Saturday, 7:40. A 40-person firm. The owner is at the kitchen table with a laptop, a phone, and a notebook that is older than the last software they bought.
QuickBooks has the invoice. The CRM has a contact and a stage that has not moved since March. The shared inbox has the actual conversation. Slack has the "can you look at this" from Thursday that never became a task. The warehouse, the shop, the field, the floor, pick your building: someone already did the work, or they are waiting, and none of the tools agree which.
He copies a job number from one screen into another. He forwards an email to himself so it will be in the right place later. He texts the person who "just knows." He writes three lines in the notebook because the notebook is the only thing that will still be true on Monday if a login breaks.
He bought ChatGPT in the spring. This morning he used it to draft the Monday note to the team. That is the AI. Then he pasted the note into Slack, attached a screenshot of the spreadsheet, and asked everyone to update their piece before the 9:30.
If you cancelled the seats tomorrow, Saturday morning would look the same.
The owner is still the integration.
Three systems do not make a company
Vendors will tell you this is a stack. It is not a stack. It is a person.
The person is usually the owner. Sometimes it is the office manager who has been there since the second location. Sometimes it is the coordinator who built the workbook everyone pretends is "the system." The title does not matter. The job is the same. Receive the truth in one place. Carry it to the next place by hand. Answer "where is this" because no screen can.
That is not heroic. That is unpaid middleware.
A 20 to 100 person firm does not have an integration team. It has one adult who will not let a job die in the gap. They check the inbox on Saturday because Saturday is when the gap shows. A customer followed up. A vendor changed a date. A tech sent a photo that is the whole update. None of it landed where the next person would see it, so it landed on the owner.
If you have never sat at that table, you will hear "we need to connect our tools" and buy another connector. The owner does not have a connector problem. The owner has work that has no home.
A seat does not carry the job
ChatGPT is a tool a person opens. Integration, in a firm this size, is what happens when a job, a deal, or a request has to survive the trip from the inbox to the person who can act.
Mixing them up is how you spend a year feeling modern and still spend Saturday as the bus.
A job has a start. An email, a PO, a call, a form that dumps into email. It has a path, even if the path is messy. It has a next step that should not require the owner. It has a state you can ask about without texting the one person who remembers.
A seat has a cursor. You type. It answers. You leave.
Watch it happen. The owner drops Thursday's email thread into the chat. The model returns a clean summary. He pastes three bullets into the Monday note. He does not paste the constraint in the fourth reply, the one that changes the date, because the summary sounded complete. Monday the team works the wrong date, politely, with confidence. The chatbot was helpful. The workflow was still a person copying.
Or someone on the team uses it to write a cleaner status update. Good paragraph. Then the paragraph lives in Slack, the number lives in QuickBooks, the next step lives in a head, and the version the customer has is a sentence from last Tuesday. You did not get an operating system. You got faster words around the same break.
This is why "we use AI" can be true and still be nothing.
Ask it at the kitchen table, not in a strategy offsite. If you cancelled the AI logins tomorrow, would the job move without you? Would the inbox still be the real system? Would "where is this" still mean a text to you? Would Saturday still be the integration window?
If the honest answer is no, you bought a writing assistant. You did not take yourself out of the middle.
The platform wants you to move in
Packaged software fails here for a boring reason. It wants the work to live inside its screens. Then it wants the other tools to visit.
A 40-person firm does not work that way. The work already lives in three places because each place won a different fight. Accounting won the ledger. Sales won the contact. The floor won the inbox, or the clipboard, or the group text, because that is where the next step actually showed up. The owner is what you get when none of those wins included the whole job.
When you buy a platform, you are asked to pour the company into someone else's idea of a pipeline. The first month is data entry. The second month is workarounds. The third month Saturday is back, because Saturday still listens when the process changes on a Wednesday.
That is not the team being difficult. That is the team protecting the customer.
Connectors fail the same way, quieter. They move a field. They do not move a next step. A new row appears in a sheet. Nobody owns it. The owner finds out on Saturday, when the customer asks a question the row cannot answer.
The companies that get something real out of software in a firm this size put the job in a system that matches how they already run, and they change that system when the floor is right. Not a chat window on the side. Not a Zap that creates a card nobody opens.
The job. The next step. The change that came in at 4pm. The Friday question that used to mean the owner opening five tabs.
One process is enough to start. Intake. "What's next on this?" The handoff that keeps landing on you. Whatever makes Saturday required.
Build that. Put it in the team's hands. Change it when they say it is wrong.
That last sentence is the part most custom software skips. The system freezes the day it ships. The owner goes back to the notebook because the notebook still listens.
The claim is the change, not the connector
This is the claim only Mantum can make.
We do not sell a platform you move into. We do not sell a connector that copies a field and calls it done. We build the broken workflow as software that matches how you already run, and we keep building it after the first version is in someone's hands. The floor flags the bottleneck. They name the data point that was missing. The system gets configured around the process they already have. The update goes live in days, not after a rigid no about scope.
We have that in the wild. ALO Property Group is a commercial real estate firm in Indianapolis. Their brokers used to keep the client in private CRMs the firm could not see. The next step lived in a login only one producer had. The firm was not the system. The producer was. They built ALO CRM around the floor they already had. When the process needed a new field, it did not die in a requirements document. It shipped in days. Hundreds of product updates in 45 days.
Your firm is not a brokerage. The pattern is the same. The real job lives in the gap between tools. The person in that gap is usually you. The system that wins is the one that can hold the next step, and move when the floor is right, so Saturday is not a second shift.
If your current vendor needs a statement of work to put the note that is already in your notebook onto the job, you do not have a partner. You have a product with your logo on a generic pipeline.
What this is not
If the seats help you write a clearer Monday note, keep them. Just stop calling that adoption. Do not ban the notebook on Friday either. If you ban the paper before the new path works, you will build a worse notebook in a notes app you check at 10pm.
Mantum builds a custom system for a broken workflow. The job. The next step. The Saturday question. Not a packaged suite with your logo on someone else's objects, and not a claim that a model will run the firm.
We are not going to tell you the owner-as-glue is a leadership style. It is a failure of the system, held together by someone who will not let a customer fall through. That is also why it works. Any replacement that cannot be corrected by the person who currently is the integration will lose to the notebook.
Watch one Saturday before you buy another login
Do it as a thought, not a procurement event.
Pick one live job. Not a hypothetical. A real one that hit your phone this week.
Write the path on a single page. Where it came in. Which tool saw it first. What you copied. What you forwarded. Who you texted. Where the next step lives right now (a head, a thread, a cell, the notebook). What would happen to that job if you did not sit down on Saturday.
Circle the first place the job needed you in order to move. That circle is more honest than an integration roadmap.
If the circle is you, you do not have a training problem. You have a system that cannot hold the work.
If you want that path built as software that matches how you already run, tell Mantum what is broken. Not the connector you wish you had. The handoff that keeps landing on you.