Look at what your office does all day. The estimator reads an RFQ and a drawing, then writes a quote. The quality lead reads a mill cert or a pressure test record, then writes a CoC. Customer service reads an email, looks up an order, and writes a reply.
Take away the part names and it is the same job each time. People read files, write files, and check files, and we think that is the work AI does best today.
The AI can make the first draft from your files, and your people check it, decide, and sign. We call that draft a first pass.
That helps you serve customers better. The quote goes back the same day, the cert packet is right the first time, and your team answers "where is my order" in minutes.
In many offices, people type the same data three times, a document sits where no one can find it, and the answer lives in one person's head. Habits like those slow down every quote and cert packet.
Why custom software fits how you work now
AI now writes software. Coding agents in 2026 work in sessions that run for hours, writing and testing code with a person reviewing the result. Your ERP, your quoting tool, and your shop floor screens are all software. When software is cheap to build, a company no longer has to bend its work to fit a tool someone else designed. The tool can be built to fit how your operation runs today and where it is going.
That moves the hard part. Connecting AI to your files is no longer difficult. Getting the real work out of people's heads and off paper still is, because your team has to name each step in words a system can follow.
Go deeperSoftware Got Cheap, Clarity Didn't: How to Define a Job Before Any AI Touches ItWhat clarity means in practice, and why we will not deploy AI without it.Cut the steps nobody needs before you automate
Clean up the work before you build screens for it.
Over the years people come and go, and each one leaves something behind: a new column in the spreadsheet, a new tab, a report someone asked for once. Ten years later, the team scrolls past forty columns every morning, and nobody can say why half of them are there. The only reason some of that work exists is that somebody left and nobody cleaned up after them.
If you point AI at that forty-column spreadsheet as it is, you just get a faster way to do work that should not exist.
Before we build anything, we sit with the people who do the job and ask why each step is there. Some steps matter a great deal, and some can go. Only then do we decide what the AI takes on.
Owners and department heads have to back the change, because good software cannot force a new way to work. We ask leaders for that promise before we start. Without it, we would only build a faster version of the old mess.
Go deeperDon't Automate What Shouldn't Exist: Find the Steps Nobody Needs FirstHow to find the steps that outlived the person who asked for them.Review screens your team will want to use
Your team has to check each draft with care, and a screen that is hard to use breaks that habit fast.
Ask the people in your office how they feel about the ERP. You will hear "clunky," "not intuitive," and "I don't want to look at it all day." If checking the AI's work is painful, people will skip the check or skip the tool.
We treat design as part of the job. To us, design is how the person using the software feels at 3 p.m. on a Friday, when the pile is still high. How it looks comes second.
We build products that show a link to the source file every time the AI works from one. The reviewer hovers over the link, and a large preview of the source document appears on the screen. They can check the value against the source at a glance, and no one has to open a second program.
It sounds minor until you multiply that hover preview by every value on every document, every day. Fast checks get done, and slow checks teach people to skip.
Go deeperThe Friday Afternoon Test: How to Judge Software Your Team Will Use to Check AIA five-point test for any software your team would use to check AI work.How the first pass works while you are away
Most people picture AI as a chat window you type into. We build AI systems that work in the background, and we call that way of working AI everywhere.
The AI watches only the queues you allow it to see. It does the first pass on new work, then asks a named person to check the draft.
Here is an example. A customer sends an urgent email while you are at lunch, asking for certs before a truck leaves. The AI reads the email, pulls the order and its mill certs, and builds a draft cert packet for you.
It sends a short note to your phone. When you get back to your desk, the draft is waiting for your review, with the context for the job already gathered.
In that lunch-hour example, the AI sent nothing to the customer. It has its own login and can see only the systems its job needs. An audit log records each step the AI took, with the time of the step and the person in charge.
The speed comes from the first pass, and the safety comes from a person signing.
Go deeperThe Colleague Who Never Clocks Out: What Always-On AI Does While You're AwayWhat an always-on AI may do alone, and what always waits for a person.The real return is headspace for your best people
Most people ask about ROI first. We think the first return is headspace, which is harder to put on a spreadsheet.
Headspace is the attention your team gets back. Your team carries a huge amount in their heads: which customer wants their certs worded a certain way, which order shipped in two partials, which sign order is still waiting on a proof approval.
On a normal week they cope. When one big customer picks up, it is all hands on deck, and everyone is drowning.
We do not promise that AI solves every problem in your operation. We do promise two things.
First, the work lives in one record, where your team can find each item, identify it, and trace it. Second, each department gets an always-on AI that knows its context and does the first pass before anyone has to go digging.
Leaders gain a view they rarely have today. They can see which work is still in progress and which is done.
Go deeperHeadspace Is AI's First Return: How to See It Before You Count Hours SavedHow to see the return in a normal week, before anyone counts hours.What comes next: humanoid robots
Our view: humanoid robots are coming to real factories, and the companies that benefit first will be the ones whose work is already digital, traceable, and clearly owned. A robot, like the AI we build, can only work from a record that exists. The foundations you lay now for your paperwork are the same ones the next wave of automation will stand on.
Go deeperHumanoid Robots in 2026: Why Your Paperwork Decides When They Can Help YouWhere humanoid robots are in 2026, and what a manufacturer should do now.The fear, named and answered
"AI is coming for my people's jobs."
AI takes the retyping first, along with the searching and the copying between systems. Nobody was hired to love that part of the job.
The person who did that work still owns the result. They review it, catch what is wrong, and handle the customer who needs a real answer.
About 1 in 4 manufacturing workers is 55 or older. Your experienced people can spend their time on judgment and on training the next generation, instead of spending the day on data entry.
What to do this week
- Pick one office role. Write down each file that person reads in a normal day and each document they write, then mark the ones copied from somewhere else.
- Open your biggest shared spreadsheet. Ask the person who uses it most which columns they cannot explain. Those columns are candidates to delete before anyone automates them.
- Wait for an urgent customer email to arrive while the person who owns the job is away. Time how long it takes before someone starts work on it. That gap is what an always-on first pass closes.
- Ask your leadership team one question. Are you willing to change how this job is done, or only to make the current way faster? Only the first answer leads anywhere.
Questions people ask
Why hasn't AI taken off in manufacturing yet?
Many offices run on paper and spreadsheets, and key knowledge lives in a few people's heads, so AI has little it can read reliably. 65% of manufacturers told the NAM they lack the right data for AI. Companies that put their records into computer files first get the most from it.
Is AI going to replace manufacturing office jobs?
AI takes the retyping, searching, and copying between systems first. A named person still reviews every draft and signs anything that leaves the building, so the job shifts toward judgment and customers.
What is "AI everywhere"?
AI everywhere is our term for an AI that works in the background. It watches the queues it is allowed to see, does the first pass on new work, and messages a person when something needs a decision. It sends nothing on its own.
Terms in this piece
- First pass
- the draft the AI prepares, such as a cert packet or a quote, before a person checks it.
- Headspace
- the attention your team gets back when they no longer have to remember and hunt for everything themselves.
- Coding agent
- an AI that writes and tests software on its own for long stretches, with a person reviewing the result.
- Audit log
- a record of every step the AI took, when, and under whose authority.
- AI everywhere
- an always-on AI that does the first pass in the background and reaches out to a person when a decision is needed.



