I’ve spent thousands of hours looking at the internal workings of businesses, from high-growth startups to established professional firms. Right now, every owner is asking me the same thing: "Penny, what should our AI strategy for SME look like this year?"
They expect me to give them a list of tools—the latest LLM for their marketing, an automated agent for their support, or a predictive model for their inventory. But usually, I give them something they don't want to hear: You don’t have an AI problem. You have a clarity problem.
We are currently witnessing the birth of The Automation Anxiety Paradox: businesses that are the most desperate to adopt AI are often the ones least prepared to use it, because their internal processes are a tangled web of 'we’ve always done it this way' and 'John knows how that works.' If you try to layer AI on top of a mess, all you get is faster, more expensive chaos.
The Hallucinating Workflow
When we talk about AI 'hallucinating,' we usually mean a chatbot making up a fact. But in a business context, the most dangerous hallucinations happen when you try to automate a process that isn't actually defined. I call this the Hallucinating Workflow.
Imagine you want to automate your lead intake. You buy a fancy AI tool, connect it to your CRM, and wait for the magic to happen. But because your Standard Operating Procedures (SOPs) are messy—maybe different sales reps follow different steps, or your 'qualified lead' definition is 'vibes-based'—the AI has to guess. It starts sending the wrong follow-ups to the wrong people at the wrong time.
In this scenario, the AI isn't failing. The engine is clogged. The AI is simply highlighting the fact that your underlying logic is broken. Before you spend a penny on a complex AI strategy for SME, you have to accept a hard truth: You cannot automate what you cannot define.
Why 'AI-Ready' is a Documentation Problem
Most business owners think being AI-ready means having a high-spec tech stack. It doesn't. It means having a business that is Functionally Literate.
A Functionally Literate business is one where every core process—from how an invoice is approved to how a client is onboarded—is documented with such clinical clarity that a stranger (or an algorithm) could execute it.
When I look at professional services savings, the biggest wins don't come from the most expensive AI tools. They come from firms that spent three months ruthlessly cleaning their SOPs before they touched a single line of code. They turned 'messy human intuition' into 'structured logical steps.' Once you have that, the AI adoption is almost trivial.
The Cost of the Clog: The 'Chaos Multiplier'
Messy SOPs act as a 'Chaos Multiplier.' In a manual business, a messy process costs you time and perhaps a bit of employee frustration. In an AI-first business, a messy process scales the error rate exponentially.
If your business accountant costs are skyrocketing despite 'automating' your bookkeeping, it’s likely because the AI is spending 80% of its time trying to reconcile data that was entered inconsistently in the first place. You’re paying for the AI to try and solve a puzzle that you haven't even finished designing.
This is why I often tell people to look at a Penny vs Business Consultant comparison. A traditional consultant might bill you £20,000 to 'optimise your strategy.' I’d rather you spend £29 to realise that your strategy is fine, but your instructions are garbage. One is a vanity project; the other is a profit driver.
The Process-First Protocol
To move from chaos to AI-driven efficiency, I recommend a framework I call the Process-First Protocol. It consists of three non-negotiable stages:
- Extract: Get the process out of the 'expert's' head. If a task requires 'intuition' or 'judgment' at every step, it’s not a process yet—it’s an art form. You can't automate art.
- Prune: Most SOPs are bloated with steps that exist only because of a software limitation from 2014. Before you automate, delete. If a human shouldn't be doing it, why should the AI?
- Standardise: Every output must have a binary definition of success. 'The client is happy' is not a metric. 'The contract is signed and uploaded to Folder X' is a metric.
Only after these three steps are complete should you even think about the tools.
The 90/10 Rule of Operations
I’ve observed a pattern across hundreds of businesses that I call The 90/10 Rule. When you clean a process and apply AI, the AI can usually handle about 90% of the heavy lifting. The remaining 10% is where the human 'edge' lives—the empathy, the complex ethics, and the high-level strategy.
The mistake most SMEs make is trying to build a strategy that automates the 100%. They fail because the 10% is too messy to codify. By focusing on the 90%—the repetitive, logic-based core—you free your team to actually do the work that AI can't touch.
Your Real Competitive Advantage
In the next two years, AI tools will be a commodity. Every one of your competitors will have access to the same LLMs and agents that you do. The tools won't be the differentiator.
The differentiator will be Process Propriety. The business with the cleanest, most logical, and most documented operations will be able to plug in new AI capabilities 10x faster than the business still running on 'messy' logic.
Your AI strategy for SME shouldn't start with a ChatGPT subscription. It should start with a blank document and a simple question: If I disappeared tomorrow, could a machine follow these instructions to keep the business running?
If the answer is no, you don't have an AI strategy. You have a clog. And it’s time to clear it.