Why most business owners are getting poor results from AI

So why are businesses getting poor results from their AI investment. It’s not the tool, writes Marie Ryan, it’s the brief.

I’ve lost count of the number of times I’ve heard someone say, “I tried ChatGPT. It wasn’t that good.”

Usually it goes like this. They open the tool, type something like “write me a LinkedIn post” or “give me a marketing strategy,” lean back in their chair thirty seconds later and conclude that AI is overrated.

The tool isn’t the problem. The brief is.

Think about walking into an architect’s office and saying, “Build me a house.” No budget. No location. No idea who’ll live there. Then being disappointed with the drawings.

That’s not the architect’s fault.

👉 Ready to reclaim an extra day a week? Secure your Google AI Professional Certificate at no cost. Apply now and get started!

AI works exactly the same way. It’s an exceptionally capable assistant waiting for clear instructions. The better the brief, the better the result. Yet most people treat it like a mind reader rather than what it actually is: a very talented new employee on their first day who knows nothing about your business until you tell them.

Prompt engineering isn’t as complicated as it sounds

Search for “prompt engineering” online and you’ll find frameworks, job titles, secret techniques and endless articles about unlocking AI’s full potential. It’s enough to make you think you need a qualification just to ask a question.

You don’t.

After delivering AI workshops to hundreds of professionals across Ireland –from marketing teams and HR managers to finance directors and leadership teams – I’ve found that the people getting the best results aren’t the most technical. They’re simply good at giving instructions.

Prompting is a communication skill, not a technical one. The same ability that helps you brief a designer, delegate to a colleague or explain a project to an agency produces better results from AI.

In every session I run, I come back to the same four things. I call them the Four Horsemen of AI Prompting: Role, Context, Task and Format. Most people only think about one of them – the task. Including all four changes everything.

Role: Build yourself a board of advisers

When people hear “give AI a role,” they usually think it means saying something like act as a marketing expert. That’s useful, but it misses the more interesting opportunity.

Role isn’t about making AI sound more knowledgeable. It’s about changing its perspective.

In one of my training sessions, I use the example of a gym owner looking for a marketing strategy. The business stays the same. The task stays the same. The only thing I change is the role.

I ask AI to act as a marketing intern. Then a marketing strategist. Then a personal trainer. Finally, an accountant.

The marketing intern comes back with energetic, creative ideas – Reels, member challenges, a dedicated selfie corner with a neon sign so members naturally post photos and tag the location. Simple, inexpensive and genuinely clever.

The marketing strategist thinks differently. Local SEO, Google Business Profile, referral campaigns, return on investment.

The personal trainer barely mentions marketing. Their focus is on helping members get results, because from their perspective the experience is the marketing.

Then the accountant walks in. Nobody invites accountants to marketing meetings. But some of the strongest ideas come from that chair. Before spending more money, can we bring former members back? Can we increase lifetime value? Can we generate more referrals from the people who already trust us?

None of those answers was wrong. Together they were far stronger than any single answer on its own.

Most of us look at business problems through our own lens. A marketer thinks about awareness. A finance director thinks about cost. AI lets you borrow those perspectives in minutes – and sometimes the best idea comes from the role you’d never have thought to ask.

Context: AI will follow the story you give it

Context is the background information you provide before asking AI to do anything. But it’s also more than that. Context shapes the story AI believes it’s operating within – and that can transform the answer completely.

One of my favourite demonstrations uses the same website, reviewed three different ways.

First, I ask for an objective review. The feedback is balanced, practical and useful.

Then I tell AI the website is excellent – clear, easy to navigate, strong content – and ask it to review it again. Suddenly it’s one of the finest websites on the internet. Every headline is compelling. Every design choice is intuitive.

Then I tell it the site is clunky, confusing and generic, and ask for a critical assessment. Within a few paragraphs it sounds as though Google’s quality team are about to arrive at my door with a demolition notice.

The website never changed. Only the context did.

This matters because AI is very good at following instructions – sometimes a little too good. If you ask it to prove something is brilliant, it usually can. If you ask it to prove something is terrible, it can do that too.

Context shouldn’t be used to confirm what you already believe. It should be used to help AI understand your actual situation. Ask for analysis, not agreement. And if you’re unsure what context to provide, ask AI directly: Before answering, what are the five most important questions you need me to answer so you fully understand the situation? What it asks is often more useful than what it eventually produces.

Task and Format: The finishing touches

The task is what most people already think about – what you want AI to do. The problem is that vague instructions produce vague answers. “Help me with marketing” is not a task. “Analyse these customer reviews and identify the five most common frustrations” is.

Start with a strong verb. Analyse. Compare. Challenge. Rewrite. Prioritise. Summarise. Each one tells AI how to think, not just what to write. And if you want genuinely useful feedback, give AI explicit permission to disagree with you. It naturally wants to be helpful, which means it can be a little too agreeable by default.

Format is the ingredient people forget most often. AI defaults to paragraphs – not because paragraphs are always best, but because they’re the safe choice. If you know how you want to use the information, say so. A table. A checklist. A one-page executive summary. A 30-day action plan. Specifying the format often saves more time than anything else in the prompt.

A better brief looks like this

Here are all four horsemen in one prompt:

Act as an experienced Marketing Director. You’re advising a professional services firm that wants to increase visibility among senior decision-makers. The organisation has a strong reputation but relies heavily on referrals, with a modest budget and limited internal time. Develop a practical 90-day marketing plan and present it as a table showing the activity, objective, expected impact, estimated effort and suggested timeline.

That’s not prompt engineering. It’s a well-written brief.

The bigger idea

The businesses getting the most from AI aren’t using better tools. They’re giving better briefs.

More importantly, they’re becoming better thinkers – clearer about what they’re trying to achieve, more precise about what success looks like, more willing to challenge assumptions before committing to a decision.

That’s the real opportunity AI offers every organisation. Not that it will think for you. But that it encourages you to think more clearly.

You don’t need better prompts. You need to ask better questions.

Top image: Photo by Immo Wegmann on Unsplash

👉 No tech background needed to bring your business into the AI age! Secure your Google AI Professional Certificate to complete the certification at no cost. Apply now and get started!

• Bank of Ireland is welcoming new customers every day – funding investments, working capital and expansions across multiple sectors. To learn more, click here

• For support in challenging times, click here

• Listen to the ThinkBusiness Podcast for business insights and inspiration. All episodes are here. You can also listen to the Podcast on:

• Spotify

• SoundCloud

• Apple