Solo founder: AI agent or your next hire?

A solo founder should choose an arrangement for a task rather than buy an abstract promise of productivity. A hire adds judgement and capacity. An AI agent handles a defined digital workflow. An external specialist supplies expertise. Manual work keeps full control while consuming the owner's time.

Name the task and its owner

Write down one week of work and sort it into tasks:

• recurring and rule-based;
• recurring with changing language or documents;
• relationship-heavy or strategic;
• regulated or high-consequence;
• one-off and unlikely to repay setup work.

For each task, record inputs, output, tools, exceptions, review time, and the person who answers when something fails. If the task has no clear output or owner, a new hire and an agent will both add coordination work.

Four ways to add capacity

Keep it manual

Manual work suits a small volume, a changing process, or a task where the founder's judgement creates the value. Track the hours and the delay to other work. A manual route can be an intentional choice while the process becomes clearer.

Hire a person

A hire fits work that needs context, initiative, customer relationship, or frequent exceptions. Include recruitment, onboarding, management, equipment, leave coverage, and the time needed to transfer the process. A person can improve the process while doing it, which may be the main benefit.

Use an external specialist

An accountant, coordinator, copywriter, or call service can own a process without adding an employee. Define the service level, data access, deliverables, handoff, and continuity plan. Outsourcing works when the task is well enough defined for another person to run it.

Build or buy an AI agent

An agent fits a repeatable digital process with accessible inputs and an explicit action set. It can classify messages, extract fields, prepare a draft, retrieve information, or route a request. The owner still reviews exceptions, maintains source data, controls permissions, and handles outages.

An agent is a poor fit when the process changes weekly, the source data is unreliable, or one wrong action could create a serious commitment without a timely human check.

Compare the total work

Estimate the full cost for the same period:

capacity cost = setup + recurring fee or labour + supervision + tools + correction time + outage cover

Then estimate the value of the hours released. The relevant result is work the founder can actually stop doing, with acceptable quality and a clear owner for the remainder.

Run a small test

Choose one task with a stable input and a measurable output. Use permission-cleared records. Define the review rubric, maximum correction time, allowed tools, and a stop condition. Keep the agent in a review queue until it meets the written target over representative cases.

Measure:

• completed cases;
• review and correction time;
• exceptions by type;
• data or permission failures;
• maintenance work after a source or tool change.

Compare those results with the current manual time and with an external specialist's quote. Do not count a draft as recovered capacity until you have included review and follow-up.

Choose the arrangement

Hire when the business needs judgement, relationships, or a person who can shape the process. Use a specialist when the task needs expertise and a clear service handoff. Use an agent when the task is digital, repeatable, and reviewable. Keep it manual when the volume or uncertainty makes setup hard to justify.

The Eurostat overview of AI use in enterprises can provide market context, while your own task log should drive the decision. Bring that log, the output rubric, and the available support time to an AI process scan.