AI isn't magic

It's just another IT Service that must earn its keep

Is your leadership team suddenly obsessed with “AI strategy”? Are vendors flooding your inbox with promises of AI-powered everything? Has your budget request for basic infrastructure maintenance been sidelined while innovation funds flow freely to anything with “AI” in the title?

Welcome to the latest technology hype cycle.

As someone who’s witnessed the breathless excitement around blockchain, big data, cloud, and countless other “revolutionary” technologies (I can even remember when COTS hardware was going to change everything), I must admit to some scepticism when it comes to AI in the enterprise. Not because it lacks value – it has the potential to deliver significant benefits if properly applied - but because the gap between hype and practical implementation in large organizations has consistently been measured in years of frustration and millions in sunk costs.

So how about we do something slightly different: What if we stripped away the mystique and treated AI as just another IT service?

Because that’s exactly what it is. AI, at its core, is simply another capability that IT departments need to deliver, manage, and support, no different in principle from virtual machines, network infrastructure, or business applications. It requires hardware, software, security controls, integration points, and skilled personnel to function properly.

The key difference? How we choose to deliver it.

For organisations, there are two basic models for delivering IT services:

• The “High Touch“ model, which relies heavily on human intervention, manual processes, and approval gates. It’s the default - but it’s painfully slow.

• The “Self-Serve“ model, which leverages self-serve experience, automation, and built in governance. It’s efficient and fast.

If you’ve read any of my previous articles, you will know I sit firmly in the “Self-Serve” camp. Yes, there is an upfront cost as you scale the learning curve, but the pay-off at the backend in terms of efficiency gains, cost reductions and quality improvements more than makes up the difference.

In this article, I will show why organizations that choose the Self-Serve path for AI won’t just be better at using AI technology, they will gain a competitive advantage in how quickly they can turn any new technology into business value.

The High Touch Reality: Hiding Technology behind Experts

Let’s be honest about how most IT organizations deliver new services today. For these organizations there are two distinct phases: First, the commissioning phase, where the new IT service “put into production” and second, the subscription phase, when users can access and use the service.

In High Touch organisations, new IT services are commissioned using the project delivery model and can be a massive undertaking – sometimes taking years! It starts with a business unit identifying a need or a new technology. An IT business case is drafted. The business case routes through multiple rounds of approval. It gets adjusted. Resubmitted. Put on hold until the next funding cycle. Eventually it is approved with reduced scope.

Finally, the project is created. Resources are assigned. Requirements are gathered. Architecture reviews happen. Security assessments are conducted. Testing is performed. Documentation is written. Training is delivered. The Operational Readiness Checklist is signed off. Then the new service is “thrown over the fence” to a brand-new operations team. And so, after many months (years!) the service is available.

Now the subscription phase:

An eager user wants access to this new service. They submit a service request. Then another form for network access. Then another for identity permissions. Then yet another for data access. Tickets bounce between teams. Status updates are requested. Clarifying questions are asked. Engineers work through multiple SOPs. Configurations are tested. More permissions are requested. More waiting ensues. Eventually‚ weeks later, the user finally gains access to the service they needed yesterday.

Two separate phrases, one consistent outcome: delay.

Sound familiar?

Despite its obvious drawbacks, this “High Touch” delivery model has become so entrenched in large organizations that we hardly question it anymore. It’s how we’ve always done things. And it’s probably exactly how your organization is planning to deliver AI capabilities.

Think about it: How many of these statements sound like discussions happening in your organization right now?

• “We need an AI governance committee to review all potential use cases.”

• “Our enterprise architecture team is developing an AI reference architecture.”

• “We’re establishing an AI Centre of Excellence.”

• “We need specialized AI infrastructure with dedicated support teams.”

• “All AI models will require security review and executive sign-off before deployment.”

Each statement sounds reasonable. Each will also ensure that the actual value of AI remains theoretical rather than practical for most of your organization.

The tragedy isn’t that these concerns exist‚ many must be addressed for risk management. The tragedy is that we automatically default to solving these risks by wrapping new technologies in layers of process and control instead of rethinking how the service could be delivered. By delivering AI capability using the High Touch model, you’re ensuring that whatever value AI might offer will be delayed, diluted, and ultimately disappointing.

Unfortunately, AI Magnifies the High Touch Problem

The High Touch delivery model creates friction for any potential IT service, but with AI, these challenges become exponentially worse. Here’s why:

1. The data problem: Unlike most other IT services, AI thrives on data, lots of it. In a high touch world, each data source typically has its own owner, its own governance process, and its own approval chain. Want to combine customer data with operational data for a predictive model? Prepare for a byzantine journey through multiple approval committees, privacy reviews, and data access agreements. By the time you get the data you need, the business opportunity has likely vanished.

2. The governance paradox: Yes, AI needs governance to manage risks around bias, privacy, and security. But excessive governance in a high touch model creates its own risk: Shadow AI. When employees can’t get timely access to sanctioned AI tools, they simply use public options like ChatGPT without oversight. Ironically, your control-focused approach creates exactly the uncontrolled environment you were trying to prevent.

3. The feedback loop problem: AI systems improve through iteration based on user feedback and performance data. Traditional high touch deployment cycles might be quarterly at best, far too slow for effective AI learning cycles. This turns what should be a key advantage of AI, its ability to continuously improve, into a non-factor.

The combined effect? Your organization’s AI initiative will likely deliver too little, too late, at too high a cost. You’ll watch competitors race ahead while your teams drown in process. And the transformative potential of AI will remain largely theoretical rather than practical for your business.

There must be a better way.

The Self-Serve Alternative: AI at the Speed of Business

Imagine a different approach. Rather than use two separate phases, which can take months, sometimes years, before anything of value is created, use an incremental approach – a product delivery approach. With this approach, your teams continually provide small value increments without the huge upfront costs nor the pain that comes with throwing solutions “over the fence” to operations.

Here’s how it works. First, start with a business problem. Suppose it takes weeks for a business analyst to handle data related to customer retention. The data is spread across multiple team silos, a dozen formats, separate IT systems. Our business analyst must interact with multiple teams to get the data required before compiling it, organising it and then using it to understand how to improve customer retention. In today’s world, this sort of month’s long cycle is just not quick enough. So, the business decides to fix it.

However, rather than build a massive business case followed by a years-long project to commission a new solution, start small. Focus on cross-team collaboration, using an Agile and/or DevOps framework and employ self-serve, on-demand cloud, data & AI technology to build version one of your solution. Since the solution itself is built in small increments, each increment provides a little more value to our business analyst, by, for example, supporting a new data source and consequently reducing turnaround time.

By version 25, our business analyst might log into a self-service portal, select an AI-powered predictive churn model template. The analyst then connects it to pre-approved customer data sources through a drag-and-drop interface. Built-in guardrails ensure privacy compliance. The system automatically provisions the necessary compute resources. Our analyst can now run initial predictions in minutes. No IT tickets. No phone calls. No waiting for specialists.

Perhaps that sounds like fantasy? You wouldn’t be the first to say it! But it’s not. Organizations at the leading edge of IT transformation already deliver capabilities this way‚ by providing robust self-serve products that put approved technology directly in the hands of business users.

AI is no different.

The results can be transformative:

• Where time to value was measured in months, now its minutes

• Where business viewed IT as a cost centre, it’s now a profitable partner

• Where once Shadow AI was attractive, now IT products are more useful

• Where compliance was once manual approval gates, now its continuous monitoring

• Where AI specialists would respond to routine calls, now they can focus on innovation
  

A Quick Reality Check: This Isn’t Easy

Let’s pause for a moment. If you’re thinking this self-serve vision sounds too good to be true, you’re partly right. The transition from a High Touch to a Self-Serve operating model isn’t a weekend project or even a quarterly initiative. It’s a fundamental transformation that will eventually touch every aspect of how your IT organization operates.

The truth is most organizations will fail to make this shift. They’ll keep adding AI capabilities (and other technology) through their existing delivery models, wondering why the promised benefits remain elusive. They’ll form committees to study the problem. They’ll hire consultants who will tell them what they already know. They’ll launch pilots that never scale. And in five years, they’ll still be wrestling with the same challenges, just with newer technology.

Why? Because the shift to self-serve requires overcoming entrenched interests, rethinking established processes, and challenging cultural norms that have been reinforced for decades. It requires:

• Leaders to invest in products rather than projects

• IT specialists to become enablers rather than gatekeepers

• Organizations to value speed and iteration over perfect-but-delayed solutions

• Security teams to trust automated guardrails instead of manual reviews

Let me be clear: I’m not suggesting you can transform your entire IT organization overnight. What I am suggesting is that AI presents a perfect opportunity to start this transformation‚ to create a pocket of excellence that demonstrates what’s possible when you deliver technology differently.

From Months to Minutes: Making AI Self-Service a Reality

The fundamental shift required isn’t just in how you deliver technology‚ it begins with how you think about delivery itself. This means moving from a project delivery model to a product delivery model. Instead of creating one-off AI capability that sits behind a wall of process and control, you’re building products, including AI products, that users can independently access and apply to their business problems.

To explain how to make this shift would require an entire book, in fact, books, but for the purposes of this article, here are the high-level steps you’ll need to consider.

Step 1: Agree to Principles

Before writing a single line of code or producing your first mock-up, align your stakeholders around core principles:

• Product delivery (this model is well understood, don’t reinvent the wheel here)

• Start small, but have a North Star

• Focus on providing incremental, measurable value

• User experience as a primary success metric

These principles will guide every decision and help resolve the inevitable conflicts that arise.

Step 2: Identify your first problem

Choose a specific problem that is impacting your business. Look for use cases that:

• Deliver visible business value quickly

• Have clear success metrics

• Balance innovation with governance requirements

Don’t, and I repeat don’t, start with an AI solution. Start with a business problem. Every business has them. It won’t be hard to find.

Step 3: Assemble a product team

This isn’t a standard project team. You need cross-functional expertise including:

• Service designers who understand end-to-end user journeys

• UX developers who can create intuitive interfaces

• Technical writers who can make complex capabilities accessible

• Engineers with automation expertise

• Product managers who understand both business and technical constraints

Most organizations underinvest in service design, UX, and documentation‚ precisely the elements that determine whether self-service succeeds or fails.

Step 4: Know your users

With your team in place, build deep connections with those who will be using your product:

• Conduct user research to identify their actual needs (not just what they request)

• Map their current workflows and pain points

• Understand their technical comfort level and domain expertise

• Involve them early and continuously throughout development

Your users will be your biggest advocates—if you design for their actual needs. Their enthusiasm and success stories will drive adoption far more effectively than any top-down mandate. Conversely, if you build something that doesn’t align with how they work, no amount of training or promotion will save it.

Step 5: Adopt a modern delivery model

Implement Agile and DevOps practices that enable:

• Rapid iteration based on user feedback

• Continuous deployment of improvements

• Automated testing and security validation

• Close collaboration between development and operations

This isn’t just about methodology‚ it’s about creating the framework for moving at the speed self-service requires.

Step 6: Secure incremental funding

Break away from the annual budgeting cycle and towards incremental funding, by:

• Defining clear success metrics tied to business outcomes

• Demonstrating value early and often

• Connect funding to achieving the business outcomes

By tying funding to business outcomes, you’ll never be in the situation where you’re asking yourself, “why are we doing this again?”.

Step 7: Create standardized user journeys

Charge your service designers with the goal to map and optimize key user journeys for your product:

• General paths such as: Subscribe, Access, Use, Unsubscribe

• AI-specific paths, including: Data connection, Model selection, Parameter configuration

• Exception paths: Support requests, Feature suggestions, Compliance verification

These journeys become the blueprint for your user experience design and automation efforts.

Step 8: Invest in exceptional user experience

Collaborate with your UX developers and technical writers to create:

• Intuitive interfaces that guide users through complex processes

• Clear, accessible documentation integrated into the experience

• Interactive examples that demonstrate capabilities

• Progressive disclosure that reduces complexity for new users

Remember: If your self-service experience isn’t dramatically better than calling IT support or turning to Shadow IT, users will always choose one of the latter options.

Step 9: Automate, automate, automate

Build automation into every aspect of the experience:

• Infrastructure provisioning and scaling

• Security and compliance verification

• User onboarding and authorization

• Monitoring and alerting

• Documentation generation and updating

Critically, embed controls into your CI/CD pipelines so that governance happens automatically with each deployment, not as a separate process.

Step 10: Close the feedback loop

Implement mechanisms to:

• Capture usage patterns and performance metrics

• Collect explicit and implicit user feedback

• Analyse patterns to identify improvement opportunities

• Prioritize enhancements based on actual usage, not assumptions

This creates a virtuous cycle of continuous improvement driven by real-world usage.

Step 11: Celebrate and market successes

Transforming delivery models isn’t just a technical challenge‚ it’s a change management challenge. Systematically:

• Measure and document time-to-value improvements

• Share compelling success stories across the organization

• Promote the users of your product

• Create communities of practice to share learning

Your most powerful weapon against organizational inertia is demonstrable success.

You may look at this roadmap and say to yourself, look at that - there doesn’t appear to be anything too specific to AI here. And you’d right, this roadmap is not about AI, it’s about how you could deliver IT services to your business. You can start with AI, and then, building on your success, shift to other IT services. Yes, even networking.

Conclusion: The Choice Is Yours

Let’s circle back to where we started. AI, despite all the hype and mystique, is just another IT service. The question isn’t about AI’s potential to solve business problems‚ the question is whether your organization can realize that potential.

You have two options:

You can treat AI like every other technology that’s come before it. You can wrap it in layers of process, gate it behind committees, and restrict it to specialists. You can ensure it takes months to access and deploy. You can watch it become another promise unfulfilled, another technology that delivered a fraction of its potential.

Or you can use AI as the catalyst for transformation. You can build a pocket of excellence that demonstrates what’s possible when you shift from high touch to self-serve delivery. You can create a product that puts AI capabilities directly in the hands of the people. You can show what’s possible when you move from months to minutes.

The technology isn’t the limiting factor. The delivery model is.

For those ready to make this shift, the path won’t be easy. You’ll face resistance from those invested in the status quo. You’ll need to secure funding for a different kind of initiative. You’ll have to build new skills and change longstanding behaviours.

But the alternative is continuing to deliver tomorrow’s technology using yesterday’s methods‚ and expecting different results.

The leaders who succeed in the coming years won’t be those who have the most advanced AI capabilities on paper. They’ll be those who put those capabilities in the hands of their people most effectively.

They’ll be the ones who transform AI from a buzzword to a business tool by changing not just what they deliver, but how they deliver it.

Which leader will you be?