A small culinary mushroom grower can hand AI four jobs today: turning voice notes into batch and flush records, calculating yield per kg of substrate, turning chefs' messages into a weekly order list, and drafting the weekly availability message. Expect $0 to $20 a month in tools. Contamination calls stay with you.
Nothing here needs a sensor network or farm-management software. It needs a phone, a spreadsheet and a general AI assistant, plus the habit of speaking to the phone at the end of each task. The jobs below suit culinary mushrooms only, such as oyster, lion's mane and shiitake sold fresh to kitchens and market customers, and each is priced for one farm so you can see whether the time saved beats the cost.
The farm we will price: 120 blocks, six regular buyers
All figures are illustrative; swap in your own. Picture a one-person grower with a helper two days a week. The fruiting room holds about 120 blocks at a time, mostly oyster with some lion's mane. Each block is 2.5 kg of wet substrate (about 1 kg dry matter). A block gives roughly 0.55 kg in the first flush, 0.25 kg in the second, and is then spent. Six restaurants order weekly by text message, and a farmers' stall takes what is left.
The grower spends about three hours a week on paperwork: transcribing scribbled notes from the lab and fruiting room, working out what sold, and replying to chefs. The aim is to give back about half of that, without letting anything touch the decisions that need a trained eye. Here are the four jobs and their price tags, then the detail.
| Job | What AI does | Tool cost a month | Time saved a week (illustrative) |
|---|---|---|---|
| Batch and flush records | Turns spoken notes into rows in a sheet | $0 to $20 | About 60 minutes |
| Yield per kg of substrate | Writes the formulas, flags odd results | $0 (spreadsheet) | About 20 minutes |
| Weekly buyer order list | Reads chef texts, builds pick and pack list | Same plan as above | About 45 minutes |
| Availability message | Drafts the Monday message from the harvest plan | Same plan | About 15 minutes |
That adds up to about 2 hours 20 minutes a week on a farm whose owner values time at $22 an hour: roughly $220 a month against $0 to $20 of tools. The tool cost is not the point; the checking time is, and each job below includes it.
The Complete Excel Course with AI Integration
From formulas and pivot tables to Power Query, DAX, VBA and AI with Copilot, ChatGPT and Claude — hands-on from the first lesson.
By Sagnik Bhattacharya and Virginia Thorn, Primetime Emmy Award winner
Buy nowJob 1: batch and flush records from voice notes
A batch record answers: what went into this block, when, and what came out. A flush is one wave of mushrooms from a block, with a rest between waves. Good records let you see which recipe, which sterilising run and which spawn supplier gives the best yield, and they matter if a buyer or inspector ever asks where a product came from. Keep the rules you must follow where you sell in mind and check them with the relevant authority; this tutorial is about the tidying, not the compliance.
The weak point is capture. Writing in a damp fruiting room is miserable, so people forget. Speak instead. At the end of each task, hold the phone and say the facts:
Batch O-0412, oyster, blue strain. Inoculated today, 2.5 kilo blocks,
40 blocks, spawn rate ten percent, supplier B lot 7. Sterilised
yesterday in run three, 90 minutes at pressure. Moved to incubation room,
temperature 23.Transcribe it with your phone's dictation or the voice feature in an assistant, then paste it into a prompt that returns a row. A prompt that works:
Turn each voice note below into one row with these columns: batch_id,
species, strain, event (inoculated, moved, fruited, harvested, discarded),
date, block_count, block_weight_kg, spawn_rate_percent, spawn_supplier_lot,
sterilise_run, location, temperature_c, notes. Leave a column empty if the
note does not say. Do not guess any value. Return a table I can paste into
a spreadsheet.A realistic output (illustrative) fills the batch, species, count and spawn rate, leaves the date empty if you did not say "today", and puts "blue strain" in the strain column. What you check: the batch ID (a mis-heard "0412" for "0421" corrupts the record), the block count, and any date the assistant filled in. Reading ten rows takes two minutes, and it is the step that stops errors spreading into yield figures.
Photographs belong in the record too. Take one phone photo of each batch at harvest and file it with the batch ID in the file name. You are not asking AI to judge it; you are building evidence you can look back at when a batch behaves oddly.
What this costs
The free tiers of the main assistants may cope with a few dozen notes a week, though free plans carry usage limits, for example on messages with uploads. A paid individual plan such as ChatGPT Plus is $20 a month. ChatGPT Business is $20 per seat billed annually or $25 monthly, with a two-seat minimum. The spreadsheet can be Google Sheets or Excel. For a one-person farm, the free tier or one $20 plan is plenty. The tutorial on voice-note record apps for livestock shows what a dedicated app adds if you outgrow this.
Job 2: yield per kg of substrate, worked through
Yield per kg is the figure that tells you whether a recipe or a spawn lot is any good. Write down what you mean by it before calculating, because two definitions are in common use.
- Yield per kg of wet substrate: total fresh mushrooms divided by the block's weight as made. Easy to measure on a farm.
- Biological efficiency: in cultivation studies, the fresh weight of all flushes divided by the dry weight of the substrate, as a percentage. Published studies report very wide ranges depending on species, substrate and method: a 2025 review table of oyster mushrooms on straw-based mixes runs from about 14% to 54%, while a 2022 study reported about 166% on alfalfa pulp. Do not compare your wet-basis number with a figure from a study unless you convert it.
For our farm, one block weighs 2.5 kg wet with about 1 kg dry matter. Total harvest over two flushes: 0.55 + 0.25 = 0.80 kg.
| Measure | Sum | Result |
|---|---|---|
| Yield per kg wet substrate | 0.80 kg / 2.5 kg | 0.32 kg per kg (32%) |
| Biological efficiency (dry basis) | 0.80 kg / 1.0 kg | 80% |
| First flush share | 0.55 / 0.80 | 69% |
| Income per block at $9 per kg | 0.80 x $9 | $7.20 |
The last row is where a grower learns something. If a block costs $2.60 to make (substrate, spawn, bag, energy) and yields $7.20, the gross margin per block is $4.60 before labour and selling costs. A batch that gives 0.60 kg instead of 0.80 earns $5.40, and a margin of $2.80, which is about 40% less. A small yield gap is a large profit gap.
Ask the assistant to write the formulas, not do the sums:
My sheet has columns: batch_id, blocks, block_kg_wet, dry_kg_per_block,
flush1_kg, flush2_kg, flush3_kg, contaminated_blocks. Write formulas for
yield per kg wet, biological efficiency on a dry basis, and the
contamination rate (contaminated blocks divided by blocks). Also write a
conditional-format rule that highlights any batch more than 20 percent
below the average of the last six batches.Test the formulas on one batch you have already worked out by hand. If the sheet says 80% and your calculator says 80%, copy it down. A common failure: the assistant divides by total blocks including contaminated ones, so yields look worse than they were. Decide whether you want yield per block started or per block that fruited, and make the formula say which.
Job 3: the weekly buyer order list
Chefs order in the way chefs write: "Hi, same as last week but add some lion's mane if you have it, 2 for Friday". Six buyers, six styles. AI reads that and builds a clean list. Give it a short template so it knows your products and units:
Products: oyster blue (kg), oyster pink (kg), lion's mane (kg),
shiitake (kg). Pack sizes: 500 g trays. Delivery days: Tuesday and Friday.
Below are this week's messages. For each buyer, list product, quantity in
kg, delivery day and any condition (for example "only if firm"). Flag
anything ambiguous instead of guessing. Then total each product.A sample output (illustrative) lists buyer A: oyster blue 3 kg Tuesday; buyer B: lion's mane 2 kg Friday, flagged "unclear whether 'some' means 1 or 2 kg"; and a total line of oyster blue 11.5 kg, lion's mane 4 kg. What you fix: the flagged lines, by replying to the chef, and any total that does not add up. This is where a misread quantity matters, so read every line against the original text. The sibling piece on how restaurants order from suppliers shows what a chef expects from the other end.
Matching orders to what you will have
Next put orders beside expected harvest. The sheet from job 2 gives yield per block. If 40 blocks are in the first flush this week at 0.55 kg a block, you expect about 22 kg. Orders total 15.5 kg, which leaves 6.5 kg for the stall. Ask the assistant to flag any product where orders exceed expected harvest, then decide yourself who gets short-changed. That is a customer-relationship call, not a maths one.
Job 4: the Monday availability message
Once you know what will be ready, a short message to your buyers writes itself. Give the assistant the harvest plan and let it draft:
Write a short, friendly message to my restaurant buyers. This week's
availability: oyster blue about 12 kg, oyster pink about 4 kg, lion's mane
about 3 kg, shiitake none this week. Orders by Sunday 6 pm. Tuesday and
Friday delivery. Plain text, no emojis, under 70 words.Read it before sending, because it will happily promise "plenty of lion's mane" when you have 3 kg. Over-promising to a kitchen costs you the relationship, so the rule is simple: quantities come from your sheet, never from the assistant's phrasing. For more on weekly produce messages and boxes, the farm-shop tutorial on weekly boxes and pre-orders covers a similar rhythm.
What stays human: the contamination calls
The note on this title is plain: contamination decisions stay with you. Green mould, bacterial blotch, sour or fermented smells, slimy caps and odd colour are judgements made by someone who has seen hundreds of blocks and knows the difference between a normal bloom and a problem. A photo-based AI guess can be wrong in both directions, and an error means either binning good product or selling something you should not.
What AI can do safely is pattern-spotting in your own records. Suppose six blocks from batch O-0412 were discarded, and your sheet shows all six came from sterilise run three. Ask:
Here are my last 30 batches with sterilise_run, spawn_supplier_lot,
location, and contaminated_blocks. Which factor, if any, is most
associated with contamination? Show counts, say how confident the pattern
is given the small sample, and list what I should check physically.A useful reply (illustrative) says run three accounts for 6 of 9 contaminated blocks across the 30 batches, that the sample is small, and that the sterilising pressure log and the door seal are the first things to inspect. It does not diagnose. It points you at where to look, and you go and look. The same discipline applies to anything a buyer or inspector might ask about: keep the decision to harvest, sell or discard with a person who is accountable for it, and follow the food-safety rules that apply where you sell, checking with the relevant authority or adviser.
A week with the setup running, hour by hour
To make the routine concrete, here is a plausible Monday to Friday, with times:
- Monday, 8:00. The grower dictates the weekend's notes in the fruiting room (5 minutes), pastes them into the assistant (2 minutes) and checks the rows (3 minutes).
- Monday, 9:00. Chef texts arrive over the weekend; the assistant builds the order list. The grower reads it against the originals and replies to two flagged lines (10 minutes).
- Tuesday. Harvest and deliver. Weights go into the sheet from the scale. Photo of each tray's batch label is filed.
- Thursday, 16:00. The grower enters Friday's harvest, checks yield per block against the last six batches, and notes any block more than 20% below.
- Friday afternoon. A 10-minute review: contamination count for the week and anything odd to inspect on Monday.
The total active admin time is about 40 minutes of checking and speaking, against roughly three hours before. The saving is not magic; it is replacing typing with speaking and arithmetic with formulas, then spending the saved time looking at the blocks.
Where it goes wrong: a mis-heard batch code that spread
A pattern you may meet. The grower dictates "batch O-0412" and the transcription reads it as "O-0421". The assistant faithfully writes that into the sheet. A week later the yield for batch O-0421, which has not yet been inoculated, looks suspiciously low, and the real O-0412 shows no harvest at all. The grower spends an hour untangling it.
Prevent it by dictating the batch code character by character ("oh, zero, four, one, two"), by having the sheet reject any ID that does not exist in the batch list (a data-validation rule the assistant can write for you), and by reading the first column of every new row before you do anything else. Small habits, large saving.
Lot codes on trays: connecting a sale back to a batch
Records only pay off if a tray can be traced to a batch. Put a short lot code on every tray label, such as the harvest date plus the batch ID (for example 1105-O0412), and write it into the sheet beside the buyer and quantity. Then, if one chef reports tough stems or a short shelf life, you can find the batch in seconds, see how it was grown and check whether other buyers received the same lot.
AI helps with the dull part: generating the weekly set of label texts. Give the assistant the harvest list and ask for a table of tray label lines (product, weight, lot code, pack date) that you paste into your label printer's software. Check the codes against your sheet. A lot code that does not exist in the batch list is worse than none, because it looks like traceability and is not. Whatever labelling rules apply to fresh produce where you sell, confirm them with the relevant authority; the assistant is not a source for that.
Pricing a flush: what the records tell you about price
With a few months of data, your sheet can answer a question growers often guess at: what should each product sell for? Cost per block was $2.60 in our example. Add labour at a loaded $22 an hour. If one person handles 120 blocks in about 14 hours of weekly work, that is $308 of labour, or about $2.57 a block per cycle. So full cost per block is about $5.17, and 0.80 kg of mushrooms must sell for at least $6.46 a kg to break even, before selling costs and waste.
At $9 a kg the margin is about $2.03 a block after labour; at $12 it is $4.43. Ask the assistant to produce a small table of price against margin and against the share of the harvest you can sell, but remember that it has no knowledge of your buyers. The table is arithmetic. Whether a chef pays $12 is a conversation.
Four more growers, four different starting points
A grower who sells only at markets
No chef texts, so skip job 3. Use the sheet for yield and sell-through: what was left unsold each week by species. Ask the assistant for a four-week trend to decide how much lion's mane to grow, since that is the product that leaves waste when overproduced.
A grower with a kitchen-garden restaurant as the only buyer
One buyer means a standing order. The useful AI job is the week-ahead plan: given 60 blocks entering fruiting and the buyer's average weekly take, how many blocks to start each week so supply stays steady. Give the assistant your incubation times and let it build the schedule table; check it with a calendar.
A grower adding dried mushrooms
Drying adds a conversion: fresh to dried weight loss. Record it per batch. Ask the assistant for a sheet column that computes dried weight as a share of fresh weight, then use your own measured ratio to price dried packs. Do not use a ratio from a website; measure yours.
A grower with a part-time helper
Share a one-page task list generated from the batch sheet each Monday: what to inoculate, what to move, what to harvest. The piece on rotas and harvest messages for farm staff adds the rota side.
Costs, all in one place
For the farm above, the ledger is short. A free or $20-a-month assistant; a free spreadsheet; the phone you already own; a few hours of setup to build the batch sheet and the order template; and about 40 minutes a week of checking. Say setup takes three hours at $22, so $66 once, plus up to $20 a month in tools. Against about $220 a month of recovered time, the arithmetic is comfortable, and it stays comfortable even if you halve the saving. The risk to price is not the subscription, but acting on an unchecked row. If you want this mapped to your own farm and kit, that is the kind of job my AI implementation consultation covers, and the contact page is the way in. Keep the books connected too: yield and sales figures feed farm accounting, and the broader tutorial on farm records and paperwork covers the rest of the office.
Questions growers ask about AI on a mushroom farm
Can AI identify contamination on my blocks?
Treat it as unreliable. A photo-based guess about green mould or bacterial blotch can be wrong in either direction, and a wrong call affects food you sell. Use AI to log photos and spot patterns in your records, such as which sterilising run the bad blocks came from, and keep the decision to bin or harvest with you.
What is biological efficiency?
It is a common yield measure in cultivation studies: the fresh weight of mushrooms from all flushes divided by the dry weight of the substrate, shown as a percentage. Studies report very wide ranges, so compare yourself with your own past batches first. Always note whether a figure uses wet or dry substrate.
Do I need special software for this?
No. A phone voice note, a spreadsheet and a general AI assistant cover the four jobs here. Specialist grow-room sensors and dedicated farm software exist, but a small grower can start with free or $20-a-month tools and decide later whether anything more is justified.
Can AI take chef orders by itself?
It can read a messy text and turn it into a tidy order line, but you should read each order before it feeds your harvest plan. A misread quantity of 2 kg as 20 kg is easy to catch if you look, and costly if you do not.
Is my production data safe to paste into an AI tool?
Batch and yield figures are low risk, though buyer names and prices are commercial information. Use a paid plan whose privacy page says it does not train on your content by default, keep chef contact details out of prompts, and check the settings before you start.
Further reads
- AI Assistant for Farmers: ChatGPT, Claude or a Farm App? — Choose between a general assistant and a farm app.
- AI Chatbot for Farmers: Answers From Your Own Farm Notes and SOPs — Ask questions of your own grow notes and routines.
- Free AI for Agriculture: What a Small Farm Gets Without Paying — What a small farm gets free and where the limits bite.
- AI in Food Retail: Jobs a Deli or Farm Shop Can Hand Over First — Jobs a food seller can hand over first.
- Turning On-Site Voice Notes Into Job Sheets and Quotes With AI — The same voice-note method on a different job.
- AI Inventory Forecasting for Small Businesses: How It Works — Forecast demand once you have a few months of sales.
- Greenhouse Management Software: What AI Climate Alerts Cost and Add — What sensor-alert greenhouse software costs against a spreadsheet and chat-assistant routine, with a break-even sum for a small growing operation.
- Microgreens Business Plan: Tray Costs, Restaurant Orders and Waste With AI — A microgreens plan built from one tray: what it costs, what it yields, what a standing restaurant order is worth and how a waste log changes the sums.
- AI for a Soap Maker: Batch Records, Ingredient Labels and Listings — A batch record template, an ingredient-list check against your safety assessment and listing copy that avoids skin-benefit claims.
- AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.
Sources: Scientific Reports review of agri-food wastes as oyster mushroom substrates (2025) and a Foods study of alfalfa pulp as substrate (2022), for the biological efficiency definition and ranges; ChatGPT and ChatGPT Business pricing pages (checked 5 November 2026).