What Tasks Should You Not Hand to an AI Teammate Yet?

Keep money decisions, legal commitments, sensitive customer situations, decisions about people, and anything irreversible with a person for now. An AI teammate is best suited to repetitive, low-risk work that follows rules you can write down in advance. Its usefulness drops as a task depends more on judgment, on relationships, or on the exact wording of a commitment.

Which tasks should stay with a person for now?

The short answer is any task where a wrong move could cost you money, damage your reputation, or weaken your legal position. A mistake in an appointment reminder is irritating, yet it is easy to correct and rarely expensive.

A mistake in a refund, a warranty promise, or a note about a customer's injury can be costly, and some of those errors cannot be retracted once the message has left your business.

Sorting tasks into three groups makes the decision considerably clearer, because the first group contains tasks with a definite answer that the teammate can locate in material you have approved. The second group contains tasks where the teammate can gather facts and prepare a draft, but a person must make the final call.

The third group contains tasks that should not leave a person's hands for the time being. Most of the work that causes trouble for small businesses falls into the second and third groups, while most of the time savings come from the first.

Hey Button sets up each AI teammate so that the owner decides where these boundaries sit, rather than accepting a generic default. That decision is made during the setup conversation, and it can be revised later, once you have observed how the teammate behaves in your particular business.

Money decisions belong with a person because each one is a trade-off

Money decisions include refunds, discounts, price exceptions, deposits, and changes to a quote after it has already been sent. Each one involves a trade-off that depends on the customer, the season, and how heavily your team is loaded that week. A teammate can explain your published policy accurately and clearly, and it can recognize when a customer is requesting something outside that policy.

Whether to grant the exception is a judgment call, and that call belongs to you. A quote that looks reasonable on its own can still conceal a larger cost.

Suppose a lawn care company promises a discount in order to win one customer. That same promise may surface in five other conversations before the month ends, and then the business has a pricing pattern it never intended to create.

Consistency is essential in pricing, and a person who understands the whole business is better positioned to maintain that consistency than a teammate reading one conversation at a time.

Under draft-and-approve, the teammate can prepare the message, summarize the request, and recommend a response, and you decide before anything reaches the customer. That arrangement preserves the speed of a ready draft without surrendering control over what a customer ultimately reads.

Read-only access is more conservative still, because the teammate only observes and reports what it sees. Our fuller explanation of the three levels appears in Read-Only, Draft-and-Approve, or Full Access: Pick the Right Level.

The downside is real and worth stating plainly, since every money decision that stays with you adds a step, and a slow reply can lose a customer who wanted an immediate answer. Accept that cost for money decisions, because a fast but incorrect answer about money is usually far more expensive than a slower, correct one.

Why are legal promises and guarantees off the list?

Words such as guaranteed, always, never, warranty, and insured carry real legal weight. A teammate asked about coverage, licensing, or insurance may sound entirely confident while being wrong about your specific policy. A customer who relies on that answer may later hold your business to it, even though no person on your team ever said it directly.

The safe approach is a short list of approved answers that you wrote and reviewed yourself. Keep the list small enough that someone can read it in roughly five minutes.

Each answer should state what is true for your business today, without adding promises that nobody has authorized. When a customer asks something the list does not cover, the teammate should explain that a person will follow up, then flag the question for your review.

This is also why Hey Button describes its teammates as configured for specific tasks. An AI teammate supports your judgment, but it does not replace that judgment, and it does not assume responsibility for business decisions. A business that understands this limitation can rely on the teammate for many tasks without asking it to speak for the business on questions of liability.

There is a trade-off in maintaining a long approved list, since the longer it grows, the harder it becomes to keep current, and outdated wording creates its own legal exposure. Review the list whenever your prices, service areas, licensing, or insurance coverage changes, and schedule a formal review at least once each quarter.

How should you handle upset customers and sensitive situations?

An upset customer often needs a human voice more than a fast answer, so a complaint about a missed appointment, damage to a home, a safety concern, a message describing a medical or family emergency, or any mention of an attorney should reach a person right away.

Write these triggers into your approval rules so the teammate knows to stop, acknowledge the message in a neutral way, and hand the conversation over.

The teammate can still contribute usefully in these moments, and it can collect the relevant facts, record the time the message arrived, and place the most urgent items at the top of your queue.

It should not attempt to soothe an angry customer with reassurances it has no authority to make. A reply that sounds caring but promises an unapproved remedy can easily make a bad situation worse.

The downside is that someone on your team must monitor the queue promptly, because a rapid handoff delivers little value if nobody reads it for several hours. Before setup, decide who watches the queue and when they check it, so that each handoff actually reaches a person in time to matter.

When this does not apply: a business that rarely receives complaints may need only a handful of triggers. Keep the emergency trigger in every configuration regardless of volume, because a rare emergency is precisely the situation in which a missed message does the most harm.

Decisions about people stay with you, even when a teammate can help

Decisions about people rank among the most consequential choices a small business makes. They include hiring, dismissal, pay, staff scheduling, and performance concerns, and these decisions involve confidential information, local employment regulations, and interpersonal relationships that an AI teammate cannot observe from the outside.

A teammate might usefully summarize an applicant's notes, yet the decision itself remains yours to make, and no summary should substitute for your own judgment.

A practical rule is to keep any task that names a specific employee, applicant, or contractor outside the teammate's default work. You can still ask for help organizing general hiring steps, such as drafting a neutral job posting for your review.

Anything that touches a named person's pay, standing, or future belongs with you, and where the law requires it, it also belongs with the professionals who understand your local employment obligations.

The trade-off is that you will personally write some material a teammate could have drafted. For most owners, that is a sound exchange, because a poorly worded note to an employee is difficult to retract, and a few extra minutes of careful writing is a modest price to pay for getting it right.

Which tasks are good to hand over first?

Begin with work that is repetitive, low-risk, and already governed by a clear rule. Strong first candidates include answering questions about published hours, confirming appointments already on the calendar, sorting incoming quote requests, and drafting a follow-up when a lead has been quiet for a set number of days.

These tasks share three characteristics: the correct answer is usually written down somewhere, a wrong answer is comparatively easy to correct, and the volume is high enough that small time savings accumulate over a season.

Consider a lawn care company in Auburn that receives most of its inquiries by text message and web form. The owner might begin by handing over quote requests alone, and the teammate sorts each inquiry, identifies the details that are missing, and drafts a reply for approval.

Once those drafts prove dependable, the owner can add follow-up drafts for quotes that have stalled. This scenario is an illustration rather than a case study, but the sequence of steps is one that most small businesses find easiest to trust.

Use draft-and-approve for anything that reaches a customer directly, including the simplest tasks. Full access becomes reasonable later, for narrow tasks you have reviewed over several weeks, and only inside the approval rules you yourself wrote.

The same logic applies to a contractor in Fort Wayne who wants help confirming job-site appointments. Begin with confirmations you review, then relax the rule once those confirmations match the way your crew actually works in the field.

Be candid about the limits here as well, because drafting slows certain tasks down. Someone must review each draft before it is sent, and for a busy owner that review can feel like a chore. It is still usually faster than composing every reply from scratch, and it catches errors before they ever reach a customer.

How do you find the edges of your own rules?

Write the rules before the teammate touches real work, and list what it may read, what it may draft, what it may do without asking, and what always returns to you first. A rule that says "answer hours questions" is too loose if your hours change during holidays.

A rule that says "never discuss pricing" is too tight if the teammate could safely quote your published rates. Effective rules name the exact material the teammate may use and the precise situations that require a person.

Then run the teammate against a week of real examples and read every draft carefully. Each draft you would have changed reveals where a rule is too loose or too tight.

Keep a short record of these adjustments, since after a month those notes usually show a clear pattern, such as one category of request that always needs a human, or one kind of reply that you approve without any edits at all.

This takes time, and owners who skip the review often discover a flawed rule only after a customer has. A focused review during the first few weeks costs far less than a correction sent to a customer who was promised something you never intended.

Expect to tighten a few rules during the first month, and expect to loosen others as well, once the drafts consistently meet your standards for accuracy and tone.

Does the list look the same for every business?

No, because a business that handles many small, recurring requests can loosen its rules sooner than a business with fewer, larger jobs. A law office, a medical clinic, or a contractor who regularly manages disputes will keep more tasks with a person for longer.

Size matters less than risk, since two people running a landscaping company may hand off quote sorting early, while a larger firm with complicated contracts keeps most client communication with its staff.

The same logic holds across Northeast Indiana, from a single-truck operation in a small town to a multi-crew shop in Fort Wayne. The question is always what a wrong answer would cost, and how quickly a person would notice it.

The list in this article is a starting point rather than a rule that applies everywhere. Your own history of mistakes, complaints, and near misses should shape it far more than general advice can. If your business experiences a recurring type of problem, add that category to the tasks that remain with a person, even when it does not appear here.

Hey Button settles these boundaries during the setup conversation, and the steps are laid out on the how it works page. The objective is a teammate that handles the work you feel comfortable delegating, and that returns everything else to you clearly and promptly.

If you want help drawing your own list, start with the Hey Button questionnaire. It asks about your business in plain language, and your answers become the first draft of your approval rules.

Bring two lists to that conversation: the tasks you dread repeating every week, and the tasks you would never hand to anyone else. Those two lists show the teammate where to start and where to stop.