You would not hire an office coordinator by saying "help out with the phones" and walking away. Yet many service businesses start with AI that way: turn on a tool, give it a vague instruction, and wait. A few weeks later the team has stopped trusting it, and nobody can say whether it worked.
This post is for owners and operations managers at service companies with roughly 5 to 49 employees who are planning a first AI employee. The fix is the same one you would use for a new hire: an AI employee job description. It is a short document that defines the job before anything is built, and it is the most useful hour you can spend before an AI build starts.
The Business Problem
When an AI employee has no clear job, a few things happen in real operations:
- It does part of a task, and nobody knows which part is left for a person.
- It handles easy cases well and mishandles unusual ones, because no one defined "unusual."
- Its output lands somewhere nobody checks, so work looks finished when it is not.
- When something goes wrong, no one is responsible for fixing it.
None of these are technology failures. They are the same problems you get with a new hire who was never told the job, the boss, or when to ask for help.
What This Costs the Business
A poorly defined AI employee rarely fails loudly. It fails quietly, which is more expensive:
- Lost follow-up. Leads or customer questions slip through because the AI handled half the step.
- Hidden rework. Staff double-check everything, taking back the time the AI was supposed to save.
- Owner dependency. Questions about what the AI should do flow back to the owner, adding to the mental load the project was meant to reduce.
- Lost trust. Once a team decides "the AI doesn't work," the next one is a harder sell, even if it is built well.
What Should Happen First
Before writing the job description, pick the job. A strong first job for an AI employee has three traits:
- It happens often enough to matter, usually daily or weekly.
- It follows a pattern your best employee could explain in a few minutes.
- A delay or miss in this job costs real money or customer trust.
If you cannot describe the job without saying "it depends" several times, map the workflow first. Our post on when one AI employee is enough covers how to judge scope before you build.
The AI Employee Job Description: Ten Lines
StrategixAI calls an AI employee built for one bounded job a Task Expert. Whatever you call it, the job description should fit on one page and answer these ten questions:
- Job: What one outcome is it responsible for?
- Trigger: What starts the work: an event, a schedule, or a request?
- Inputs: What information may it read?
- Tools: Which systems can it use, and which can it change?
- Outputs: What does it produce, and where does that land?
- Accountable human: Who owns its results and reviews its work?
- Limits: What is it never allowed to do?
- Escalation: When must it stop and hand off to a person?
- Quality standard: What does a good result look like?
- Evidence: What record proves the work was done?
The two lines owners skip most often are Limits and Escalation. They are also the two that protect the business. An AI employee that knows what it must not do, and when to hand off, is far easier to trust.
A Worked Example: Missed-Call Follow-Up
Here is what the outline might look like for a plumbing or HVAC company that loses calls during busy hours:
- Job: Make sure every missed call from a new or existing customer receives a follow-up the same business day.
- Trigger: A call goes unanswered during business hours.
- Inputs: Caller number, CRM contact record if one exists, and the approved list of services and service area.
- Tools: Reads the CRM. Sends text messages from the business number. Creates a follow-up task. Cannot change job records or pricing.
- Outputs: A text to the caller, a CRM note, and a task assigned to the office for any caller who needs a person.
- Accountable human: The office manager.
- Limits: Never quotes prices, promises arrival times, or diagnoses a problem.
- Escalation: Anything mentioning water actively leaking, no heat, gas, a complaint, or a billing dispute goes straight to a person.
- Quality standard: Messages are short, polite, use the company name, and match the approved wording.
- Evidence: A daily list of missed calls, the follow-up sent for each, and which ones were escalated.
Notice what the job description leaves to people: pricing, scheduling promises, diagnosis, and anything urgent or emotional. That is intentional.
Where AI or Systems Can Help
Once the job description exists, the build is mostly configuration. The AI employee works inside the CRM and phone system you already use, with the permissions written on the page. In most cases no new system is needed. A light configuration change, such as a new pipeline stage or a custom field for "AI follow-up sent," is common. A new integration may be needed if the phone system and CRM do not share data today.
When a business has several AI employees, a manager role coordinates them, checks outputs, manages handoffs, and reports to a human leader. Our post on who manages an AI team explains how that oversight works.
What Changes and What Stays the Same
What changes: a repetitive job gets done consistently, every time, with a record behind it. The office manager reviews exceptions instead of chasing every missed call.
What stays the same: the people who make judgment calls still make them. The customer still talks to your team when it matters. Your existing systems stay in place.
The job description also becomes a maintenance tool. When services, pricing rules, or staff change, update the page first, then the AI employee. Like any role, it needs ongoing review.
When On-Site Discovery Helps
A single, well-documented job like missed-call follow-up can often be defined in a working session. Discovery matters more when the job crosses departments, when nobody agrees on how the work is done today, or when several AI employees will share information. In those cases, StrategixAI may spend one, three, or five days on-site to see how work actually moves before writing any job descriptions. The goal is zero ambiguity before the AI employees are built.
Where StrategixAI Fits
StrategixAI helps service businesses define, build, and maintain AI employees and AI Teams that work alongside their staff. That starts with the work in this post: picking the first job, writing limits and escalation rules, assigning a human owner, and configuring the systems you already have. You can see the full process on our How We Work page.
Practical Next Step
Pick one job your team repeats every day and fill in the ten lines above. If you get stuck on Limits or Escalation, that is a useful signal, not a failure.
If you want help turning that page into a working AI employee, schedule a consultation to map the first AI employee or AI Team for your business.