Markets
The valuable step after AI advice is deciding what to implement
Use a constraint filter so a plausible suggestion does not become an expensive distraction.
Zocri · 2 min read · Sources checked
Advice quality and implementation choice are different
The Kenyan entrepreneur experiment reports outcome differences associated with which advice participants implemented. This does not establish a universal mechanism or justify labeling owners by initial performance. It highlights a decision that can be overlooked when advice becomes cheap: which of the many plausible suggestions deserves action now?
An assistant can produce more initiatives than a small business can execute. A new promotion, a revised product range and a redesigned booking process may each sound sensible. Pursuing all of them can consume attention without resolving the constraint that currently limits the business.
Write the current constraint first
State the problem in observable terms. Orders may be delayed because one supplier is unreliable, inquiries may go unanswered or staff may be unclear about approvals. Keep the description specific enough that a proposed action can be connected to it.
Ask whether the suggestion addresses that constraint directly, merely improves a nearby activity or introduces a separate objective. The separate objective may be worthwhile later, but it should compete for time openly rather than enter the plan disguised as a solution to the original problem.
Apply three practical filters
Check affordability, operating capacity and reversibility. Can the business fund the action, can someone execute and maintain it, and can it be stopped without leaving a disproportionate burden? These are our proposed filters, not an experimentally validated scoring system.
Include the cost of distraction. An inexpensive software subscription can still be costly if it requires the owner to reorganize a process during the busiest period of the year.
Choose one accountable action
Assign a person, a limited scope and a review date. Define the evidence that would justify continuing or changing the action. If the suggestion depends on information nobody has checked, make obtaining that information the first task rather than proceeding as though it were already true.
Keep a list of deferred ideas so choosing one does not feel like losing all the others. The list should support focus, not become an obligation to implement every generated recommendation eventually.
Review the selection process itself
After a trial, ask whether the original filters identified the important costs and constraints. Improve them when they did not. Over time, the business should become better at selecting useful experiments, regardless of which assistant supplies the initial ideas.
The economic value of advice depends partly on what follows it. This article's filter makes that step explicit while preserving the experiment's limits. It offers a practical way to use abundant suggestions without letting a small company's scarce attention become the unpriced resource in every new plan.