When Should You Automate vs Hire?

This question comes up in every boardroom, every team meeting, every strategy session. And the answer is almost never "automate everything" or "hire everything." But the framing is usually all wrong.

People ask: "Should we build AI to do this, or hire someone?" Like it's a binary choice. It's not. The real question is more complicated, involves more variables, and requires you to think like a business operator, not a technologist.

Here's how I actually think about it.

First: Is This Problem Expensive To Solve Right Now?

Let's get concrete. Let's say you have a problem that costs you $500K a year in labor. Either you're paying someone to do repetitive work, or you're paying smart people to do work that could be automated, or you're just not doing the work at all because it's too expensive.

Now you have three real options:

• Hire someone. Salary, benefits, hardware, management time. Let's call it $80K all-in. That solves your problem at $80K a year.
• Automate it. Build or buy software. Upfront cost, maintenance, edge cases that break. Let's say $150K to build and two years to get it stable. That's $75K a year over time, plus risk that it doesn't work as well as you think.
• Reorganize. Sometimes the real answer is "don't solve this problem the way we're currently solving it." Maybe you don't need the work done. Maybe you need it done differently. Maybe you need it done less often.

If automation costs $150K and a person costs $80K, and the problem is actually solvable by a person, you probably hire the person. Especially if you're uncertain about automation. Certainty has value.

I've seen too many teams pour $300K into automation to save $80K a year in labor. That's a bad trade.

Second: What Kind Of Problem Is This, Really?

Not all problems are created equal. Some are automation-friendly. Some aren't.

Good automation problems:

• Repetitive with clear rules. "Sort these by category." "Extract data from this format." "Flag items that meet X criteria."
• High-volume, low-stakes. You do this 1,000 times a week and if you get one wrong, it's not a disaster.
• Stable. The rules don't change every month. The input format doesn't shift. The success criteria are clear.
• Clear success metric. You can measure if it's working.

Bad automation problems:

• Requires judgment. "Is this customer important?" "Does this proposal make sense?" "Should we take this risk?" These require context, intuition, understanding nuance. AI can help, but humans are still cheaper and better.
• Ambiguous requirements. "Make this better." "Improve customer experience." "Optimize for efficiency." If you can't define what success looks like, automation will disappoint you.
• Constantly changing. If your business process changes every quarter, automation is expensive because you're always rewriting the logic.
• Mission-critical with low tolerance for failure. If this process breaks and costs you $50K per outage, automation better be bulletproof. Humans are more flexible about errors.

Here's something that makes me frustrated about tech hype: people assume AI is the answer to judgment problems. It's not. AI is very good at finding patterns in data and doing things consistently. It's not good at novel situations, nuance, or saying "I don't know, ask a human." You can build that on top of AI, but at that point you've added a human anyway.

Third: What's The Hidden Cost?

This is where most automation fails. Calculating the direct cost is easy. Nobody calculates the hidden cost.

Hidden costs of hiring:

• Management overhead. Someone has to manage them, which takes time.
• Onboarding. Three months before they're productive.
• Turnover risk. Great people leave. Then you've got a gap.
• Seasonal variation. If the work is bursty, you're paying someone for slow periods.

Hidden costs of automation:

• Maintenance. Someone has to maintain it. That person better know what they're doing.
• Edge cases. The system works 95% of the time. What happens the other 5%? You need escalation paths, error handling, monitoring. That's person-work you didn't budget for.
• False confidence. Your system works fine for six months. Then the world changes slightly, and suddenly it's making bad decisions and nobody notices for three weeks.
• Integration friction. The automation needs to talk to your other systems. That's usually more work than you think.
• Training and change management. People need to learn to use the system. People need to learn to trust it. People need to know when to override it. That's your hidden HR budget.

I worked with a team that automated their expense report processing. They saved maybe $50K a year in manual processing. But they spent $200K building it, $30K a year maintaining it, and they had to hire someone to handle the 5% of reports the system couldn't parse. Net cost? Probably negative. And that's with everything going well.

If the automation breaks, or the requirements change, or integration takes longer than expected? Now you're spending money on automation AND you still need the person. That's the hidden cost.

My Framework For Deciding

Here's how I actually make this call:

Step 1: Is this problem expensive right now? (More than $150K a year.) If no, don't solve it at all.

Step 2: Is this problem stable? (Requirements haven't changed in a year, won't change for the next year.) If no, hire flexibility instead of building rigidity.

Step 3: Does this require judgment? (A human needs to understand context or make a call.) If yes, automation is 30% of the solution, and you need the human anyway.

Step 4: What's the failure cost? (If the system breaks, how much does it hurt?) High failure cost = hire someone with accountability. Low failure cost = automate and monitor.

Step 5: What's the certainty level? (Do you know this will work?) Uncertain = hire someone while you're figuring it out. Certain = automate.

Step 6: Cost comparison. Build a real spreadsheet. Include everything. Compare the total cost of ownership over three years. Hire the cheaper option, accounting for risk.

The honest answer is usually: hire someone for the next year while you're figuring out if automation makes sense. Once you understand the problem deeply, you can build something that actually works.

The Thing Nobody Says

The real answer to "automate vs hire" is usually "do both, sequentially." Hire someone smart. Have them do the work while documenting exactly how they do it. Once you understand the process deeply, automate the parts that are actually repeatable. Keep the person for judgment calls and exceptions. You've now got efficiency AND reliability.

This costs more upfront. But it's cheaper over time, and it actually works.