You put a date and a dollar figure on every selection before the job starts, then let software do the back-scheduling. Selections stall because nobody told the client when the decision was due or what waited on it. Pull your allowance lines into one list. Have Claude work backward from your build schedule to set a decide-by date for each one. Keep every approval written down somewhere your team can search. Two tools, about ninety minutes, and you only build it once.

Not sure which part of your operation is leaking first? The Skill Match assessment takes five minutes and tells you.

Because a selection is a schedule task and almost nobody treats it like one.

Think about how it usually goes. You send a selections list at kickoff. It has items on it. It has no dates. The client reads it as homework she can do whenever.

Meanwhile your framer, your plumber, and your tile setter all need answers on different weeks. Nobody told her that. So the first time she hears about urgency is when you call, annoyed, three days before the trade shows up.

Then it gets expensive. Every late decision turns into a change order written under pressure. Roughly 40% of change orders never get invoiced. About $2,847 a project walks out the door, and three disputes per job get argued over work nobody wrote down.

The delay is the cheap part. The unbilled work is the expensive part.

One list. Every selection on the job, with five things attached to each row.

That is the whole idea. You are not buying a new portal. You are adding four columns to a list you already have.

Builders tend to keep allowances in one system, lead times in a supplier email, and the schedule somewhere else. The numbers that should answer one question sit in three places. Scattered data is the whole job on the commercial real estate side too. It is where people like Kevin Shtofman at Cherre spend their days. Same shape of problem, much bigger buildings.

Three steps. Claude does the arithmetic, Gemini Notebook holds the record.

Open your budget in whatever runs the job and export the line items that carry an allowance. In JobTread that is a budget export to CSV. Buildertrend and Ressio both do the same thing from the estimate.

You will get the item, the room, and the dollar figure. You will usually not get the lead time. Add that column yourself, once, from your own supplier list. This is the part nobody tells you, and skipping it is what makes the rest of this fail.

(Worth saying plainly: GO First implements JobTread for builders. Use whatever you already own. The workflow does not care.)

Paste the sheet into Claude along with your build schedule. Then give it this:

The second half of that prompt is the part that pays. You get a sorted list of what is already late, by how much money is sitting on it. That list is your Monday morning.

Create a notebook in Gemini Notebook, the tool you may still know as NotebookLM. Add the decision calendar, the signed contract with its allowance schedule, and a running document of selection confirmations.

Then anyone on your team can ask it questions and get an answer with a citation:

The free tier covers this. Check the current plan page before you count on any one feature.

The whole build takes about ninety minutes the first time. Maybe twenty minutes a job after that.

Four places. Every one of them is worth knowing before you start.

Claude will invent dates it has no business inventing. Give it lead times or it will guess, confidently, and your calendar becomes fiction. Garbage lead times produce a garbage schedule faster than a spreadsheet ever could.

Your export will not have everything. Budget exports carry the allowance, not the supplier lead time. That column is manual the first time. Build it once as a reusable list and stop rebuilding it.

Notebook only knows what you loaded. It cannot tell you about the approval that happened on a phone call and never got written down. If the answer is wrong, the fix is a better document, not a better prompt.

Bad source data costs you the client's trust. One builder I sat with had already written AI off entirely. It had handed him wrong shop sizes for windows, and that was that. He was not wrong to be careful. Run this on one job, check the dates against what you know, and only then run it on the rest.

None of this replaces the conversation. Software sends a reminder. It cannot tell a client the tile she loves is nine weeks out and the shower is already framed. You still have to say that, early, in plain words.

Some will not. Usually your highest-budget one.

Stop fighting it. Take the call, then send a short written confirmation the same day that repeats the choice and the dollar figure. Two minutes. Then paste it into the notebook so it exists somewhere other than your memory.

The goal was never portal adoption. The goal is a written record of the yes.

This is the same lesson builders keep learning about anything that sits outside their control. Permit timelines are the other classic one, and it is why permit platforms for residential builders exist at all, like Robbie Leer's TurboPermit. You cannot make the other party move faster. You can put their piece on the calendar with a date and stop being surprised by it.

Yes, and you should say so before you sign.

Put a re-selection or late-decision fee in the contract, next to the decision calendar, and walk the client through both at the same meeting. Sold that way it reads as competence. Sprung in month four it reads as a penalty.

A defined sequence you can sell is most of what separates a trade business that scales from one that does not. Sales-process work like Mike Christie's Roofing Revenue System exists for that reason on the roofing side.

Most builders never charge the fee. That is fine. Having it in writing is what makes the date real.

When should the selections schedule be built?
Before the contract is signed, during pre-construction. The dates have to be yours to set, and once the job starts they are not. Building it at kickoff is already late.

What is a selection allowance in construction?
A dollar amount carried in the budget for an item the client has not chosen yet, like tile or plumbing fixtures. The client picks inside the allowance, and anything above it becomes a change order.

How far in advance do clients need to make selections?
Work backward from the install date, not forward from today. Add the supplier lead time, then add two weeks of buffer for anything over six weeks out. Long-lead items like custom windows and cabinets get pulled to the front regardless of build order.

Can AI make selections decisions for my client?
No, and you would not want it to. AI handles the back-scheduling, the sorting, and the record-keeping. The taste and the money are the client's.

Why did my allowance number change mid-job?
Often because the baseline moved. Some tools drop the original allowance figure when an estimate is reopened and locked again after a selection was approved. Test it on a dummy job before you trust the number in a client conversation.

Do I need new software for this?
No. You need a list, a schedule, and somewhere searchable to keep the approvals. Whatever already runs your jobs will export the first two.

Pick your two biggest active jobs. Find one selection with no date on it. There will be one.

Put a date on it today, then build the rest of the calendar this week while your fall jobs are still in pre-construction.

If you want to know which part of your operation is costing you the most first, start with the Skill Match assessment. Five minutes, no call required.

For more of how builders are running this, Beyond the Bid is the podcast.

Grant Fuellenbach builds operating systems for residential
builders and remodelers doing $1M to $15M. He came up through construction technology,
starting with drones, sUAS, photogrammetry, and LiDAR, before founding GO First
Consulting. He has spent the last year putting selection calendars and allowance records
into builders' own systems, on real jobs, and watching which parts of it hold.

Published September 18, 2026. Last updated September 18, 2026.