Why you can’t AI your way out of fragmented IT

Many enterprises will have hundreds of projects running simultaneously, with the same people being pulled by competing priorities. There is only limited visibility into what is actually happening across the landscape, the connections and dependencies. When something goes wrong, it can feel like a Reservoir Dogs moment: everyone pointing the finger at everyone else.

In my experience, organisations that have grown through acquisition will recognise this problem immediately. The logic of consolidation is sound: combine functions, serve more customers, create economies of scale and take costs out of the business. That assumes, however, that the businesses being acquired run in broadly comparable ways when, often, they don’t.

There are often different tools and versions, standards, processes, governance models, and even diverse ideas of what “good” looks like. Most teams haven’t had the time to do the unglamorous work of understanding and standardising how teams operate. That work matters. However, it rarely gets the same attention as a new transformation programme or a major platform investment. If it’s overlooked, the debt compounds with every acquisition and every new initiative, eroding the foundations upon which IT operates.

The advent of SaaS models promised effortless upgrades, continuous innovation and a move away from capital-intensive data centre costs. However, without strong governance, it often layers in more complexity and makes spending less visible.

The result is fragmentation across the enterprise. Tools do not connect properly. Projects cannot be prioritised with confidence. Dashboards tell different stories depending on which system you’re looking at.

These dynamics create a real leadership problem for IT. Senior teams are often making the best decisions they can based on an incomplete view. They can see spend and activity, but they cannot always see the relationship between investment, risk and value. Often, the real picture only becomes visible in the rear-view mirror, by which time the cost of fixing it has usually multiplied.

Security adds another layer of difficulty. Large enterprises can look like battleships: heavily governed, well-defended and built to withstand sophisticated threats. But they can be exposed by faster, less visible risks – the tools that teams bought without telling anyone, the integrations that were never properly governed, the cloud services that expanded beyond anyone's visibility.

Every new tool and every new integration likely adds value, but it may also create another point of exposure. That’s why this problem cannot be treated as just an IT hygiene issue. It’s a business resilience issue.

The cost picture is already complex. Many organisations are paying for tools that overlap in some places and leave gaps in others. They may be running integration work to connect systems that shouldn’t have been disconnected, or even using manual efforts to bridge gaps between processes.

Shadow procurement compounds the issue. Finance, HR, marketing and business teams often buy SaaS directly because waiting for central IT ‘takes too long’. In the short term, that can solve a problem. In the long term, it creates governance blind spots, technical debt and security exposure. But the deeper issues are visibility and traceability.

If you cannot see how work flows through your organisation, you cannot diagnose where value is being lost. You cannot tell which projects deserve your best people, which ones should be stopped, where duplication exists, or where the business is leaving money on the table. A connected enterprise can trace work from idea to delivery to outcome. It can understand where decisions were made, where delays occurred, where costs increased and where value was delivered. Without that traceability, organisations spend too much time firefighting symptoms rather than fixing causes.

The temptation is to believe that AI will strip away all these challenges. The promise is attractive – AI will automate fragmented processes, replace manual effort, reduce cost, accelerate delivery and make the whole system work faster.

Most large organisations today carry a people cost in IT that dwarfs their tool and application costs. AI is beginning to shift that ratio toward fewer people managing more machines. In principle, that's transformative. In practice, the arithmetic only works if token costs stay where they are – and they're currently subsidised. Even if they don’t rise, I think that misunderstands the problem. If the underlying tools are disconnected, the data is poorly structured, ownership is unclear, and governance is inconsistent, AI does not fix the mess. At best, you just get a faster mess – harder to trace, govern and unpick.

This is particularly acute in software development. AI can accelerate coding, testing, documentation and analysis, but it can also weaken the thread of intent. Why was a decision made? What problem were we solving? What constraints were in place? Who owns the outcome? Technical debt used to accumulate slowly enough that teams could usually follow the trail. AI can make it accumulate at machine speed.

The organisations that will navigate this well start by being honest about where they are. That means assessing the technology estate as it really exists. What tools do they have? What do those tools connect to? What is duplicated? What is being used properly? What is costing more than it returns? Where are teams working around the process because it doesn’t work?

The solution is to build a practical view of the current estate and use it to decide what to do next. From there, the question becomes more straightforward: what needs to change, in what order, and where will the value come from?

It is rare to have the luxury of solving everything at once, so sequencing matters. The organisations that make progress pick a clear starting point, fix the foundations, create visibility and build from there. That might mean rationalising a toolset, standardising delivery methods, connecting portfolio planning to delivery execution, improving traceability, tightening governance around SaaS, or creating clearer patterns for how teams adopt cloud and AI. The point is to move with purpose – do the work that removes friction, improves control and creates measurable value.

Organisations don’t need another vendor. What they need is someone who sits between the tools and can look across the technology estate and into the gaps, understanding the commercial and operational reality. They help the organisation make better decisions about what to keep, what to connect, what to rationalise and what to stop. That means doing the hard work of discovery, assessment, sequencing and execution.

It is about getting more value from what they already have, while being clear about where change is genuinely needed. In my experience, this is where real value gets unlocked. Not through another big promise, but through a practical plan that connects technology decisions to business outcomes.

AI has a role to play. As do SaaS platforms, cloud services, automation tools and modern delivery methods. But they contribute sustained value only when they sit on sound foundations. That’s how you move from complexity to control.