Why AI moonshots miss, and small steps hit - Version 1

Think big, start small

We’re experiencing an extraordinary moment in technology. Each day seems to bring a fresh breakthrough, sparking imagination and inspiring leaders everywhere to reconsider what’s possible. It’s easy to get swept up in the excitement. AI feels like a thrilling new frontier, promising bold, industry-defining leaps forward, and who wouldn’t want to be part of something groundbreaking?

Yet, as the initial hype settles, reality is proving far less straightforward. Despite billions invested in ambitious AI initiatives and a constant stream of impressive demos, most businesses haven’t experienced the dramatic transformations initially promised. Grand moonshot projects often stall, not because AI lacks potential, but because sweeping transformations inevitably bring sweeping complexity. Many companies get caught up in the vision, overestimating AI’s immediate capabilities while overlooking the practical, incremental applications that could drive real impact today.

So, how can businesses effectively harness AI’s potential without becoming overwhelmed by elusive moonshots? The key is to think big – but start small.

The temptation of the big vision

Generative AI has undeniably fuelled the appetite for big ideas. It’s now easy to imagine a virtual assistant that does it all: handling customer inquiries, drafting strategy documents, maybe even serving as an ever-present AI executive. After all, if ChatGPT can whip up an email in seconds, why not an AI-powered CEO?

This is where expectations start to run ahead of reality. The idea of an all-knowing, do-everything AI is alluring, but it doesn’t match where the technology is today. Building an AI that can reliably and intelligently handle every business challenge is incredibly complex. Too often, projects that try to do too much become unwieldy, take longer than expected, and ultimately fall short.

Instead of aiming for an AI that can chat about anything and everything, companies see far better results when they focus on specific, high-impact use cases. For example, an AI trained to assist employees with HR policy questions, pulling answers only from official company documents, can be both effective and easy to maintain. Or trained to assess large volumes of documentation against a set of business rules, saving valuable manual processing hours. Similarly, a customer service bot that handles simple password resets or frequently asked questions can provide real value without over-promising. The key is to narrow the focus so that AI remains useful, reliable, and easy to manage.

The power of small, iterative wins

The companies making real progress with AI aren’t betting on a single moonshot – they’re taking a steady, incremental approach. They identify specific problems and use AI to solve them, one step at a time. These small wins don’t make for dramatic headlines, but they create tangible value and build a strong foundation for future AI expansion.

AI pioneer Andrew Ng has long advocated for this approach: start with small projects that deliver value within six to twelve months, use those successes to build internal AI expertise, and gradually take on bigger challenges. A company’s first AI initiative doesn’t need to be groundbreaking – it might be as simple as automating a routine data-entry task or generating weekly sales reports.

These small wins build momentum, establish trust in AI, and help teams understand what the technology can and can’t do. Over time, as confidence grows, the business can take on more ambitious AI projects with a much higher likelihood of success.

group of professionals working in a bright office around computers

Case in point: Delphi, a small AI with a big impact

Take Delphi, our own internal bid-support assistant. At first glance, it’s hardly a moonshot. We weren’t trying to build an AI that could write entire client proposals from scratch or automatically win new business. Instead, Delphi was designed to tackle a specific, common frustration: finding and reusing past bid content.

Anyone who has worked on sales proposals knows the pain of digging through old documents to find that perfect paragraph. It’s tedious, time-consuming, and inefficient. So, we trained Delphi on our repository of past bids and responses, allowing it to quickly surface relevant examples. Instead of sifting through endless folders or sending out frantic emails – “Does anyone have a cybersecurity risk section from a past bid?” – our team can now ask Delphi.

The impact was immediate. What used to take hours now takes seconds. Bid managers save time, proposals maintain consistency, and we avoid reinventing the wheel for every new opportunity. Delphi isn’t flashy, but it’s invaluable. It solves one problem really well, and in doing so, it fundamentally improves a key business process.

From Small Wins to Lasting Change

But does a series of small AI projects really add up to transformation? Absolutely. In fact, a collection of well-executed, practical AI solutions often delivers far more impact than a single over-ambitious initiative. Think of it as compounding gains: each small AI success lays the groundwork for the next, gradually reshaping how the organisation operates.

Consider a company with dozens of AI-powered tools, each solving a different but meaningful problem. A chatbot handling routine customer inquiries. An AI system streamlining expense approvals. A model that helps predict inventory needs. Each one might only save employees a few hours per week, but together, they create a business that is significantly more efficient, data-driven, and competitive.

A smarter approach to AI adoption

So, how do you make AI work for your business? Keep these principles in mind:

ROI, not sci-fi

Prioritise AI projects that deliver measurable business value over those that simply sound impressive

Narrow the focus

AI works best with a clearly defined, genuine problem. A tool that solves one problem well is more valuable than one that tries to solve five and does none of them effectively

Iterate and scale

Start with small pilots, learn from them, and expand what works. This approach minimises risk while building organisational buy-in

Educate and empower

Use each AI initiative as an opportunity to increase AI literacy within your company. The more people understand AI, the better they can identify useful applications

Build a strong foundation

Good AI relies on good data. Investing in data quality, governance, and integration will pay dividends as you scale AI across the business

Human centric AI

Underpinning all these principles, successful AI adoption requires thoughtful governance and a relentless commitment to user-centred design, ensuring the solutions you create consistently address genuine human needs

This isn’t about playing it safe – it’s about being strategic. Grand AI visions aren’t inherently bad, but they need to be grounded in reality. Practical AI adoption isn’t about abandoning ambition; it’s about earning the right to pursue bigger opportunities through a track record of success.

The path to AI transformation

Ultimately, the best way to build an AI-powered organisation isn’t through a single dramatic leap – it’s through steady, thoughtful progress. Avoid the siren song of the AI moonshot that promises everything but delivers little. Instead, break your AI journey into focused, achievable projects that create real value now. Over time, those small wins will add up to something truly transformative.

In the end, the companies that truly thrive with AI won’t be chasing flashy breakthroughs – they’ll be steadily advancing, step by step, towards lasting innovation.

For more detail on how our AI expertise is driving innovation and helping businesses transform, click here.