If you are an SMB owner or manager, and have read my first two posts from this series, you may be wondering how you can get started. In this session, I move from basic information and concepts to strategy and tactics. I’m going to show you how to design your first AI experiment -the process to follow. In my next one, we’ll talk about the guardrails and governance that can keep you safe. My hope is that this will help you to move from strategic inertia to strategic adaptation.
To begin, you need to ask yourself – “What are our most pressing business problems today?”, “What is holding us back from higher growth or profitability?” Your answers will direct you toward the right AI pilot. Let’s walk through an example.
Start With a Business Problem
Something that is common to almost all small businesses is the need to improve marketing outcomes. For example - “We need to increase the number of net new customers we are adding per month, without increasing our costs.” This is a great target for an AI-enabled solution because of its potential for high impact with low risk.
Establish Measurable Baselines
We need at least one baseline metric against which to measure success. For this example, we’ll use the marketing portion of our customer acquisition cost (CAC). This is calculated by taking our total marketing spend over a period of time and dividing by the number of net new customers acquired in that period. If we added 120 customers over the last 12 months and spent $30,000 on marketing, that’s $250 per customer. Keeping the same budget for the next three months ($7,500), let’s target to reduce the CAC to $200. That means we have to add about 38 new customers – 8 more in this quarter than we did in the same quarter last year.
Run a Pilot – One Change, One Metric, 90 Days
Next, we should review our current marketing plan and pick an area that we think can be improved to use as a pilot. Top possibilities for AI include (1) content creation and copy editing, (2) email marketing automation and (3) campaign scheduling and platform publishing. In most cases, you will already be using various tools for these tasks. Many of the latest versions will include AI capabilities. Check with your vendor and start there before buying anything new. If you do need new tools, an initial subscription for things like email automation or campaign scheduling can likely be had for less than $100 per month – a small fraction of your overall marketing budget. For content creation and copy editing, you have access to free versions of multiple LLMs (e.g. ChatGPT, Claude, Gemini, Grok).
For the purpose of this example, let’s focus on content creation and copy editing. Your marketer should use the chosen LLM to create drafts, refine language, add reference links, etc. If that person currently spends 8 hours a week updating social media content and, during your pilot, that drops to 3 hours, then 5 hours a week can be redirected to higher value tasks, like figuring out how to convert a higher percentage of website browsers into customers. Keep in mind that the marketer must be the human-in-the-loop that performs final edits and approves all materials for publication.
Plan to run your pilot for at least 90 days. That’s long enough to start seeing results but short enough to limit your investment, especially if it’s not working.
Measuring Results and Capturing Truth
In our example, the result of our pilot is easy to verify. At the end of the 90-day trial, the CAC will either hit $200 per customer, or it won’t.
Whether the pilot is successful or not, it is essential to satisfy yourself that there is a clear link between the change that was made using AI and the result. Be honest with yourself about causality. If you happen to have a great quarter due to external market factors, don’t attribute all of your wins to the pilot. By the same token, if the pilot fails, don’t automatically assume AI is the villain. Try to get to root cause and then decide if a pivot is needed or a longer trial is necessary or perhaps your marketer needs more training. Learn from it.
Finally, to verify true success, it will be important to capture sentiment from the people involved. Is your team excited and ready to embrace AI or are they fearful? If the reactions you get are not enthusiastic, you will have some change management to do before you start the next pilot. If you're going to scale, not just run pilots, you'll need the whole team behind you.
Summary
I chose the example above to illustrate the following principles, which are essential to a successful pilot.
Difficulty is low. We are simply using an LLM to speed up the process. Business risk is low, because we still have a marketer overseeing the work. Potential impact is high; we are targeting a reduction of 25% in the CAC.
We selected one thing to change – how content is developed. If all other things remain relatively the same and we achieve our desired outcome, chances of establishing a strong correlation or causation are higher.
We targeted a single SMART metric (specific, measurable, achievable, relevant and time-bound) that is easily baselined and calculated.
The CAC result is easy to verify.
If you are familiar with design-thinking, you’ll recognize the method. This is how you move from strategic inertia to strategic adaptation.
The discipline of decision-making — weighing competing goals, understanding nuance, balancing speed with risk — is a defining skill for small business owners. Implemented properly, AI can handle many tasks faster and more accurately than humans ever will, but it cannot bear the cost of a wrong decision or suffer the reputational loss that might accompany it. Leadership demands accountability and, for now, that is beyond the grasp of any AI system. In my next instalment of this blog post series, we’ll finish building the framework by talking about guardrails and governance, topics that are extremely relevant for AI usage.
I hope you’ve enjoyed this series so far, and I invite you to provide your feedback. Thanks for reading.
© 2026 by Roy Gowler. All rights reserved.
This article was originally published in December 2025 and posted on Medium.com. As its author, I have updated it and posted it to my own website to increase visibility and reach.
Back to Insights.