Where to Start with AI: A Practical First Project for an SME

Few topics generate as much pressure and as little clarity for a business owner as artificial intelligence. The headlines insist that AI will transform every company, the competitors claim to be using it already, and the founder is left with a genuine question that no headline answers: where, specifically, do I start. The honest answer is that the first AI project matters far less for the value it creates directly than for what it teaches the organisation. The goal of project one is not transformation. It is a clean, visible win that builds confidence and capability for project two.

This reframing matters because the most common way SMEs waste money on AI is by starting too big. Inspired by ambitious case studies, they attempt a sweeping initiative, discover that their data is messy and their processes undefined, and conclude that AI is not for them. The discipline that avoids this is the same one that governs any good investment: start where value is high and difficulty is low, prove it, and expand from a position of strength rather than hope.

Choose with a value and feasibility lens

The simplest tool for selecting a first project is a two by two grid that scores candidate use cases on two axes: the value they would create and the feasibility of delivering them. Value asks how much time, cost, or risk the use case would remove. Feasibility asks how ready the data, the process, and the team are to support it. The first project should sit firmly in the high value, high feasibility quadrant. The transformational ideas that are high value but low feasibility are not wrong, they are simply not first. They become realistic once the organisation has built the muscle that an easier project develops.

Where the easy wins usually hide

For most growing businesses, the high value, high feasibility quadrant contains the same handful of opportunities, because they share three traits: the work is text based, it is repetitive, and it does not require perfect accuracy on the first pass. Drafting routine documents, proposals, job descriptions, standard replies, is one. Summarising long inputs, meeting transcripts, research, customer feedback, into usable briefs is another. Answering common internal questions from your own documented policies is a third. Each of these uses widely available assistants such as ChatGPT, Claude, or Microsoft Copilot, requires no custom development, and delivers value within days rather than quarters.

A worked example: a professional services firm chose a deliberately modest first project: using an AI assistant to turn consultants' rough notes into first draft client reports. The drafts still needed editing, but the blank page disappeared. The firm estimated each report saved roughly two hours, and across a team producing dozens of reports a month, that recovered well over a hundred hours of senior time in the first quarter, from a project that took a week to set up.

Define success before you begin

Even a small AI project deserves a clear measure of success, set in advance. Without one, the initiative drifts into a vague sense that the team is experimenting, which is impossible to evaluate and easy to abandon. The measure should be concrete and modest: hours saved per week, turnaround time reduced, error rate lowered. A baseline taken before the project, even a rough one, turns a subjective impression into evidence, and that evidence is what unlocks the budget and the confidence for the next, more ambitious step. The McKinsey research on generative AI suggests the technology could add trillions of dollars of value across the economy, but that value accrues to organisations that capture it deliberately, one measured use case at a time, not to those that adopt it as a gesture.

Protect the basics from day one

A first project should be small, but it should not be careless. Two guardrails belong in place from the start. The first is data: do not put confidential client information or personal data into a public AI tool without understanding where that data goes, and prefer the business tiers of these tools, which offer clearer protections. The second is the human check: AI drafts, humans decide. For anything that leaves the building or informs a real decision, a person reviews the output before it is used. These two habits, established on a low risk first project, become the foundation of responsible AI use as the organisation expands its ambitions.

The businesses that succeed with AI are rarely the ones that started with the boldest vision. They are the ones that started with a clear, modest, measurable project, proved it worked, learned from it, and let each success fund the next. Start small, start where the value is obvious, measure honestly, and expand from there. The destination can be ambitious. The first step should not be.

You probably do not need to build anything

One misconception keeps growing businesses from starting at all: the belief that using AI means commissioning a custom system, hiring specialists, or undertaking a costly technical build. For a first project, and often for the tenth, this is simply untrue. The overwhelming majority of early value comes from off the shelf tools that already exist and require nothing more than a subscription and some thought about how to apply them. The general assistants, the AI features now embedded in the software you already pay for, and the lightweight automation platforms cover an enormous range of useful work without a line of custom code. The build versus buy question, which consumes so much energy in larger technology decisions, has an easy answer at this stage: buy, configure, and learn. Custom development becomes worth considering only once you have proven, through simple tools, exactly where the value lies and why the off the shelf option falls short. Starting with what already exists keeps the cost low, the risk small, and the time to value measured in days, which is precisely what a first project needs.

Unsure where AI would actually pay off in your business? We will help you find a first project worth doing and a clear way to measure it.

Book a free 30 minute call

Innovate. Optimize. Grow.