Concrete Pour Schedule: Plan Pours Around Weather and Crews With AI

Give ChatGPT or Gemini three things: your go or no-go weather rules, the next five days' forecast pasted in, and which crews, supplier slots and pump are free. Ask for a ranked list of pour days with reasons, then draft messages to the supplier, pump company, crew and client. You make the call; the assistant sorts and writes.

An assistant only knows the weather you give it, and it cannot see your sub-base or the sky over the site. What it adds is consistency: the same written rules applied to every job each day, and a full set of messages ready to send before the supplier's cut-off. The rules, the cut-off hour and the final decision stay with a person.

Follow me on Instagram@sagnikteaches

Write your weather rules once, as a table the assistant must obey

An assistant can only apply rules that exist. Write each one as a condition, a number and an action. The table below mixes the ready-mixed concrete industry's own guidance (the NRMCA Concrete in Practice sheets) with triggers a firm sets for itself, and the last column tells you which is which. Treat the "your rule" lines as examples to replace with your own.

Connect on LinkedInSagnik Bhattacharya

Notice the gap between the industry definition and the rule in row two. The NRMCA sheet defines "cold weather" as more than three consecutive days with an average temperature below 5 °C (40 °F), and air not above 10 °C (50 °F) for more than half of any 24 hours. A single hard frost on fresh concrete can do serious damage long before a definition like that applies, so a firm's own rule should act earlier. Write it as one sentence the assistant cannot misread: if the forecast shows 2 °C or lower at any time up to 48 hours after the planned start, flag a cold-weather plan.

Subscribe on YouTube@sagnikteaches

Two more details belong in the rules because they change what a pour day costs. Do not use water to help finishing: the NRMCA guidance says never to sprinkle water or cement on the surface to ease finishing, because it causes dusting or scaling. And rain on a fresh slab is not rescued by working the water in: NRMCA's dusting sheet names finishing while water is on the surface, and leaving fresh concrete unprotected from rain, among the causes of a weak, dusty top. Push standing water off with a float or squeegee, and have plastic ready and weighted down at the edges; the curing and discoloration sheets warn that sheeting which wrinkles or touches the surface unevenly can leave dark streaks or blotches. If the slab is a decorative or exposed finish, say so in the rule, because the allowed method changes.

Where the rules sit matters as much as what they say. Keep them where every day's prompt can start from the same text: a Project in ChatGPT or Claude, or a Gemini Gem (becoming a skill from November 2026). The tutorial on ChatGPT Projects explains the setup, and the quote clauses that tell a client about weather delays are covered in the concrete quote template.

Match truck spacing to what your crew can place

The weather decides whether you can pour; the crew decides how fast concrete can safely arrive. The NRMCA hot-weather sheet puts it plainly: schedule the rate of delivery that can be managed by the available crew and placement equipment, and make sure trucks can discharge immediately. So before the assistant sees a forecast, give it a grid of what is actually free.

The sum that links crew to supplier is short: truck spacing in minutes = load size ÷ placing rate × 60. Say a driveway needs 14 m³, the supplier sends 7 m³ loads, and your crew's own records show it places 7 m³ an hour. Spacing is 7 ÷ 7 × 60 = 60 minutes, so two loads an hour apart and a placing time of about two hours. Now the age check. Suppose the supplier's limit is 90 minutes from batching to the last of the load being placed (ask for yours) and the plant is 25 minutes away. Each load is 25 + 60 = 85 minutes old by the time it is empty. That passes, with five minutes to spare, and a truck held at the gate uses them up. The NRMCA sheet advises against adding water to concrete more than 1½ hours old, which is why the limit is not a formality.

If the same crew places 5 m³ an hour, a 7 m³ load takes 84 minutes to empty and is 109 minutes old at the end. Three options follow: more hands, smaller loads, or a different day. Smaller loads, for example 3.5 m³ every 42 minutes, keep each one at 67 minutes old, but they mean more trucks and possibly more per-load charges. That trade-off is the owner's to price, and the assistant is useful for laying the options side by side once you have given it the numbers.

The day-before check belongs in the grid too

The NRMCA finishing guidance says to complete sub-base excavation and compaction, formwork and the placing of mesh or other embedments before the concrete is delivered, because delays after it arrives reduce the quality of the slab. Add a line to the grid for "ready to pour" with a named person and a time, usually late afternoon the day before, so a forecast-driven "go" is never confirmed on a site that is not ready.

The daily prompt: forecast in, ranked days out

Here is a week for a six-person firm with two crews of three. Three jobs are waiting: a 14 m³ driveway that needs both crews and a pump, a 6.5 m³ garage base for one crew and a pump, and a 4 m³ patio to barrow. Crew 2 is away on another site on Tuesday and Wednesday. The supplier has no slot on Monday and a 07:30 first truck on every other day. The forecast, pasted from whichever source you trust, is below (illustrative figures).

You are my pour-day planner. Apply ONLY the rules below; do not add rules of your own.
RULES: [paste the weather rules table]
RESOURCES: [paste the crew, supplier and pump grid]
JOBS: [list each job: m3, crews needed, pump or barrow, earliest and latest date]
FORECAST: [paste the 5-day forecast with the issue time]
For each job give: best day, a verdict for every day (Go / Go with plan / No-go),
the rule or constraint that decided it, and the kit and supplier requests the
plan needs. Quote the forecast figure behind every weather verdict.
If anything is missing or ambiguous, ask me before you answer.

An illustrative reply, trimmed:

That reply is tidy, and it still has to be read against the rules, line by line. Typical corrections, each of which has a rule behind it:

• A rule applied half-way. The assistant may call Friday a plain "Go" because 2 °C is above freezing. Your rule says 2 °C or lower within 48 hours triggers a plan, so the verdict is "Go with plan". Tell it which rule it missed and ask for the table again.
• Units mixed. A forecast in km/h pasted next to a rule in mph produces confident nonsense. State the unit in the prompt and ask the assistant to repeat the numbers it used.
• A constraint ignored. It may suggest a 05:30 start to beat the wind, forgetting that the supplier's first truck is 07:30. Constraints in the grid are as binding as weather rules; say so in the prompt.
• A rule tightened without saying so. The patio row says no suitable day, yet your rain rule allows a small pour with sheeting on site and one named decision-maker, so Wednesday morning with Crew 1 is really a "Go with plan" you may or may not take. Ask for the rule behind every No-go, put a number on "small" in the rule (up to 5 m³, say), then make the call yourself.
• A window the forecast does not reach. A Friday pour's 48-hour cold check runs to Sunday morning, past the end of a Monday-to-Friday forecast. The prompt tells the assistant to ask when something is missing, and it should have; if it did not, paste a seven-day forecast and run it again.

Keep the prompt, the pasted forecast and the reply together in the job's folder. If a slab shows a problem later, that record shows what the plan said and what you decided.

Put it on a timer with ChatGPT tasks or Gemini schedules

Once the prompt works by hand, run it automatically. ChatGPT has scheduled tasks, created and managed on a Scheduled page. OpenAI's help page lists them for eligible Free, Go, Plus, Pro, Business, Enterprise and Edu users, with limits on active tasks that run from 15 on Pro and Enterprise and 10 on Business and Edu down to 5 on Plus and 3 on Free and Go. Event-triggered tasks, which respond to activity in apps such as Gmail, are not available on Free or Go. You choose push, email or both in the notification settings. The same page has two catches for pour planning. On Free, a task runs at most once a day in a loose window such as morning or afternoon, and exact times need a paid plan (OpenAI's pricing page has at times shown no tasks on Free at all), so a fixed 13:00 run means a paid plan. And a task created inside a project cannot read the files uploaded to that project, so paste the rules into the task's own instructions rather than leaving them as a project file. The ChatGPT scheduled tasks tutorial walks through the setup.

Gemini calls its version scheduled actions: up to 10 active schedules at a time, available on a personal Google account (being rolled out gradually) or a work account on a qualifying Workspace edition, and only with Keep Activity turned on. Two details in Google's help text matter for pouring. Each response is prepared within the hour leading up to its delivery time, so a forecast in it can be up to an hour old when it lands. And a schedule that depends on location uses the place where you created it for every later run, so name the site's location in the prompt or the forecast may be for your office, not the slab. Use it for the day-before plan, not for a 05:00 go or no-go. The Gemini scheduled actions tutorial covers the steps.

Whichever you use, give the task three instructions beyond the prompt: quote the forecast source and the time it was issued, say "forecast unavailable" instead of guessing if it cannot find one, and send the reply to the person who owns the cut-off. A sensible rhythm is one run at 13:00 for the next two days, ahead of a 15:00 decision hour, plus a manual re-run at the decision hour if the forecast has moved. If you later automate the forecast pull yourself through a weather data service, read its terms first. Open-Meteo's pricing page, for instance, says its free API is for non-commercial use and that a subscription grants a commercial licence, which a contractor's use would need.

Links for the two help pages: OpenAI's scheduled tasks page and Google's scheduled actions page. Limits and eligibility change, so check both before you rely on a number.

The five messages to send once the day is chosen

Once the day is chosen, the same facts go to five readers who each need a different part of them. Ask the assistant to draft all five from the decision table, then edit every one before it goes out. The wording below is illustrative; names stay as placeholders.

1. Supplier

Hi [first name], it's [your name]. Can you hold 14 m³ for Friday at 07:30, as two 7 m³ loads an hour apart? Because of the forecast I need the concrete delivered at no less than 13 °C. Please confirm the latest time I can cancel with no charge, and that no water is added on site without my say-so, with any addition written on the delivery ticket. I'll confirm go or no-go by 15:00 on Thursday.

The water line comes from the NRMCA guidance on adding water on site: any water added must be measured and recorded on the delivery ticket, and extra water beyond the design mix delays setting, reduces strength and raises the risk of cracking. Asking for it in writing gives the driver and your foreman the same instruction.

2. Pump company

Hi [first name], confirming Friday: set-up from 07:15, first load at 07:30, pumping done by about 09:45. Please confirm hose length for the driveway, the call-out terms and your cancellation cut-off. I'll message by 15:00 on Thursday with a firm yes or no.

3. Crew

Friday pour, driveway. Yard at 06:45, van loaded with insulated covers, sheeting, curing compound, floats and squeegees. Frost is possible at dawn, so check the forms and steel before the first truck. I'll text go or no-go by 15:00 Thursday. If it's no-go, Tuesday is the fallback.

4. Client heads-up

Hi [first name], we're planning to pour on Friday morning if the weather allows. The concrete needs dry conditions and temperatures above freezing for the first couple of days, so I'll confirm by 3 pm on Thursday. Please keep the access clear and keep vehicles and people off the area until I tell you it can be used.

5. The rain-off text

This is the one that goes wrong, because an anxious draft reads as an apology. Compare a typical first draft with the edit.

First draft: Hi [first name], I'm really sorry but unfortunately we're not going to be able to pour tomorrow because the weather isn't looking good and I wouldn't want to risk anything going wrong with your lovely new driveway. I'll let you know about a new date as soon as I can.

Edited: Hi [first name], we're not pouring tomorrow. The forecast shows showers from 13:00 with a 60% chance, and rain on a fresh surface can leave it weak and dusty. New slot: Tuesday at 07:30, or Wednesday if the forecast looks better; I'll confirm by 3 pm on Monday. Your price is unchanged and there is nothing you need to do except keep the access clear.

The edit states the decision, the reason in plain words, two options and a confirmation time. It does not offer a date you cannot hold. A good prompt for the whole set: draft five messages from this decision; each must state the decision, the time of the next update, and nothing the recipient does not need; use [first name] and [your name] placeholders; do not promise a date. Before sending, check each number in each message against the decision table, because assistants copy times confidently and sometimes wrongly.

Set the cut-off hour with one sum

Every pour has a moment after which cancelling costs money. The percentage on a forecast matters less than whether you decide before or after that moment. Put the costs in a table, using your own figures. These are illustrative.

The break-even is the cost of holding divided by the cost of damage: $770 ÷ $2,800 = 27.5%. If the chance of rain between the start and the end of finishing is above that, holding has the lower expected cost; below it, pouring does, provided sheeting is on site. At the 60% in the Wednesday forecast, pouring carries an expected $1,680 of damage against $770 to hold, so it is a hold. At 20% the expected damage is $560, so it is a pour with a plan. The line that matters most is the second row: if you decide before the cut-off, holding costs nothing at any percentage. That is why the decision hour in your rules should match the supplier's and pump company's free-cancellation times, and why the daily run is worth setting up.

Treat the 27.5% as a way to set your own threshold, not as a law. A forecast percentage describes a chance of measurable rain in an area and a window, not on your slab, and the damage figure should include the cost of a client who now distrusts you. Ask the assistant to do the arithmetic once you supply the three figures, and read the working.

After the pour: the saw-cut window, curing and a log

The schedule does not end when the last load is placed. Three timings sit after it, and the first is easy to miss because it is set by the concrete, not by your day.

• Saw-cutting. The NRMCA joints sheet says conventional saw-cuts should be run within 4 to 12 hours after the concrete has been finished, and early-entry dry-cut joints generally 1 to 4 hours after, depending on how the concrete sets. If the edges ravel, delay the cut; if you wait too long, sawing becomes difficult and uncontrolled cracking can occur. A slab finished at 10:30 has a conventional window of 14:30 to 22:30. Book the saw operator for it, and move it later on a cold day, when setting is slower.
• Curing. Start as soon as finishing is complete. NRMCA's curing sheet describes typically 3 to 7 days, with liquid compound applied once the water sheen has gone and two coats at right angles described as desirable for even coverage. Its cracking sheet adds that a second application the next day is a good quality-assurance step, so put that visit in the schedule.
• Rain cover. If showers are forecast within hours of finishing, sheeting sealed at the edges goes on straight after curing compound, and comes off when the rain stops.

Then keep a pour log: date, job, cubic metres ordered and placed, the forecast from the 13:00 run, the forecast at the decision hour, the decision, the actual weather, any surface problems (dusting, plastic shrinkage cracks, scaling), the saw-cut time and the cost of any wrong call. After ten pours, paste the log into an assistant and ask: which of my rules would have changed a decision, and did any hold turn out to be unnecessary? Adjust one rule at a time. The roofing weather-delay tutorial uses a similar message-and-log loop for another outdoor trade.

Where an AI-drafted schedule goes wrong

The failures are predictable, which makes them cheap to guard against.

• A forecast nobody supplied. An assistant with no forecast will still produce plausible weather. Make "forecast unavailable" an acceptable answer and require the issue time on every forecast it uses.
• Daily numbers standing in for hourly ones. "60% chance of showers" for the day says little about the four hours of placing and finishing. Ask the assistant to quote the hours that overlap the pour window.
• Rules that collide. "Pour early to beat the heat" meets "supplier's first truck is 07:30". An assistant may resolve the conflict silently. Add a line to the prompt: list any two rules or constraints that conflict before giving a verdict.
• Dates and units. 03/04 can be March or April, a forecast pasted from a different source may use km/h while your rules use mph, and a clock change can shift an hour. State the format and unit in the prompt and have the assistant repeat the figures it used.
• Things the assistant cannot see. A saturated sub-base, standing water in the formwork, frost on the steel and a crew member who rang in sick are invisible to it. The 06:00 look at the ground is not optional.
• A schedule that never says no. If every verdict comes back "Go", the rules are not being applied or are too loose. Test them with a forecast you know should be a no-go.
• Client details in the wrong plan. Use job references rather than names and addresses in prompts. If the crew uses personal ChatGPT or Gemini accounts, find the model-training switch in each one's privacy settings; on team and business plans, your content stays out of model training by default.

None of this removes the need for a person at the site at dawn. It changes the day-before routine: instead of an hour of phone calls and guesswork, you check one table, edit five messages and send them before the cut-off. For the wider picture of where AI fits across a concrete firm's week, including estimating, see AI for concrete contractors.

Questions about AI and pour-day planning

Can ChatGPT or Gemini check the weather by itself?

Both can search the web, but treat a summary as a pointer, not a forecast you can book trucks on. Paste the hourly forecast you trust, or tell a scheduled task to quote the source and the time the forecast was issued. Then check the assistant's reading of it against the original before you confirm anything with the supplier.

What if the forecast changes after I have booked?

Run the same prompt again at your decision hour with the new forecast and send the matching message. Set the decision hour to match the supplier's and pump company's free-cancellation cut-offs, so a change costs nothing. If the change comes after the cut-off, use the cost sum in this tutorial to decide whether a cancellation charge is cheaper than the risk.

Who decides whether to pour, me or the assistant?

You, or your foreman on site. The assistant applies the written rules to a forecast and drafts messages. It cannot see standing water, frost on the steel, a soft sub-base or a crew member who has not turned up. The last check on the morning is always a person looking at the ground and the sky.

Further reads

• Concrete Contract Template: Payment, Cracking and Delay Terms With AI — Put delay and weather terms into the contract once the price is accepted.
• Concrete Estimate Template: Volume, Pump and Finishing Lines With AI — Work out the order volume and pump hours this schedule relies on.
• AI Tools for Home Builders: Build Schedules, Delays and Sub Chasing — Build schedules, delays and subcontractor chasing across a whole house job.
• How to Set Up Claude Projects as a Shared Team Assistant — Keep your weather rules in a shared Claude Project for the whole crew.
• How to Build Gemini Gems for Repeat Business Tasks — The Gemini route for saved prompts, now moving to skills.
• Asphalt Paving Proposal Template: Scope, Exclusions and Weather Terms — How another outdoor trade writes weather terms into a proposal.
• AI Scheduling for Small Builders: Subcontractors and Deliveries — Subcontractor and delivery scheduling across a wider building job.
• AI Software for Construction Scheduling: Programme Updates After Delays — One small job walked through a late delivery and two weather days: what the assistant does, what you calculate yourself, and what you tell each sub.
• How to Bid Concrete Jobs: Quantities, Pump Costs and AI Checks — A concrete bid built line by line: volume, waste, pump, labour, finishing and washout, with AI prompts that recheck your arithmetic and catch missing lines.
• Tree Service Scheduling Software: Can AI Rebook Weather Delays? — Weather-hold triggers, hazard-job priority, rebooking texts and a crew reshuffle, written as rules an AI can follow and a person can overrule.
• AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.

Sources: NRMCA Concrete in Practice sheets on cold-weather concreting (CIP 27), hot-weather concreting (CIP 12), plastic shrinkage cracking (CIP 5), joints (CIP 6), curing (CIP 11), finishing flatwork (CIP 14), jobsite addition of water (CIP 26), dusting (CIP 1), discoloration (CIP 23) and cracking (CIP 4); OpenAI help page on scheduled tasks; Google Gemini Help page on scheduled actions; Open-Meteo pricing page.