For solo businesses, AI’s practical value lies in taking on defined work that would otherwise consume an owner’s time. App development, customer follow-ups and revenue analysis are among the tasks Ben Angel demonstrates in a video describing seven ways to deploy AI agents. The appeal is straightforward: a small operator can assign work to software without adding staff or needing to write code. But the usefulness depends on the task, the quality of the instructions and the owner’s ability to check the result.
Start with the bottleneck
The useful first step is to identify a business task with a clear result. An owner might want to understand a change in sales, contact customers who’ve gone quiet or create an internal tool. These are different jobs, but each can be framed as a request with an observable outcome. A brief should give the AI relevant context, set boundaries and specify what a successful result looks like.
That structure also helps divide work among specialist agents. The owner sets priorities and approval points; each agent receives a specific responsibility. In Angel’s demonstrations, AI works with browser-based tools, passes a project to Lovable and coordinates tasks across agents. The examples are framed as ways to move a job from instruction to execution without adding staff or writing code. The distinction matters: delegating a bounded task is easier to assess than handing over a vague goal such as “grow the business.”
There’s a practical reason to begin with a bottleneck. A solo owner has limited time to supervise new systems, so a workflow that already causes friction offers a natural test. If the assignment is routine and the output can be checked, the owner can compare the AI-assisted process with the existing one. If the task involves judgment that’s difficult to describe or verify, delegation may create more review work than it saves.
Productivity claims need a business test
An Intuit QuickBooks article citing its 2026 AI Impact Report says 78% of U.S. businesses reported improved productivity from AI, while 43% reported higher revenue. Those figures offer context, but they don’t establish what a particular business will gain. Surveyed businesses may use different tools for different jobs, and a reported improvement doesn’t show how much came from AI or how consistently it can be repeated. The numbers here are useful as a signal of adoption and reported experience, not a forecast for any one operator.
For a small business, the relevant question is whether one workflow improves in a way that can be checked. That might mean measuring time spent on customer outreach, tracking whether follow-ups bring customers back or checking whether an analysis helps explain a revenue change. The owner should decide what evidence would count before handing over the task; otherwise, a polished output can be mistaken for a useful one.
The underlying work still needs review. A clear assignment can make delegation more useful, but it doesn’t remove the need to assess the output. AI may organize information or draft a next step, while the owner remains responsible for deciding whether the result fits the customer, the business and the available evidence. Here’s the thing: saving time on production has little value if that time is then spent correcting errors that a simple review process could have caught earlier.
What the demonstrations show—and what they don’t
Angel presents the approach as a way for a one-person business to take on work normally spread across a team. His book, The Wolf Is at The Door, also identifies AI delegation as a skill for adapting to changing work. The video includes seven experiments, among them a customer recovery brief, a ChatGPT-to-Lovable handoff and a revenue diagnosis assignment.
Those examples show how a task might be passed between tools or divided among agents. They don’t, by themselves, establish that every business will get a reliable result or a financial return. Demonstrations tend to make the handoff visible; the less visible work is often deciding what information an agent can use, catching weak assumptions and handling cases that don’t fit the brief (and this is the part most people overlook). That review burden is part of the workflow, not an exception to it.
If you’re working in this space, treat each demonstration as a prompt to define a test, not as proof that a process is ready to run unattended. The numbers from a customer recovery task, for instance, would matter only if the owner can connect the follow-up to an outcome and distinguish that result from other causes. A revenue diagnosis may help organize questions, but the explanation still needs to be checked against the business’s actual records.
Build from one assignment
The practical takeaway is to choose a recurring bottleneck, define the expected result and test whether the AI-assisted workflow saves time or supports a business outcome. Keep the first assignment narrow enough that the owner can review the result without creating a second job. One well-scoped assignment gives an owner a concrete starting point for deciding where AI belongs next.
This is more significant than it looks: a small business doesn’t need to automate everything to benefit from a better division of work. It needs to know which tasks can be described clearly, which outputs can be checked and where human judgment remains necessary. Start with one assignment. Then judge the process by what it actually changes.
