SMEs that experiment with AI quickly stop asking whether to adopt it and start asking how to use more of it. The shift happens once teams see AI working across operations, marketing, reporting, and decision-making. Companies that reach this question first build an advantage that is hard to reverse.
This is why I hope more SMEs start experimenting with AI.
Because once they experience how much it can help — with operations, marketing, reporting, decision-making, and execution — the question will no longer be “Should we use AI?”
It will become: “How do we use more AI in our business?”
I have watched this shift happen in real time. In PAIBA workshops, in deployments with Olern clients, in conversations with business owners who started with hesitation and ended asking when they could do more. The pivot is always the same. Once people see what AI actually does inside their workflows, the skepticism evaporates. Not because the technology is flawless. Because the value is visible, and visible value changes behavior.
The companies that experience this shift early have a serious edge.
Why most SMEs are still stuck on the wrong question
Most SMEs are still asking whether to adopt AI. That sounds like responsible due diligence. It is often a delay dressed up as caution.
The “should we” framing puts AI on trial before anyone has experienced the outcome. And because AI is unfamiliar, the default verdict is skepticism. Teams argue about risks they imagine rather than problems they have already seen solved. They read about hallucinations and conclude the technology is not ready, rather than running a prompt and seeing what actually happens.
Companies that move past this question are not reckless. They are further along the learning curve. They have run enough experiments to stop debating whether AI is useful and start asking where to use it next.
That is a much more productive place to operate from. And the gap between companies in “should we” mode and companies in “how do we use more” mode is widening every month.
Token maxxing and outcome maxxing — two approaches to getting there
You hear two phrases circulating in serious AI discussions: token maxxing and outcome maxxing.
Token maxxing is the volume play. Feed the model everything. Use it for every task. Maximize the number of workflows that have an AI step. Some teams treat this as a metric: how many processes touched AI this week? How many reports had a first draft from the model?
Outcome maxxing is the strategic play. Identify the tasks where AI changes the result, not just the speed. Focus on the decisions, the reports, the customer interactions where better output actually matters. Fewer use cases, higher impact.
Both approaches have a place — and understanding when to use each one is part of the learning.
Token maxxing builds the muscle. Teams that run AI through many tasks develop the instinct for where it helps and where it does not. That instinct is hard to teach in a classroom. You build it by using the tools repeatedly, across different types of work, and paying attention to what happens. Comfort with AI comes from volume.
Outcome maxxing builds the results. Once teams know where AI moves the needle, they concentrate effort there. They stop chasing novelty and start compounding gains. The reports that matter improve. The decisions that cost time get faster. The customer touchpoints that affect retention become more consistent.
The companies I work with through PAIBA and Olern tend to move from token maxxing to outcome maxxing over time. They start with volume, which builds confidence. Then they narrow toward impact, which builds measurable results. The sequence matters — rushing to outcome maxxing before building comfort leads to underuse.
What gets in the way for SMEs specifically
For most small and medium enterprises, the obstacle is not the technology. AI tools are accessible, affordable, and improving every month. The obstacle is the gap between the first experiment and the second.
Teams try a tool, get inconsistent results, and conclude it is not ready. What usually happened is the prompt was too vague, the use case was poorly chosen, or the output was judged against a standard AI was never going to meet. The tool did not fail. The experiment was not designed to succeed.
This is a solvable problem — but only if teams understand what went wrong.
The most common failure mode is starting with a task that is too open-ended. Asking AI to “write a marketing plan” produces a generic output. Asking it to “write a first draft of a Facebook post announcing our September sale for home furniture buyers” produces something usable. The difference is specificity, and specificity is a skill that develops with practice.
The second most common failure mode is treating one bad output as a verdict. One prompt does not tell you much. Twenty prompts on the same type of task, refined over two weeks, starts to reveal patterns. You learn where to be specific, where context matters, where to check the output more carefully before using it.
The companies that move fastest are not the ones with the most technical skill. They are the ones who treat AI adoption like any other operational learning process: run small experiments, observe what works, refine, and repeat.
How to actually move from “should we” to “how do we use more”
The shift does not happen from a webinar or a seminar. It happens from doing.
Start with a task you do every week that has forgiving output. Drafting internal updates. Summarizing meeting notes. Creating first versions of sales proposals that a human will edit and personalize. These are not the highest-stakes uses of AI — but they are the ones where the learning cost is low and the improvement is visible enough to be motivating.
Run the same task through AI for two full weeks. Not one prompt. Two weeks of the same task, with small refinements each time. At the end of the two weeks you will have a working approach, not just a one-time experiment. You will also have a concrete answer to “how long does this take with AI versus without it?” — and that comparison, once you have seen it yourself, changes the conversation.
Pay attention to what surprises you. Every time AI does something you did not expect it to do well, that is a signal worth following. In PAIBA workshops, the moment that most often shifts skeptics is not when AI does something impressive. It is when it handles something tedious that was eating two hours a week, and it does it in four minutes. That is the moment “should we?” becomes “where else can we do this?”
Ask your team the same question. Once you have one working use case, ask the people doing the work: where else could this apply? The people closest to the repetitive work almost always have answers. They know which tasks eat time without producing value. Give them permission to experiment, and give them a low-stakes space to do it.
One thing is clear. The companies that learn how to work with AI faster will have a serious advantage.
How about you? Are you maxxing out your tokens?
I would love to hear where your team is on this. What was the first use case that made you stop asking “should we?” — share it in the comments.
Frequently Asked Questions
What does it mean for an SME to move from “should we use AI?” to “how do we use more AI?”
The shift happens when a business stops debating whether AI is worth trying and starts asking where to expand it. It typically follows the first time a team sees AI working inside one of their real workflows. Once the value is visible in a familiar context, the resistance drops and the question becomes operational rather than philosophical.
What is the difference between token maxxing and outcome maxxing?
Token maxxing means maximizing the volume of tasks that go through AI — using it broadly across the business to build comfort and instinct. Outcome maxxing means concentrating AI use on the tasks where it changes the result, not just the speed. Most businesses benefit from starting with token maxxing to build confidence, then shifting toward outcome maxxing once they know where AI has the highest impact.
Why do AI experiments fail for small businesses?
Most failed AI experiments fail because of how they are set up, not because the technology does not work. The two most common causes are prompts that are too vague (asking for a generic output instead of a specific one) and judging the tool based on one or two tries rather than a two-week run of the same task. AI output improves significantly when prompts are specific and when teams develop the instinct through repetition.
How long does it take for a small business to see real results from AI adoption?
A team that picks one repetitive task and runs it through AI consistently for two weeks typically has a working approach by the end of that period. Results come faster when the task is specific, the output is forgiving (so errors are low-cost), and someone is paying attention to what works and what does not. The timeline is shorter than most business owners expect once they start.
What kinds of tasks should SMEs start with when adopting AI?
The best starting tasks are ones that happen every week, have low-stakes output (a human will review before it is used), and currently eat time without producing strategic value. Examples include drafting internal updates, summarizing meeting notes, writing first versions of sales proposals, and generating social media post drafts for review. These tasks build the team’s AI instinct without exposing the business to high-risk outputs.
How do SMEs build competitive advantage through AI?
The advantage comes from compounding small gains over time. Companies that adopt AI earlier develop faster instincts for where it helps, which lets them find higher-impact applications sooner. The gap between early movers and late adopters is not just the tools they use — it is the organizational knowledge about how to use them well, which cannot be acquired overnight.
