Navigating the AI Landscape

AI Resources Issues & Experimentation Limitations

Let me guess — you’ve been hearing non-stop about AI. It's on every tech blog, in every boardroom conversation: "AI will save you time, reduce costs, unlock new markets, revolutionize your business!"

Naturally, your next question is: “Okay, how do I actually get started with AI?”

Having worked with AI for the past 8 years, let me be honest — the hype is real, but the journey is often underestimated. One of the first challenges we encountered was a scarcity of AI expertise. We had big ideas, especially in the service management space, but quickly hit a wall: not enough skilled people to experiment the way we wanted.

We tackled this by investing in talent — building internal capability, hiring data scientists, machine learning engineers, and neural network specialists across 3 different geos! More recently, we went on the hunt for two senior AI architects. That search took 6 to 9 months — and not for lack of applicants. Quite the opposite: we had to sift through what felt like a flood of AI-generated résumés and superficial profiles.

We built a rigorous hiring and interview process to cut through the noise and ensure we brought in the right people. 

Why does this matter?

Because experimentation is at the heart of successful AI adoption. You can’t unlock value from AI without testing. Start with clearly defined use cases — pain points where AI can genuinely make a difference. Then find or grow the right resources to work on them.

One final tip: if senior talent is scarce, invest in young, eager minds. More universities now offer AI-specific programs, and with proper mentorship, these grads can grow into your most valuable assets.

Data Security & Privacy Considerations

Once you’ve assembled the right team and begun experimenting, another critical challenge quickly emerges: data security and privacy.

With evolving regulations like the EU AI Act taking effect, businesses must go beyond building powerful models — they must ensure those models are also safe, ethical, and compliant. Great algorithms alone aren't enough; where your data comes from, how it's stored, and how it’s used has never been more important.

Here are key factors to consider:

• Model Hosting & Infrastructure: Deciding where and how your models are hosted impacts both performance and compliance. For instance, services like Microsoft Azure offer access to OpenAI models via API — but some tiers may not allow opting out of data sharing for model training. Lower cost options may compromise privacy. Would a fully local, air-gapped deployment give you better control and assurance?
• Training Data Protocols: Avoiding data leakage is essential. Large models, even open-source ones, can inadvertently memorize and regurgitate sensitive information. You need robust protocols to ensure that no private or personally identifiable data makes its way into the training process.
• Security Strategy & Governance: Is your model operating in your environment or a third-party cloud? What data governance policies are in place? Consider aligning with recognized standards like BSI (British Standards Institution) for best practices. And importantly, how do your customers and stakeholders perceive your handling of their data?

Understanding Cost Models

As the AI ecosystem matures, cost structures are rapidly evolving — and understanding them is just as important as evaluating performance or accuracy.

Traditionally, AI service providers charged on a token-based model, where usage was metered by the number of input and output tokens. While this model offered some predictability, it often led to cost surprises as usage scaled.

Now, we’re seeing a shift toward fully hosted, subscription-style services and tiered enterprise offerings — each with its own tradeoffs, here are a  few examples:

• Token-Based APIs: Flexible, pay-as-you-go, but can scale unpredictably with usage.
• Hosted Platforms: Simplified billing and infrastructure management, but may involve data lock-in, long-term commitments, or opaque pricing.
• Self-Hosted/Open Source Models: High upfront cost and operational complexity, but long-term control and potentially lower variable costs.

For finance, procurement, and engineering teams, it’s essential to model costs under realistic usage scenarios, consider future growth, and understand where the true cost drivers lie — whether compute, storage, API calls, or vendor markups.

The financial model you choose will shape your deployment strategy, your risk profile, and your ability to scale sustainably.

Choosing Impactful Use Cases

AI isn’t cheap — in money or time. That's why choosing the right use cases is critical.

The most successful AI initiatives don’t start with, “Let’s build something cool!” They start with, “What’s a real problem we have today that AI can solve better, faster, or smarter?”

Your goal should be impact over complexity. Pick use cases where value can be measured — whether it's internal efficiency gains or customer based cost savings, improved service resolution times, or better customer experience.

Some guiding principles:

• Start small but meaningful: Quick PoCs with defined success criteria.
• Fail fast and learn faster: Not every experiment will work, but each one should teach you something.
• Keep your eyes on ROI: AI should either save money, make money, or unlock something you couldn’t do before, like offering a new differentiating service that help you sell more or grow relationships. 

AI Journey Thoughts

The AI journey is exciting and full of potential — but it’s far from plug-and-play. Success requires a thoughtful investment in talent, training, security, infrastructure, and continuous experimentation.

If you’re just starting out, keep these principles in mind:

• Solve real problems first — don't build tech for its own sake.
• Invest in people — a strong talent pipeline is your most valuable AI asset.
• Experiment often and stay agile — small pilots can lead to big insights.
• Reimagine the customer journey – You have more tools than ever to completely revolutionise the way customers interact with your business

Above all, treat AI not as a one-time project, but as a core business capability. A clear, flexible, and responsible AI strategy isn’t optional anymore — it’s a competitive necessity.

With the right foundation and mindset, AI won’t just optimize what you already do — it can fundamentally reshape what's possible.