Field NotesRecognition
What work should a system do instead of a person?
A simple way to decide who should own each task: a person, a system, or an AI agent. Start with the work, not the tool.
Short answer
Work a system should do is work that repeats often, follows the same steps each time, and does little harm if a first draft is wrong. A person should keep work that needs judgment, trust, or context that isn’t written down. The quickest test: if a clear brief is enough to do the task well, a system can own it.
How do you decide whether a person or a system should do a task?
Rate each task on two things: how often it repeats in the same way, and what happens if the output is wrong. Repetition tells you whether a system can do it. Consequence tells you how much a person needs to review it.
Put the two together and every task lands in one of four places:
| The task | Who owns it | Examples |
|---|---|---|
| Repeats often, low cost if wrong | System runs it, person spot-checks | Lead follow-up, meeting summaries, reports from structured data, scheduling, replies to common questions |
| Rare, low cost if wrong | Agent drafts, person finishes | Researching a vendor, a one-off announcement, background on a prospect before a first call |
| Repeats often, high cost if wrong | Agent drafts, person approves | Proposals, financial summaries, performance reviews with recommendations |
| Rare, high cost if wrong | Person | A hard client conversation, a hiring call, pricing a complex job, next quarter’s direction |
What is the one question that tells you whether to hand a task to AI?
Ask: does doing this task well require knowing something that isn’t in the brief? If a clear brief and the right information are enough, a system can own it. If it depends on relationship history, trust, or unwritten context, keep it with a person.
Most of what owners hold on to could be written into a brief. Most of what they shouldn’t hand off can’t be. This question sorts the two in seconds.
What tasks should a small business automate first?
Start with a repeated handoff: work that has clear inputs, predictable decisions, and an outcome you can measure. Following up on new leads, moving form entries into your CRM, appointment reminders, and the weekly report are common first choices.
A handoff is the moment work passes from one person or tool to the next. It’s where leads sit, information gets retyped, and follow-up depends on someone’s memory. That makes it the best place to start.
What do you need to decide before handing work to a system?
Five things: what people should own, what systems should handle, where an AI agent can assist, where a person must approve, and how you’ll know the change helped.
- What people should own: judgment, relationships, and accountable decisions.
- What systems should handle: the steps that run the same way every time.
- Where an AI agent can assist: reading, sorting, and drafting that follows a pattern.
- Where a person must approve: anything that reaches a customer or affects money.
- How you’ll know it helped: one number to compare before and after.
Which work should always stay with a person?
Keep work with a person when it carries news that shapes a relationship, needs a judgment call, or affects strategy. An AI assistant can help you think it through, but a person makes the call.
- Any message to a frustrated or confused customer
- News, good or bad, that changes how someone feels about working with you
- Pricing a complex job
- Hiring and people decisions
- Deciding what the numbers mean and what to do next
A useful split for reporting: the system pulls the data, a person writes the recommendation.
Is tedious work always a good candidate for automation?
No. Tedious isn’t the same as automatable. Some tedious work needs context only you carry, and some work that doesn’t feel tedious, like research, summaries, and first drafts, is a strong fit for a system.
Judge the task by repetition and consequence, not by how much you dislike doing it.
Does fully automated work still need a person?
Yes. Every automated task needs a named owner who sets the quality standard and checks a sample on a schedule. High-consequence work needs a person to approve it before it goes out.
Handing work to a system changes your job from doing each instance to owning the standard. That’s the point: fewer repetitive tasks, more room for judgment.
How do you map a repeated handoff?
Pick one handoff that happens every week and write it down from start to finish: what starts it, each step, each decision, who owns it, and what “done right” looks like. Then decide which steps a system should take.
- Name the trigger: what starts the work.
- List every step in order, and the decisions inside them.
- Mark each step: person, system, or agent.
- Mark where a person approves.
- Measure it now, build it, then measure again.
For a fuller explanation of the three kinds of workers, see What is a hybrid workforce?
More questions
Is a system the same thing as AI?
No. A system can be a simple automation that follows fixed rules, an AI agent that reads and drafts, or both. Many tasks need only plain automation.
Should I pick an AI tool first?
No. Decide who should own the work first. The tool comes after you know the steps, the decisions, and the standard for done.
What if the steps aren’t written down?
Then writing them down is the first job. A system can’t follow a process that only lives in someone’s head.
Does this work for a business with only a few employees?
Yes. In a small team, the same few people answer every inquiry and send every follow-up, so moving repeated work to a system frees the most time for judgment.
How often should I revisit who owns each task?
Review it when the work changes, and on a regular schedule, such as once a quarter. Tasks can move as your process and standards get clearer.