An AI teammate costs its usage charges each month plus the owner's time for setup and review, and the usage part rises with the work you assign. Although the monthly figure is less fixed than a salary, it becomes reasonably predictable once an owner measures real activity over several consecutive weeks rather than relying on initial assumptions.
What makes up the monthly bill for an AI teammate?
It helps to separate the bill into two distinct columns. The first column covers what the underlying software and the language models cost to operate, which is the category most owners mean when they say "usage." Every message the teammate reads, every draft it composes, and every record it retrieves consumes a measurable quantity of that usage.
The second column captures human effort. Someone must describe the business accurately, decide which systems the teammate may touch, examine its early drafts with care, and audit its performance on a regular schedule. Because that effort never appears on a usage statement, owners tend to overlook it, yet it represents a genuine expense that belongs in any honest calculation.
Many owners consider only the first column. They set the usage figure beside the wage of a part-time employee, conclude that the teammate is inexpensive, and move forward without accounting for supervision.
That comparison omits the review hours and it omits the costly errors that occur when no one examines a message before it reaches a customer. A rigorous estimate therefore includes both columns from the outset, even though the second column is harder to quantify.
Usage charges are identified before you commit, which allows you to understand the economic drivers of your deployment before you agree to any arrangement. That transparency offers a sounder foundation for planning than a general assurance that the teammate is affordable, because it lets you connect each category of work to a specific expense.
Which kinds of work push the usage charges up?
Work varies considerably in its consumption of usage. A brief reply confirming your Saturday hours represents a small, contained exchange. A request that directs the teammate to read a lengthy email chain, consult a calendar, compare two competing quotes, and produce a consolidated summary represents a far larger undertaking.
The second request may require many discrete steps, and each step consumes usage in its own right. The teammate may also re-read the same thread several times as it assembles its response, which compounds the total quantity consumed.
Four categories of work tend to exert the greatest influence on the bill:
- Lengthy threads and attachments, because the teammate must process more text before it can formulate a response.
- Multi-step tasks such as booking, verifying availability, and confirming details in strict sequence.
- Concentrated message volume during peak periods, when a large number of customers write within the same window.
- Follow-up exchanges and rechecks, in which a single task loops back for an additional review pass.
Simple, repetitive work sits at the opposite end of the spectrum. A confirmation message for a booked visit, a reminder regarding a Tuesday appointment, or a standard answer describing your service area requires comparatively little processing.
These tasks consume modest usage and tend to be the most predictable components of the total. A business that begins with lightweight work can observe its own patterns carefully before it delegates heavier responsibilities to the teammate.
Does the model you choose change the monthly cost?
Yes, and the effect is substantial. Different language models carry different operating costs, and they vary in their capacity to handle subtle or ambiguous requests. A lighter model typically manages a routine reminder competently and at lower expense.
A heavier model usually serves better when a single message combines a complaint, a refund question, and a request to modify an existing booking. Routing every task to the most capable model is the most expensive way to operate a teammate, whereas routing every task to the least capable model introduces the greatest risk of error.
The productive question therefore concerns which category of work requires which model. Evaluate the teammate against your own typical messages before reaching a decision, and document the results.
Retain routine work on the lighter option, and reserve the more capable option for interactions in which an incorrect answer could cost you a customer. For a plain explanation of how a teammate selects among models, consult what model routing means for an AI teammate.
A less expensive model does not automatically represent the better purchase. When a cheaper model produces an inaccurate response that a person must subsequently correct, the labor required for that correction may exceed whatever usage the cheaper option saved. Evaluate each model according to the total cost of obtaining a correct result, rather than according to the usage line alone.
How can a small business estimate its own monthly usage?
Begin with a deliberate, written log. Over one representative week, record the number of messages your business receives, the tasks you would prefer the teammate to handle, and the number of steps each task demands. Group those figures by category so that patterns become visible.
Multiply the weekly totals by approximately four to produce a monthly projection. Then adjust that projection for your seasonal workload, because a lawn care company's spring and a tax preparer's April differ dramatically from a quiet winter week.
Carry the estimate forward for one complete month before committing to a fixed budget. A solitary week seldom reveals the follow-up messages, the complaints that require three replies to resolve, or the surge of activity that follows a storm.
Once you possess a genuine month of observations, compare it against your projection and revise the work you assign accordingly. In most cases you can influence the total by adjusting which tasks the teammate undertakes and which model handles each one.
Maintain a companion note recording the tasks you were relieved to delegate and those you wish you had retained. That record informs decisions when you expand the teammate's responsibilities later. It also highlights tasks that appear inexpensive on paper yet generate disproportionate follow-up work, which is precisely the kind of hidden expense that a careful estimate should expose.
How do you keep the monthly bill predictable?
Predictability originates from boundaries that you establish deliberately rather than from hope. Decide in advance which categories of work the teammate may accept during a given month and which categories remain with a person. Light, repeatable tasks can run without interruption. Long, open-ended assignments should carry a defined stopping point, such as a summary of a single thread rather than an unbounded review of every conversation from the past year.
During the first month, review your usage weekly, and thereafter review it monthly. Identify the category of work that expanded most quickly. Frequently a single new habit produces the surprise, such as customers beginning to attach photographs to every request, or a staff member asking the teammate to rewrite the same reply five times in succession.
Once you identify that habit, decide whether it justifies the expense or whether a simpler process would achieve the same result at lower cost.
Establish a provisional ceiling for each month and monitor performance against it. A ceiling does not impose a limit on quality. It prompts a deliberate question about whether the additional work is generating adequate value. If a particular task repeatedly exceeds its allocated share but saves meaningful hours, retain it and adjust other categories. If it exceeds the allocation while delivering little measurable benefit, return that work to your team.
What does a monthly cost look like for a local service business?
Consider a hypothetical lawn care company in Auburn. This illustration describes a plausible scenario and does not represent an actual customer result. The owner sends a short confirmation after every booked visit, responds to the same questions about service areas, and asks the teammate to draft a reply whenever a customer reports a missed mowing day.
Most of that activity is light. The teammate issues a standard confirmation, checks the service area, and drafts a brief, courteous apology that the owner reviews before sending.
Now introduce the spring season. During a busy April, customers across Fort Wayne and Northeast Indiana inquire about recurring cuts, aeration, and spring cleanup within the same week. Their messages grow longer because they describe their yards in considerable detail, which increases the material the teammate must process.
The owner also asks the teammate to compare two quotes for a larger project and to prepare a summary for a crew lead. Each of those assignments consumes more usage than a confirmation text, so the monthly bill climbs even though the customer count has not changed significantly.
The lesson is not that spring is unaffordable. The lesson is that the owner should plan explicitly for the season. A business that understands its busy weeks can decide in advance which demanding tasks to keep with the teammate and which to leave with a person. That planning converts an unpredictable invoice into a budget the owner can explain confidently to a business partner or an accountant.
Is your review time part of the cost?
It is, and it is the cost owners most frequently forget. An AI teammate that drafts replies, books appointments, or answers questions still requires a human to verify its work. During the early period, read its drafts closely and annotate any errors. Over time, you may shift toward periodic spot checks, yet a regular review remains essential, particularly for matters involving money, refunds, and commitments about timing.
Schedule your review deliberately. Select a fixed interval each week, examine a representative sample of the messages the teammate sent, and record each reply that required correction. When corrections accumulate around a particular kind of task, return that task to a person or revise the instructions you give the teammate. This discipline keeps the total cost controlled without surrendering oversight.
Review also guards against expensive mistakes. An incorrect price, an unintended promise, or a message delivered to the wrong customer can impose costs that far exceed a month of usage. The review is therefore not additional labor layered on top of the teammate. It forms part of what makes the teammate responsible to operate, and its cost belongs in the monthly figure.
Where does an AI teammate not save money?
An AI teammate does not reduce costs uniformly across every task. It produces its greatest savings on work that is repeated, precisely defined, and straightforward to verify. Confirmations, reminders, intake questions, and first-draft replies fit that profile well.
Work that requires fresh judgment on each occasion does not. A complicated dispute with an extensive history, a negotiation over a large contract, or a decision that alters your pricing structure typically still requires your direct involvement.
Be candid with yourself about where that boundary lies. If you assign the teammate a task that you would ultimately have to redo yourself, you pay for the usage and then pay again in labor. A sound approach is to begin with the tasks you already perform identically each time, then expand the teammate's scope only after its results remain reliable for several consecutive weeks.
A tradeoff also exists between responsiveness and accuracy. A teammate can answer at any hour of the day, and many customers value that availability highly. Nevertheless, a rapid incorrect answer does more damage than a slower correct one. Configure limits so that any matter involving money or binding commitments waits for a qualified person to review it.
What should you ask before you commit to a monthly plan?
Prepare five questions before agreeing to any monthly arrangement. First, how are usage charges billed, and are those charges identified before you commit? Second, which kinds of work count toward usage, and how do long conversation threads affect the total?
Third, can you examine usage by type of task, so that you learn which activities carry the highest cost? Fourth, what access level will the teammate receive, and may you begin with read-only or draft-only permissions before authorizing actions? Fifth, what happens to your data and your work if you later decide to discontinue the service?
Record the answers in writing. A precise response to each question tells you how the monthly cost will likely behave over time. A vague response deserves suspicion, regardless of how confident the sales presentation sounds.
Hey Button runs on OpenClaw, and Hermes is part of the underlying runtime. Those names matter because they identify the software that performs the work behind your teammate. To begin, start your setup at the Hey Button start page, and bring a written list of your typical messages and recurring tasks. Those real examples will produce a far more accurate estimate than any generalized figure.
Frequently Asked Questions
- Can I cap what I spend each month?
- A ceiling you set yourself works better than a hope. Decide which categories of work the teammate may take, review usage weekly during the first month, and treat any overrun as a prompt to review that work rather than as a surprise.
- What is a sign a task is too expensive for its value?
- Watch for work that keeps looping back. If a task needs several rechecks, repeated replies, or a human rewrite every time, its usage buys little value. Move that work back to your team or simplify the request, then compare the cost again next month.
- Should I start with full access to my systems?
- Usually not. Starting with read-only or draft-only access lets you see the teammate's output before it can act. Widen its permissions only after several weeks of reliable results, and recheck the monthly cost each time you widen them.
- How do I explain the monthly cost to a partner or accountant?
- Show two columns: usage charges and the owner's hours for setup and review. Attach one month of measured activity and a list of the tasks the teammate handles. A partner can then see which tasks drive the number and decide whether they are worth it.
- Is an AI teammate worth it if I only send a few messages a week?
- It depends on the work, not the message count alone. A few messages that each require a long, multi-step task can cost more than dozens of short confirmations. Estimate by task type, then compare the monthly total with the hours the work would take you.
Hey Button
Want a teammate that handles this for you?
Answer a few questions about your business and Hey Button will show you what an AI teammate would take off your plate, with you approving anything that goes out.




