The Simple Test That Tells You Whether to Hire or Automate

Before you approve the next hire, write down the 10 things you expect that person to do.

Not the job title. Not the qualifications. The actual recurring work.

Now separate the list into four categories: judgment, relationship, repetition and coordination.

That exercise has become one of the most useful ways I know to tell whether a growing company has a headcount problem or a workflow problem.

Headcount is an expensive default

When a team is stretched, hiring feels like the responsible answer. Sometimes it is. But growth creates work in two ways. It creates valuable new work, and it also creates coordination work around the valuable work.

The second category is where job descriptions quietly become storage units for broken workflows.

The U.S. Census Bureau found that overall business AI use was still only about 17% to 20% across late 2025 and early 2026, even as larger firms and knowledge-intensive sectors adopted at much higher rates. The more detailed Census research on AI diffusion shows that among adopting firms, Sales and Marketing, Strategy and Business Development, and IT are among the most common functions using AI.

That matters because many of the tasks creating headcount pressure in growing companies sit exactly in those functions: lead routing, follow-up, scheduling, reporting, data movement, content operations and customer communication.

AI can create capacity, but only on the right work

The Stanford AI Index 2026 reports AI use in at least one business function at 88% of surveyed organizations, yet agent deployment remains early. Adoption is moving faster than operating-model redesign.

The strongest research on productivity also argues for precision. An NBER study of more than 5,000 customer support agents found a 14% average productivity increase from AI assistance, with substantially larger gains for less experienced workers. A Harvard Business School field experiment found major speed and quality gains for consultants on tasks inside AI’s capabilities, but also showed that performance can suffer outside that boundary.

The message for founders is not “automate everything.” It is “stop treating every task in a job description as equally human.”

Run the 4-bucket test

1. Judgment: Decisions with consequences, ambiguity or accountability. Strategy, negotiation, creative direction, hiring, clinical or legal judgment and sensitive customer decisions belong here. AI can assist, but a named human should own the outcome.

2. Relationship: Work where trust is part of the value. Sales conversations, leadership, coaching, conflict, partnership development and customer recovery often lose value when treated as a pure efficiency problem.

3. Repetition: High-frequency work with stable rules. Reminders, routine follow-ups, scheduling, data entry, status updates and templated communication are the obvious automation candidates.

4. Coordination: Work that exists because systems or people are disconnected. Copying information between tools, chasing approvals, reconciling versions and asking the same status question repeatedly usually need redesign before automation.

This sequence matters because automation amplifies process design.

If the workflow is clear, the inputs are reliable and exceptions are understood, automation can remove a remarkable amount of administrative friction.

If ownership is fuzzy and the data is wrong, automation makes the confusion faster.

A job description should not be a storage unit for broken workflows.

How to decide whether to hire

I use a simple founder-level heuristic. It is not an industry standard, but it forces the right conversation.

• If the bottleneck is mostly judgment or relationship work, hiring or developing a person is probably the right answer.
• If the bottleneck is mostly repetition, test automation before adding permanent headcount.
• If the bottleneck is mostly coordination, redesign the process first.
• If the job contains all three, separate the workflow so the human role is built around the work that deserves a human.

Do not measure AI by how often people use it

Microsoft’s Work Trend Index describes the fragmented attention environment many knowledge workers now operate inside. Adding another tool to that environment is not automatically progress.

For an AI or automation pilot, baseline the workflow before you touch it. Measure cycle time, throughput, quality, error or exception rate, human review time and customer impact. Then run the pilot on one defined workflow.

If the tool gets used constantly but the workflow does not improve, you have adoption without value.

Put a price on the workflow before you put a salary on the role

Founders tend to compare automation cost with salary, but that is too narrow. The current workflow already has a cost, even if it is buried inside several salaries.

Take one recurring process and estimate how many minutes it requires, how often it happens and which people touch it. Then add the cost of delays and avoidable rework. You do not need accounting-grade precision. You need enough visibility to know whether the proposed hire is solving valuable work or absorbing friction.

This is also where small companies can compete with larger ones. They may not have the budget to add a specialist every time volume increases, but they can often redesign a workflow faster because there are fewer layers protecting the old way of doing it.

Use a reversible pilot before making an irreversible operating choice

A new hire is a meaningful commitment. A workflow pilot can be much more reversible.

Pick one process, keep the scope narrow, and set a quality threshold before the test starts. If automation saves time but creates more exceptions, customer confusion or manual review, stop. If it improves throughput without degrading quality, you have evidence to scale.

The important part is that the pilot answers an operating question. “Can we automate this?” is weaker than “Can we handle twice the volume at the same quality without adding another handoff?”

Add failure cost to the decision

One more filter matters: What happens when the workflow fails?

A missed social post is not the same as a missed patient message, payroll error or customer escalation. The higher the consequence, the more explicit the human review, exception path and accountable owner should be. Automation should be scaled according to failure cost, not only according to how repetitive the task looks.

The next hire may still be a person

This argument is not anti-hiring. It is anti-lazy diagnosis.

Some companies are understaffed. Some need deeper expertise. Some need more relationship capacity because the founder is still carrying every important customer or partner conversation.

But headcount should buy human capability, not permanently subsidize administrative friction that better workflow design could remove.

Try this before your next requisition

Take the proposed role and list the recurring tasks. Label each one J, R, P or C for judgment, relationship, repetition or coordination.

Then ask three questions:

• Which repetitive tasks could be automated safely?
• Which coordination tasks should disappear if the workflow were redesigned?
• What is left that genuinely requires a person with this level of skill and salary?

That final list is the job you actually need to hire for. The rest is an operating-model decision.

I think of this as capacity-first growth: Diagnose the constraint before buying the solution. The same principle applies whether the eventual answer is a person, a process change or an AI-enabled workflow.

Before you approve the next hire, write down the 10 things you expect that person to do.

Not the job title. Not the qualifications. The actual recurring work.

Now separate the list into four categories: judgment, relationship, repetition and coordination.