively simple to just increase the amount of speed or power available, but what happens when you also need to respect confidential customer data or control the provenance of the information being used?
• AI-native systems will demand change. I don’t just mean they need new technology. As indicated, this involves an organizational and cultural change. People need to revise how their job works. Is their job even required? How will reporting lines change? How will information hoarding be eliminated and no longer used as a source of corporate power? Entire processes and departments need to change so they can be planned around a team of agentic employees – working with humans. What would your business look like if it were designed from scratch today?
• Investment needs to increase. You need to plan seriously for hardware, infrastructure, cloud, and energy costs. This may require trade-offs or a squeeze on other areas of the budget.
• International services get more complex. If you are providing a service across multiple borders then you need to think even more carefully about regulatory compliance and how data sovereignty issues can be managed. Regional cloud sovereignty and confidential computing are becoming important in managing these processes.

The clear message is that the most transformational benefits with AI can only be achieved by making data available across the entire enterprise, but this means that you need to rethink how data is controlled. Who can access it and why? What happens once you add agentic employees to the existing human ones?

For the past couple of years it feels like everyone has talked about AI as if it were just a spreadsheet or some other software tool. It’s just something that works right out of the box.

This probably explains the project failures. To achieve the potential transformation you need to think carefully about who has  the required data, where it is located, and when it can be used. It requires a completely different approach to the traditional hierarchical corporate structure.

Designing a future-ready enterprise needs planning and effort. There are other technology trends on the horizon – look at how quantum computing is emerging now – but until we start restructuring how businesses are organized we will never be able to make the most of AI and agentic automation, let alone emerging trends and developments.

It’s time to get serious about AI and to finally realize that companies designed before generative and agentic AI almost certainly need their hierarchy and data flows redesigned to be more effective today. This is the real story for AI in 2026 – it really does work, but it needs thought and planning as you need to redesign the way that data is accessed and used across the entire enterprise.

Click here for more information on data management, analytics, and AI services from IBA Group.