AI Scheduling for Small Builders: Subcontractors and Deliveries

Partly. AI can draft a programme from your trade sequence, work out order-by dates from merchant lead times, write the confirmation messages to subcontractors, and re-plan the next few weeks when something slips. It can't know whether your plasterer is actually free, so a person still confirms every date before it counts.

That split matters because the failure mode is a tidy programme full of dates nobody agreed to. Subcontractors' availability lives in their heads and phones, not in any system an AI can read, and merchants' delivery slots change daily. Treat AI as a planner's assistant that does the sequencing, arithmetic and writing, and keep the phone calls and confirmations with you.

Follow me on Instagram@sagnikteaches

What AI can and can't see when it plans your job

Build the sequence once: a trade-dependency sheet

The single most useful thing you can give an AI is a dependency sheet: every task, who does it, how long it takes, what must finish first, and what has to be ordered. Build it once for each type of job you do, then copy it for each new project. Here is an illustrative sheet for a single-storey rear extension run by a three-person building firm. The lead times are examples only; use what your own suppliers quote.

Connect on LinkedInSagnik Bhattacharya

Keep it in a spreadsheet, not in a chat. The chat is where you ask questions of the sheet; the sheet is the record.

Subscribe on YouTube@codingliquids

Turning lead times into order-by dates

Late deliveries are where most small-builder programmes fall apart, and it's usually an ordering problem, not a delivery problem. Once the sheet has planned start weeks, AI can work backwards from each one. A prompt that works:

Using the attached dependency sheet and a start date of week 1, calculate
the planned start week of every task. Then, for every item in the
"Order needed" column, give an order-by week = planned start week minus
lead time minus a 1-week buffer. Show the working in a table.
Flag any item whose order-by week is before week 1, and any item whose
lead time only starts after an earlier task (e.g. a survey of the built
opening).

An illustrative extract of what comes back:

Item                 Needed on site   Lead time   Order by
Matching bricks      Week 3           6 weeks     Week -4  FLAG: before start
Steel beam           Week 5           3 weeks     Week 1
Rooflights           Week 5           3 weeks     Week 1
Bifold doors         Week 6           8 weeks     Week -3  FLAG: before start
Kitchen units        Week 10          8 weeks     Week 1

FLAG, bifold doors: if the supplier must survey the built opening
(end of week 4), doors arrive around week 13, long after plastering.
Options: order to drawing sizes now with the opening built to suit,
or board up the opening and fit the doors late.

That last flag is the kind of thing that costs a builder weeks when it's spotted too late. Two things to check in any output like this: that the week arithmetic is right (spot-check two rows yourself), and that the lead times are this month's, not the ones in your sheet from last spring. A quick call to the merchant beats any AI on that second point, and tracking supplier prices and lead times with AI helps keep the sheet current.

The Friday three-week look-ahead

Whole-job programmes go stale within a fortnight. What keeps a small builder's jobs on track is a short, regular look-ahead: every Friday, what is happening on each job over the next three weeks, who needs confirming, and what needs to arrive. AI makes this a twenty-minute routine instead of an evening.

1. Update actual progress. Record a two-minute voice note per job: what finished, what's behind, anything new. Paste the transcripts into the chat with the dependency sheet.
2. Ask for the look-ahead. "Update planned dates from this progress. List every task, trade and delivery for the next three weeks, by job and by day. Mark anything unconfirmed."
3. Get the messages drafted. "Draft a confirmation message for each subcontractor starting in the next two weeks and a delivery booking for each merchant order."
4. Send, and log replies. Mark each date confirmed only when the sub or merchant replies. No reply by Tuesday means a phone call.

An illustrative confirmation message the AI produced for a plasterer, after light editing:

Hi, plastering at the rear extension is still planned for Monday to Thursday, 13th to 16th, four days. Boards and beads arrive Friday 10th. First-fix electrics and plumbing will be finished by Thursday 9th. Can you confirm by Tuesday? If anything's changed your end, tell me now and I'll rework the week.

And a delivery booking to a builders' merchant:

Please deliver order 4471 (12 packs of plasterboard, beads, 20 bags of multi-finish) on Friday 10th, morning slot, not before 8:00. Narrow access: the driveway takes a rigid lorry only, and we'll need kerbside offload with the lorry's crane. Site contact on the day is the number below.

Notice what makes the second one useful: the order number, the time window, the vehicle limit and the offload method. Those are the details that decide whether a delivery lands or goes back on the lorry. Put them in the dependency sheet so the AI includes them every time. If you'd like the routine to prompt you automatically, ChatGPT's scheduled tasks can remind you on Friday mornings.

When the roofer drops out: re-planning a slip

Here is where AI earns most of its keep, because re-planning by hand means redoing every downstream date and remembering every message. Suppose the roofer on the extension above calls to say he can't start until a week later than planned. The prompt:

The roofer can't start until Monday 6th instead of Monday 30th.
Re-plan the next four weeks using the dependency sheet. Keep every
dependency. List every trade and delivery that moves, the new dates,
and anything that now clashes. Draft a message to each affected party.

An illustrative reply moved the window fitter, first fix, plastering, the screed booking and the skip exchange by a week, and drafted six messages. It also suggested "starting first-fix electrics inside before the roof is on, to recover two days". That is the line you'd delete: first fix before the building is watertight risks wet cable and wet insulation, and your electrician will refuse anyway. AI re-plans by logic and has no sense of rain. Read every suggestion that breaks the sequence you set.

The re-plan is also the moment to check the client-facing side. A week's slip is news the client should hear from you, with the new dates, before the kitchen fitter mentions it. Writing client updates with AI covers that message.

Weather: letting the forecast move the right tasks

Some tasks don't care about rain; others are ruined by it. Mark the weather-sensitive ones in the dependency sheet with the rule that applies, for example: concrete pours and screeds need no frost overnight; flat-roof membranes need a dry day; external render and brickwork stop below a set temperature; roof stripping only with a dry two-day window. Then, on Thursday, paste the next week's forecast into the look-ahead chat:

Here is next week's forecast and the look-ahead for the extension.
Using the weather rules in the dependency sheet, list every task
that clashes with the forecast, suggest the smallest swap with a
task that isn't weather-sensitive, and draft the messages.

An illustrative reply: "Tuesday and Wednesday show heavy rain. The flat-roof membrane (Tuesday) clashes. Swap: bring forward the internal stud walls planned for Friday, and move the roofer to Thursday, the first dry day. Knock-on: rooflight fitting moves from Wednesday to Friday." That's sensible, but check one thing it can't know: whether the roofer is free on Thursday. The AI's swap is a proposal until the phone call. Also watch for forecasts pasted in from a week earlier; an old forecast gives confident, wrong advice.

The day-before delivery check

Many delivery problems are decided the day before, when nobody checks the basics. Have the AI turn tomorrow's deliveries into a short checklist for whoever is on site. An illustrative one for a plasterboard drop:

• Delivery confirmed for the morning slot, order number 4471, 12 packs of board plus beads and plaster
• Rigid lorry only; driveway clear of skips and vans by 7:45
• Crane offload to the front garden; boards off the ground on bearers, under a sheet
• Site lead on site from 7:30 to check the delivery against the order
• Next-door neighbour told the lorry will park across their drive for 20 minutes

The fourth and fifth lines are the ones that get skipped. On one illustrative job, boards were left on the drive with no cover because the only person on site was a subcontractor who didn't know they were coming; a shower later, two packs were scrap. A one-line checklist item would have saved the cost of the packs and a day's plastering.

When a subcontractor confirms, then doesn't show

A confirmation on Tuesday doesn't guarantee a van on Monday. Add a second, lighter check the working day before each start: "Morning, just checking you're still on for the kitchen at 8 tomorrow. Anything you need from us on site?" AI can draft these in a batch from the look-ahead. Keep a short list of backup subcontractors for each trade in the sheet as well, so that when a no-show happens the re-plan prompt can include "the backup plasterer is available from Wednesday" instead of leaving a gap for a week.

Two jobs, one crew: catching double-bookings

Small builders rarely run one job at a time, and the clash that hurts is usually your own people or your favourite bricklayer booked on two sites in the same week. AI is good at spotting this if you give it both jobs' sheets and a simple resource list: who you have, and how many days a week each is available.

Here are the dependency sheets for Job A (extension) and Job B (loft
conversion), and my resource list: own team of 2, bricklayer 5 days a
week, electrician 3 days a week. List any week where a resource is
booked for more days than it has. Suggest the smallest change to fix
each clash, and say which job's finish date moves.

An illustrative reply: "Week 4: the bricklayer is booked for 5 days on Job A and 2 on Job B. Smallest fix: move Job B's chimney breast removal make-good to week 5; Job B's finish date doesn't move because that task has 4 days of float." Float, meaning spare time before a task delays the finish, is exactly the thing most builders carry in their heads and never write down. Once it is in the sheet, the AI can use it to choose which job absorbs a clash.

Check its suggestion against what you know: the bricklayer may simply prefer to do both jobs with a labourer, or the loft client may have a moving date you can't touch. Add fixed dates like that to the sheet as "must finish by" so the next re-plan respects them.

Where the programme should live

For one to three jobs at a time, a spreadsheet dependency sheet plus a general AI assistant (ChatGPT Plus or Claude Pro, about $20 a month each) is enough. Past that, or when more than one person books trades, a scheduling module in job-management software stops the "which version is current?" problem.

Much of what these modules do is rules-based rather than AI. Buildertrend's schedule, for example, sends notifications to subcontractors and clients about upcoming tasks and changes, lets you send a subcontractor a confirmation request so they acknowledge a booked task, and has workday exceptions for weekends and holidays. JobTread includes tasks and scheduling alongside sub and vendor portals. Those features solve the confirmation problem better than a chat does, because the sub's acknowledgement is recorded against the task. AI then sits on top: drafting the look-ahead, writing messages, and re-planning. For a wider view across trades, see AI job scheduling and dispatch for field service teams.

Take an illustrative firm running four jobs at once. The owner keeps each job's dependency sheet in the job software, runs the Friday look-ahead through a chat that reads an export of the schedule, and uses the software's confirmation requests for every subcontractor booking. Nothing counts as booked until the sub has pressed confirm.

Signs an AI-assisted programme is drifting from reality

• Dates without confirmations. If more than a couple of next-week bookings are unconfirmed on Friday, the programme is a wish list.
• Deliveries with nobody to receive them. Every delivery should have a named person on site, a place to put it and, for plasterboard or timber, somewhere dry.
• Two trades in one small room. AI will happily book the plumber and the plasterer into a 3 m by 4 m kitchen on the same day. Add a "can't share space with" note to your sheet.
• Lead times that never change. If the sheet still says six weeks for bricks from last year, the order-by dates are fiction.
• Friday deliveries of anything perishable. Ready-mix, plaster and screed booked for late Friday are a weekend risk. Tell the AI your rules and it will follow them.

Once variations start arriving, each one's time impact should flow straight into this programme; pricing variations and change orders with AI shows how to capture the days as well as the money, and the estimating software comparison covers tools that tie estimates to schedules.

Programming questions small builders ask

Can AI book subcontractors directly by text or email?

Technically an automation can send the messages, but for a small builder it is better to have AI draft them and send them yourself, at least until the routine is settled. A subcontractor who gets an automated booking with a wrong date loses trust quickly, and a reply like 'can do Tuesday, not Monday' needs a person to decide what moves. Automate the reminder to confirm, not the booking itself.

Do I need Gantt chart software for this?

Not for one to three jobs. A spreadsheet dependency sheet with planned start weeks does the same work, and AI can produce a simple bar chart from it if a client wants to see one. Job-management software with a schedule becomes worth it when several people book trades, when you want subcontractors to confirm inside the system, or when clients expect a portal.

How far ahead should I confirm subcontractors?

Book them provisionally when the job is won, using the dependency sheet, then confirm each one roughly two weeks before their start and again at the Friday look-ahead the week before. The two-week confirmation catches clashes with their other work; the final one catches last-minute changes. Anything still unconfirmed a few days out deserves a phone call rather than another message.

Further reads

• Can AI Produce Accurate Estimates for a Small Building Firm? — Realistic estimates give you realistic task durations for the programme.
• How to Compare Supplier Quotes Side by Side With AI — Compare merchants' prices and lead times side by side before ordering.
• Can AI Manage Your Calendar and Book Meetings for You? — What AI calendar tools can and can't do with your own diary.
• The First 30 Days of AI in a Trades Business — Where programming fits among the first AI jobs to automate.
• Turning On-Site Voice Notes Into Job Sheets and Quotes With AI — Capture site progress by voice so the programme stays current.
• What Tradespeople Actually Use AI For: 12 Real Jobs It Handles — Other everyday jobs AI handles for trades businesses.
• AI Shift Scheduling vs a Rota Spreadsheet for Small Teams — A rota spreadsheet is fine until swaps, availability and rules outgrow one person. Costs, a tested AI-drafted rota, and the team size where apps win.
• AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.

Sources: Buildertrend help articles on schedule notifications, subcontractor confirmations and workday exceptions; JobTread features page; OpenAI and Anthropic plan pages.