You probably met AI in a chat window. You typed a question, got a decent paragraph back, and filed it under "fancy search box." That was fair. Chat apps now search the web and read files you upload, but they still do nothing for the stack of certificate requests in your quality inbox, because they cannot open your orders.

A chatbot is something you talk to, back and forth. A new hire is someone with a role, a purpose, and a job to do whether or not you are talking to them. The AI we build for manufacturers is the second kind, and the easiest way to understand it is to onboard it like one.

25.2%of US manufacturing workers are 55 or olderBLS CPS Table 18b
1.9Mmanufacturing jobs that may go unfilled, 2024 to 2033Manufacturing Institute and Deloitte
51%of manufacturers already use AI in operationsNAM and Manufacturing Leadership Council

How it works on the floor

Take a company with a shared team inbox, where customers email requests for Certificates of Conformance and Certificates of Origin.

A team member opens each email, finds the order and the parts, pulls the details, and types the certificate. Each request is a small job, but together they eat hours of someone's week.

Now put an agent in that inbox. It reads each request, matches it to the order, and drafts the certificate. A person checks the draft and sends it.

Your team stops typing certificates and keeps every decision about them.

The role can grow wider. At a building-products company, for example, one agent can follow each order from the mill certificates through receiving, production, the yard, and dispatch, then pass each order's record into the next quote.

Most of the time, it keeps the record straight. When a mismatch or a delay needs a decision, or when only a person can answer the question, it calls in the right person.

AI does the first pass. People verify, decide, and sign.

Onboard it like any new hire

A new coordinator starts the first week with a short list. They need an email address, a computer, and logins for the systems they will use.

Your team trains them on how your shop does the job and tells them what they are responsible for. A manager checks their work until they have proven themselves.

An AI employee needs the same list, and that is how we set one up.

The inbox can be a new address that customers or colleagues write to, or the AI can connect to a person's existing inbox and read along.

The AI can read and sort the mail, but it cannot send any. Every reply stays a draft until a person checks it and hits send.

Your people stop being glued to their email. The AI reads the requests that land after hours and over the weekend, matches them to orders, and drafts replies, so Monday morning begins with a queue of first passes ready for review.

The computer is a virtual machine on a server where the AI opens files, runs programs, and keeps its work in its own folders without touching anyone else's desk.

The logins come one system at a time, and reading comes before writing. For a certificate job, the AI might read the shared quality inbox and the order history and write only to a drafts folder.

The training starts with your standard operating procedures plus 50 to 200 real past examples of the job done well. Each example pairs the customer's email with the certificate that went out and notes anything unusual about the case.

The model underneath does not learn on its own. Those examples go into files it reads every time it works, and we add new ones as it meets new cases.

The job description gives the AI one role and one purpose, and spells out what good work looks like.

The manager is a named person on your team who reviews the AI's work.

It will make mistakes, so we build the checks in

AI will make mistakes. A model writes by predicting what should come next. When the order record is in front of it, the prediction matches the record. When the record is missing, it still predicts, and a made-up heat number looks exactly as confident as a real one.

People make mistakes too, and your operation already has ways to catch them. We build the same kind of checks for the AI.

First, we map the decisions with your team before anything goes live. We write the job down step by step and mark each place where a person has to verify the work. We call that written map the decision map.

It shows what the AI can do on its own (sort, match, and draft) and where it must stop and ask a person.

Second, the AI flags what it does not know. Say a heat number on the mill cert does not match the receiving log, or a request matches no open order. The AI does not guess. It stops, flags the item, and brings in the right person.

Third, the AI keeps an activity feed. A reviewer can see every step in order, with the time of each step and the source the AI used. Nothing is hidden inside the software.

Test the AI before it goes live

Before the AI touches live work, it has to pass a test set of 50 past requests with answers your team already knows. Your team grades the work field by field against a bar you wrote down first, and that bar allows zero critical errors.

Then the AI spends two weeks in shadow mode, drafting on live work while nobody sends what it writes. Only after that stretch can it go live, and a person still approves every output.

Reviewing the AI's work is a skill, and most manufacturers have not trained anyone for it yet.

AI training manufacturers give frontline workers percent of manufacturers, select all that apply
None41.43%
Introductory or awareness40.00%
Data or interpreting AI outputs25.24%
AI for quality, production, maintenance21.43%
Operating AI-enabled equipment18.57%
Advanced or role-specific13.33%

Source: NAM Manufacturers' Outlook Survey, Q2 2026.

The reviewer needs to know what a wrong draft looks like. Walk that person through the AI's real mistakes from the test. We think that hour is worth more than a general awareness course.

The deciding question is whether the job is clear

The usual questions are "Can AI do accounting?" and "Can AI do quoting?" We think that is rarely the question that decides anything.

Today's AI can do most of what a person does at a computer. It can use a web browser, read and write files, and analyze a spreadsheet far larger than anyone wants to scroll through. Voice is coming too, although we do not deploy AI on phone calls today.

We think AI still struggles with long, unclear jobs, while short and clear jobs work well when a person checks the result.

So you need to name the requests it handles, the documents it reads, the decisions it makes, and the person who checks the work.

If a company cannot get to that clarity, we will not deploy an AI system, because an AI employee with a vague job does vague work.

We think plain software solves many shop problems first. Replace a paper traveler with a digital form, keep one record for each order, and clean up the parts list.

We do not throw AI at a problem. We start with clarity about the job, then build the software and systems the AI needs. The AI uses those systems as its tools, the same way a new hire uses the ERP and the shared drive.

This approach works in departments that run on reading and writing files. Quality, sales, purchasing, customer service, and accounting all fit that pattern.

An AI employee stays on through nights and weekends. It can also work on thousands of files and data points at once, where a person has to take them one at a time.

The fears, named and answered

"My people will think they're being replaced."

Some will, at least at first. Say it out loud instead of hoping it will go away.

AI takes the work nobody wanted first, like retyping data into certificates, copying numbers between systems, and hunting for the right PDF.

The person who did that work still owns the result. They review the draft, catch problems, and handle the customer who calls with a real question.

US manufacturers had 580,000 open jobs in July 2026 (preliminary). Most shops already have work waiting for those hours.

"I'm not technical."

You don't need to be. The hardest part of an AI employee is clarity, and clarity is a management job.

You know which requests come in most often, which documents your team types by hand, and who checks what. You know your operation better than any engineer.

Someone else can handle the technical setup. You cannot hand off the job design or the choice of reviewer.

What to do this week

You do not need to buy anything for these steps.

  1. Pick one office role and write the onboarding checklist you would use for a new hire in that seat. Include the inbox, computer, logins, training, job description, and manager. That list is also the checklist for an AI employee.
  2. Open that role's inbox and tally last week's requests by type. Find the one that shows up most often, and check whether your team answers it the same way each time.
  3. Write down each step for that request and mark the ones where a person has to verify something. That marked list is the start of your decision map.
  4. Write down what should happen when the answer is not clear, such as a heat number that does not match. Those are the moments when the AI should stop and flag.
  5. Name the person who would review the drafts, and ask them which parts of the job they would happily hand over.

Questions people ask

What is an AI employee?

An AI employee is an AI agent set up like a new hire: an inbox, a virtual machine to work on, logins to specific systems, training on your SOPs and past examples, one defined job, and a named person who reviews its work. It reads and drafts inside your systems, and a person approves anything that leaves the building.

Can an AI employee send emails on its own?

In the systems we build today, no. It can read an inbox, sort requests, and draft replies, but a person sends every email. That keeps a human signature on everything that reaches a customer.

What happens when the AI is not sure?

It flags the item instead of guessing, for example when a heat number on a mill cert does not match the receiving log, and brings in the right person. Every step it takes is written to an activity feed that anyone checking its work can read.

Terms in this piece

Model
the AI "brain" a company such as Anthropic, OpenAI, or Google trains on huge amounts of text, code, and images. It predicts, and knows nothing about your business until you show it.
Agent
a model put to work. Software that takes actions, like reading an email and drafting a document, instead of only answering questions.
Chatbot
a chat window where you type a question and get an answer back. It does not act inside your systems.
Virtual machine
a computer that runs as software on a server. The AI employee works on one the way a person works at a desk.
Decision map
a written, step-by-step map of a job that marks where the AI may act and where a person must verify.
Activity feed
the step-by-step record of everything the AI did, with times and sources, that a reviewer can read.
Test set
50 past cases with known answers, set aside and used to grade the AI before it touches live work.
Shadow mode
two weeks in which the AI drafts on live work, nothing it drafts is sent, and someone compares its drafts with what people sent.