Why AI is overwhelming IT departments: tool sprawl, not talent

The real problem is fragmentation across platforms, models, and vendors, making it harder to standardise, govern, and scale value.

Article link: https://open.substack.com/pub/johanosteyn/p/why-ai-is-overwhelming-it-departments?r=73gqa&utm_campaign=post&utm_medium=web&showWelcomeOnShare=true

Audio summary: https://youtu.be/M9MJZ9EdSog

Follow me on LinkedIn: https://www.linkedin.com/in/johanosteyn/

Most organisations think their AI challenge is skills. They assume the reason IT is struggling is that teams don’t understand models, prompting, or the latest vendor stack. In reality, I’m seeing something more mundane and more dangerous: sprawl. AI rollouts often start with good intentions and a handful of pilots. Then every department gets excited, buys or enables its own tools, and suddenly IT inherits a fragmented estate: multiple platforms, duplicated capabilities, inconsistent governance, and a growing queue of integration and security problems.

This is not a theoretical risk. A recent TechCentral article, AI complexity is crippling IT departments, describes how AI is adding complexity and technical debt rather than simplifying IT environments.

CONTEXT AND BACKGROUND

AI is being embedded into everything: email, CRM, analytics tools, meeting platforms, design software, and development environments. That sounds like progress, but it also means AI is no longer a single programme with a single owner. It becomes dozens of micro-decisions made across business units, each one rational on its own, but chaotic in aggregate.

In practice, this looks like “pilot inflation”. A team trials one chatbot. Another buys a summarisation add-on. Someone builds an internal agent using a low-code tool. A vendor bundles AI into an existing contract. Within months, you have an AI estate, whether you planned for it or not.

The scale of the problem is visible in market surveys. A December 2025 enterprise survey by Zapier reported that only 35% of leaders say all AI tools used in their business go through proper approval channels, and it highlights tool sprawl as a major blocker to integration.

INSIGHT AND ANALYSIS

Tool sprawl creates three immediate problems for IT:

First, integration becomes the tax nobody budgeted for. AI systems are not useful in isolation. They need access to data, documents, workflows, and identity controls. When every department picks different tools, IT ends up stitching together connectors, permission models, audit logs, and policies. The work is invisible until it breaks, and then it becomes urgent.

Second, governance becomes inconsistent. Different teams apply different rules around data privacy, retention, prompt logging, and acceptable use. This is where risk creeps in quietly: sensitive information in the wrong place, outputs that can’t be audited, and decisions being made with no clear accountability.

Microsoft has been responding to this reality by pushing more centralised governance features for Copilot and agents. In December 2025, Microsoft outlined new security and governance innovations, emphasising the need for a unified view across controls and admin tooling rather than scattered oversight.

Third, costs start drifting upward in ways procurement does not easily see. AI-related spend is often distributed across departments, bundled into subscriptions, or hidden inside cloud consumption. Then IT gets blamed for “AI costs” while lacking the authority to rationalise the toolset.

This sprawl problem is not just operational; it is strategic. Gartner’s CIO survey (published November 2025) expects AI to touch all IT work by 2030, which means the organisations that don’t standardise now will simply scale their complexity later.

IMPLICATIONS

For CIOs and IT leaders, the practical move is to treat AI like an enterprise capability, not a collection of experiments. That does not mean killing innovation. It means setting guardrails early: approved platforms, identity and access standards, data handling rules, model evaluation practices, and clear ownership for “who supports what”.

For business leaders, this is also a leadership issue. If every division can buy tools freely, IT will become the clean-up crew. And the clean-up crew will always be behind. The right model is a partnership: business units bring problems and priorities; IT brings standards, architecture, and the ability to scale safely.

For security teams, the warning is blunt. Tool sprawl is already showing up as a measurable pain point, with unmanaged tools creating blind spots and inconsistent policy enforcement. A February 2026 analysis highlighted how sprawl contributes to weak asset visibility and accountability, which is exactly the environment where modern attacks thrive.

CLOSING TAKEAWAY

AI is not overwhelming IT departments because IT people can’t learn new technology. It is overwhelming them because organisations are allowing AI to grow like mould: everywhere, in different shapes, with no plan for containment. The fix is boring, but necessary: rationalise tools, standardise governance, build shared platforms, and measure value in a way that encourages fewer, better deployments. AI will absolutely reshape business, but it will only deliver sustainable value if we stop mistaking tool adoption for strategy and start designing an AI estate that can be operated, secured, and trusted.

Author Bio: Johan Steyn is a prominent AI thought leader, speaker, and author with a deep understanding of artificial intelligence’s impact on business and society. He is passionate about ethical AI development and its role in shaping a better future. Find out more about Johan’s work at https://www.aiforbusiness.net