Why Most Businesses Get Bad Results from AI (And How to Fix It)

Someone asked for access to my system after a LinkedIn post I wrote recently.

I had to say no. Not because I’m gatekeeping, but because the system only works because it knows my business inside out. It knows how I talk, who my ideal clients are, what my pipeline looks like, how I price my work, and what I’m optimising for this quarter. Hand it to someone else and it would be useless.

But here’s the thing. The reason it works isn’t the code. It’s not the commands or the automations. It’s the context I fed it before I built any of that.

And that’s exactly what most businesses are getting wrong with AI.

The Real Reason AI Gives You Generic Output

39% of UK SMEs say their biggest barrier to AI isn’t cost or skills. It’s that they can’t identify where it would actually help.

That stat used to surprise me. It doesn’t anymore.

Most business owners open ChatGPT, type “write me a cold email” or “summarise this document,” and get back something that reads like it was written by a corporate robot. Then they close the tab and tell their mates that AI is overhyped.

The problem isn’t the AI. The problem is what you gave it to work with.

AI doesn’t know your business. It doesn’t know your clients, your pricing, your voice, your processes, or what you’re actually trying to achieve. Without that, it can only give you generic answers. Because generic is all it has.

Context vs Commands

Here’s a distinction that changed how I think about this.

Most people use AI by giving it commands. “Write me a follow-up email.” “Summarise these meeting notes.” “Give me a marketing plan.” Every single time, they start from scratch. They explain the situation, describe what they want, and hope the output lands close enough to be useful.

That’s using AI as a tool. Tools do what you tell them in the moment. They have no memory, no understanding, no judgement. You get out exactly what you put in, every single time.

A system is different. A system knows your business well enough to do the right thing without being told.

When I run a command to draft outreach for a prospect, I don’t tell the AI how to write. It already knows my voice, the prospect’s pain points, what’s worked in previous emails, and what my pricing looks like. The output sounds like me because it knows what “me” sounds like.

The difference isn’t in the technology. It’s in the preparation. The commands are just the interface. The real product is the context underneath them.

What I Actually Wrote Down Before Building Anything

Before I wrote a single line of automation, I sat down and documented everything about my business. At the time, it felt completely unproductive. No code, no tools, no visible progress. Just writing.

That turned out to be the single thing that made everything else work.

Here’s what I wrote down, and what I’d recommend any business owner documents before using AI for anything serious:

1. Who you are and what you actually sell

Not your elevator pitch. The real version. What does a client actually buy from you? What does the delivery look like? What are your typical deal values? What makes you different from the alternatives?

AI can’t position your business if it doesn’t understand your business. Write this down in plain language. If someone who knew nothing about your industry read it, they should understand exactly what you do and why it matters.

2. How you talk

This is the one everyone skips. Your voice is what separates your emails, proposals, and content from sounding like you versus sounding like generic AI slop.

Write down the rules. Are you formal or casual? Do you use industry jargon or plain English? What phrases do you use naturally? What phrases would you never say? What impression should someone have after reading your emails?

I wrote a full voice guide. It includes real examples from emails I’d already sent, patterns I noticed in my own writing, and explicit rules like “never use em-dashes” and “never say leverage or synergy.” That guide gets read before the AI writes a single word on my behalf.

3. Your ideal client profile

Who are you actually trying to reach? What size company? What sector? What does the decision-maker look like? What problems do they have, in their own words, not yours?

This is critical for prospecting and outreach. Without it, AI will target everyone and connect with no one.

4. Your current pipeline and client status

Every deal, every contact, every follow-up date, every conversation you’ve had. This is the data that lets AI help you manage your sales process instead of just being a writing tool.

When my system flags that a prospect’s follow-up is overdue, it’s because this data exists and stays current.

5. Your pricing and objection handling

How do you price? What are the common objections you hear? How do you handle them? What are the real reasons people say no?

This is the stuff that turns AI-generated proposals from “here’s what we offer” into something that actually addresses what the prospect is worried about.

6. Your goals and strategy

What are you optimising for right now? Revenue? Client retention? Market expansion? What decisions are open? What trade-offs are you making?

This is what lets AI give you strategic advice instead of generic suggestions. It’s the difference between “you should do more marketing” and “your pipeline has three deals going stale this week, and you haven’t sent outreach in six days.”

The Compounding Effect

Here’s where it gets interesting. Context isn’t a one-time thing.

Every interaction I log, every deal I close or lose, every outreach pattern I learn from feeds back into the system. The context gets updated constantly. And because the context gets smarter, the output gets smarter.

After a meeting, I capture what happened, what the prospect said, what objections came up, and what the next step is. That information updates my pipeline data. Next time I draft a follow-up for that prospect, the AI already knows the full history.

When I close a deal, I capture why it worked. When I lose one, I capture why it didn’t. Over time, that builds a pattern library. The system starts to recognise what kinds of outreach work best for what kinds of prospects. It spots patterns I wouldn’t notice manually.

It doesn’t just do what I tell it either. It pushes back. It tells me when my pipeline is going stale, when I’m avoiding outreach, and what the highest-leverage thing I should be doing today. Less of an assistant. More of a co-founder who doesn’t let me coast.

That compounding effect is only possible because the context exists and stays current. Without it, every interaction starts from zero.

What Context Looks Like for Different Businesses

This isn’t specific to what I do. Any service business can do this. Here’s what it might look like in practice:

A recruitment agency would document their candidate sourcing process, their client verticals, their fee structure, how they screen candidates, their CRM workflow, their email templates, and their typical objections from both clients and candidates. Context like “we specialise in finance roles, £40k–£80k, and our biggest competitor is Robert Walters in this region” turns AI from a generic writing tool into something that actually understands your market.

An accounting firm would write down their service tiers, their client onboarding process, how they handle year-end versus monthly bookkeeping, their pricing per service, their compliance deadlines, and how they communicate with clients. The AI could then draft client communications that reference specific deadlines, flag overdue submissions, and suggest upsells based on client activity.

A trades business (plumber, electrician, builder) would document their service area, their pricing for common jobs, how they handle quotes, their booking process, their follow-up approach for quotes that haven’t converted, and their supplier relationships. Even something as simple as “we charge £75 call-out plus £45/hour, minimum 2 hours, parts extra” gives AI enough to draft accurate quote follow-ups.

A consultancy would document their methodology, their engagement structure, typical project timelines, their positioning against competitors, how they scope work, and their proposal format. With that context, AI can help draft proposals that follow your structure and speak your language instead of producing something you have to rewrite from scratch.

The specifics change. The principle doesn’t. Write down how your business actually works, and AI has something real to work with.

Common Mistakes

Using AI without any context

This is the big one. Opening a chatbot and asking it to do something for your business without telling it anything about your business. You wouldn’t hire a new employee and expect them to write client emails on day one without any onboarding. AI is the same.

Treating AI as a search engine

AI isn’t Google. If you’re using it to look up facts, you’re using about 5% of what it can do. The real value is in applying knowledge to your specific situation. But it can only do that if it has your specific situation.

Expecting it to know your voice without examples

“Write this in a professional but friendly tone” is meaningless. Professional and friendly to who? Show it real examples of how you’ve written before. Give it rules. Tell it what you’d never say. The more specific you are, the less you’ll have to edit the output.

Giving up after one bad output

The first output is rarely perfect. That doesn’t mean AI doesn’t work. It means you need to iterate. Give it feedback, add more context, try again. The people who get the most from AI are the ones who treat it as a working relationship, not a vending machine.

The AI Is Only as Good as What It Knows About You

You don’t need to build what I built. You don’t need 42 custom commands or an AI-powered sales system. But if you want AI to actually be useful in your business, start here: write down how your business works.

Your clients. Your processes. Your pricing. How you talk. Give it that before you ask it to do anything.

The businesses that get the best results from AI aren’t the ones with the fanciest tools. They’re the ones that took the time to give AI something real to work with.

If you had to write down everything AI would need to know about your business to actually help you, where would you start?