Here’s a situation that’s become incredibly common: a business leader knows their company should be “doing something with AI,” but they don’t have a data scientist, machine learning engineer, or anyone on staff who’s built an AI system before. The budget doesn’t support hiring a full AI team. And the vendor pitches they’ve received range from $20,000 chatbot implementations to $500,000 “digital transformation” projects that sound impressive but vague.

So what do you actually do? How do you build an AI strategy when you don’t have an AI team?

I’ve watched a lot of businesses wrestle with this over the past two years. The ones that got it right followed a surprisingly similar pattern. Here’s what it looks like.

Step One: Forget the Technology

This is counterintuitive, but the first step in building an AI strategy has nothing to do with AI. It starts with understanding your business problems clearly enough to know which ones are worth solving with technology.

Sit down with the people who run your operations — the ones who deal with the actual work every day — and ask them three questions:

1. What tasks take the most time relative to their value?
2. Where do mistakes happen most often?
3. What information do you wish you had but don’t?

The answers to these questions will give you a shortlist of problems. Not all of them will be AI problems. Some will be process problems, training problems, or plain old “we need to buy better software” problems. That’s fine. Eliminating non-AI problems from the list is genuinely useful because it stops you from buying AI solutions for problems that don’t need them.

Step Two: Categorise What’s Left

The problems that remain after filtering usually fall into a few categories:

Pattern recognition tasks. Sorting emails, categorising support tickets, identifying anomalies in data, flagging quality issues — anything where a human currently looks at information and makes a classification decision based on patterns they’ve learned.

Prediction tasks. Forecasting demand, estimating project completion dates, predicting which customers are likely to leave, projecting cash flow.

Content generation tasks. Drafting reports, writing email responses, creating summaries, generating descriptions.

Optimisation tasks. Scheduling staff, planning routes, allocating resources, managing inventory levels.

These categories map to well-understood AI capabilities. You don’t need to know the technical details of how natural language processing or predictive modelling works. You just need to know that your problem fits a category that AI tools can address.

Step Three: Start With Off-the-Shelf Tools

This is where most businesses with no AI team should begin. Before you build anything custom, explore what’s already available.

The AI tool landscape in 2026 is vastly more mature than it was even two years ago. There are ready-made solutions for most common business problems:

• Customer support automation (Intercom, Zendesk AI, Freshdesk)
• Document processing and data extraction (various OCR and document AI tools)
• Sales forecasting (built into most modern CRM platforms)
• Content drafting (Claude, ChatGPT, Gemini, with business-grade versions)
• Scheduling optimisation (multiple purpose-built platforms)

These tools don’t require an AI team. They require someone willing to learn the tool, configure it for your context, and test it properly. That person is probably already on your team — they just haven’t been asked to do this yet.

Step Four: Know When You Need Outside Help

Off-the-shelf tools have limits. If your problem is genuinely unique to your business — if the data is proprietary, the workflow is non-standard, or the integration requirements are complex — you’ll need custom work.

This is where working with an AI consultancy can save you significant time and money. Not because you can’t figure it out eventually, but because the expertise gap between “we know what problem we want to solve” and “we know how to build and deploy the right solution” is wider than most people assume.

The key is engaging outside help after you’ve done steps one through three. If you approach a consultant with a clearly defined problem, a sense of which category it falls into, and evidence that off-the-shelf tools don’t quite fit, the engagement will be focused and efficient. If you approach them with “we want to do AI,” you’ll get a broad, expensive scoping exercise that may or may not produce useful outcomes.

Step Five: Measure Before and After

This is the step that almost everyone skips, and it’s the one that matters most for long-term AI strategy.

Before you implement any AI tool or system, measure the current state. How long does the task take? How often do errors occur? What does it cost in time and money?

After implementation, measure the same things. Did the tool actually improve performance? By how much? Was the improvement worth the cost?

These measurements do two things. First, they tell you whether the specific implementation was successful. Second, they build your organisation’s ability to evaluate AI investments rationally. Over time, this capability becomes your AI strategy — not a document that sits in a drawer, but a proven process for identifying, testing, and scaling AI applications.

The Strategy Is the Process

The businesses that succeed with AI without having an internal AI team are the ones that treat it as an iterative process, not a one-time project. They start small, measure results, learn from what works and what doesn’t, and gradually expand.

They don’t try to hire a chief AI officer or build a machine learning platform from scratch. They find practical problems, test practical solutions, and build confidence through demonstrated results. The strategy emerges from the practice.

That’s not as glamorous as a 50-page AI roadmap presented to the board. But it works, and it’s achievable for businesses of any size with any level of technical sophistication.