Production and harvest planning is not a spreadsheet problem.
The spreadsheet is usually where the problem becomes visible. Demand changes. Weather shifts the harvest window. A field is not ready. A supplier is late. A packing line is constrained. Labor is short. A batch goes on hold. Inventory looks available until shelf life, grade, location, or customer commitment is considered.
Then the planner has to make a practical decision with imperfect information: harvest now or wait, produce this SKU or that one, allocate to this customer or another, hold inventory for a later order, switch source, add a shift, move a batch, or escalate the risk.
That is why production and harvest planning should be treated as an operating workflow, not just a forecast.
A forecast can suggest what might happen. A planning workflow decides what to do next, using the demand, capacity, quality, inventory, labor, logistics, and field data the business actually has.
The practical test: if the plan changes and the team has to ask five people which orders, fields, batches, lines, staff, or customers are affected, the planning workflow is not connected enough yet.
What the workflow is
A production and harvest planning workflow is the recurring process that turns demand signals, crop or raw-material availability, production capacity, inventory, labor, quality status, and logistics constraints into a plan the team can execute.
For a grower, processor, packer, food manufacturer, distributor, or vertically integrated food business, this workflow often sits between the commercial team and operations. Sales may know orders and forecast demand. Operations may know capacity and labor. Farm or procurement teams may know harvest readiness and inbound supply. Quality may know what is on hold. Warehouse and logistics may know what can actually ship.
The workflow is the mechanism that brings those pieces together often enough to make good decisions.
In practice, the workflow usually has four planning horizons:
- Seasonal or campaign planning: what volumes, crops, products, facilities, suppliers, and contracts are expected over the season.
- Weekly planning: what should be harvested, procured, produced, packed, held, allocated, or escalated this week.
- Daily planning: what the field, line, warehouse, quality, and logistics teams should execute today and tomorrow.
- Exception handling: what changes when demand, readiness, quality, yield, labor, equipment, or transport does not match the plan.
The first mistake is trying to solve all four horizons with one static file. They need to connect, but they do not all need the same view.
Where it usually breaks
The first break is disconnected demand and supply.
Orders, forecasts, promotional plans, customer commitments, and channel demand sit with sales or commercial teams. Field readiness, production capacity, packing constraints, and raw-material availability sit with operations. Inventory and shelf life sit somewhere else. The plan becomes a negotiation instead of a shared operating view.
The second break is planning at the wrong level of detail. A top-level forecast may be useful, but the team often needs the SKU, grade, pack size, field, lot, batch, customer, site, and delivery window. If the plan stays too high-level, the real constraints appear late.
The third break is stale inventory. Food and agriculture inventory is not just "units available." It can be blocked by shelf life, location, grade, quality status, customer allocation, certification, cold-chain history, or batch hold. A plan that treats all stock as interchangeable will create surprises.
The fourth break is no exception queue. Everyone knows the plan will change. But if exceptions are handled in calls, inboxes, and side spreadsheets, the team loses the history of what changed, who approved it, and what else was affected.
The fifth break is weak lot or batch continuity. If harvest lots, production batches, quality checks, warehouse movements, and shipments are linked only after the fact, planning becomes harder and traceability becomes slower.
What good looks like
A good production and harvest planning workflow does not pretend the world is stable.
It gives the team a rolling view of what is planned, what is constrained, what has changed, and what needs a decision.
The weekly planning view should show expected demand, available or expected supply, production capacity, packing capacity, inventory position, quality holds, labor constraints, logistics constraints, and the main exceptions. It should also show which assumptions are still uncertain.
The daily plan should be more operational. It should show what to harvest or receive, what to produce or pack, which lots or batches are involved, which inventory should be used, what is on hold, which customers are affected, and who owns each exception.
Most importantly, the workflow should make changes visible. If harvest volume changes, the team should see which production plan, inventory plan, allocation plan, and customer commitments are affected. If a production line is constrained, the team should see which orders or SKUs are at risk.
The minimum useful planning loop
- Demand view: orders, forecasts, promotions, customer commitments, channel demand, and priority rules.
- Supply view: crop readiness, inbound supply, raw materials, expected yield, supplier status, and quality caveats.
- Capacity view: field crew, production line, packing, labor, storage, cold-chain, and transport constraints.
- Inventory view: stock, location, shelf life, grade, hold status, committed quantity, and batch or lot links.
- Plan: what to harvest, procure, make, pack, allocate, hold, discount, substitute, or escalate.
- Exception queue: decisions needed, owner, reason, affected orders, affected batches, due time, and resolution.
- Review rhythm: daily execution review, weekly planning review, and post-week learning.
What data is needed
The exact data depends on the business. A fresh-produce packer, dairy processor, grain operator, seafood exporter, prepared-food manufacturer, and ag input distributor will not plan the same way.
But the same data families show up again and again.
| Planning area | Data you usually need | Common checks |
|---|---|---|
| Demand and commitments | Confirmed orders, forecasts, promotions, customer priority, delivery windows, substitutions, margin, channel demand. | Old forecasts, duplicated orders, missing delivery dates, demand not linked to SKU or pack size. |
| Crop or raw-material readiness | Field, block, variety, expected harvest window, maturity, yield estimate, quality grade, supplier availability, inbound date. | Readiness not updated, yield estimate not versioned, supplier date changed without downstream update. |
| Production and packing capacity | Line capacity, changeover time, labor, shift plan, pack format, equipment status, storage capacity, cold-chain constraints. | Capacity treated as fixed, downtime not reflected, changeovers ignored, labor not linked to the plan. |
| Inventory and shelf life | SKU, lot, batch, location, quantity, shelf life, grade, hold status, allocation, committed quantity, movement history. | Available stock includes holds, wrong location, expired or short-life stock, committed stock counted as free. |
| Quality and compliance | QA status, lab result, certificate, non-conformance, release decision, audit requirement, customer-specific requirement. | Quality holds not visible to planning, expired certificates, missing release timestamp, unclear approval owner. |
| Logistics and cold chain | Truck booking, route, temperature requirement, loading time, delivery window, warehouse slot, delay, carrier exception. | Transport changes not reflected in production plan, cold-chain evidence disconnected from lot or shipment. |
| Exceptions and decisions | Exception type, reason, affected order, affected batch, owner, deadline, decision, approval, resolution. | Decisions captured in messages but not in the system, no history of why the plan changed. |
The goal is not to collect everything. The goal is to collect the fields needed to make the next planning decision reliably.
What tools and systems are involved
This workflow often crosses ERP, MRP, farm management, production planning, warehouse management, quality management, lab systems, transport tools, order management, CRM, spreadsheets, and BI.
That sounds large, but the first useful version can be smaller. Many teams can start with a controlled planning layer that pulls or uploads the most important inputs: orders, forecasts, expected supply, inventory, holds, capacity, and exceptions.
The key is to make source and update logic visible. If field readiness is updated by the farm team, the plan should show when it was last updated. If inventory comes from WMS but QA hold status comes from a spreadsheet, the plan should show both. If a planner overrides the recommended allocation, that decision should be captured.
The planning tool itself may be a dashboard, spreadsheet-like internal tool, planning board, workflow queue, or BI view. The shape matters less than whether it connects the right data and supports the review rhythm.
Where AI can help
AI can help with planning, but it should not be framed as magic forecasting.
The useful AI layer is usually more practical. It can summarize changes since the last plan, classify exceptions, explain why a line or customer is at risk, compare demand and inventory, extract facts from supplier or quality documents, draft planning commentary, and help planners run "what changed?" reviews faster.
AI can also support scenario work. For example: if the harvest window moves two days, which orders are at risk? If available inventory is short-life, which customers can take it? If packing capacity is reduced, which SKUs should be prioritized based on commitment, shelf life, margin, and customer importance?
But AI should not silently choose the plan. Food and agriculture planning involves commercial commitments, food safety, quality risk, labor realities, customer relationships, and local operating knowledge. The system can suggest, summarize, and flag. Operators still decide.
Good AI use: "Summarize the changes since yesterday and highlight which orders, batches, or fields need review."
Bad AI use: "Let the model decide the harvest and allocation plan without showing the assumptions, constraints, and exceptions."
Where human review still matters
Human review matters where the data does not capture reality.
A crop may be technically ready but not commercially ideal. A batch may be within specification but not right for a specific customer. A forecast may look strong but be tied to a promotion that sales no longer trusts. A line may have nominal capacity but a known staffing problem. A shipment may be possible but too risky for the cold-chain window.
Those decisions need people who understand the work. The system should bring the decision to them with the right context, not bury the issue in another report.
Human review also matters for quality, compliance, and customer commitments. If the plan changes something that affects food safety, certification, traceability, customer specification, or a contractual promise, the workflow needs explicit approval.
What to fix first
Start with one planning rhythm that already hurts.
For many teams, that is the weekly production and harvest meeting. For others, it is the daily execution plan, the order allocation review, the fresh inventory review, or the late-change exception queue.
Do not start by trying to replace every planning file. Start by making one review more reliable.
A practical first build
- Choose the planning decision: harvest, produce, pack, allocate, substitute, hold, or escalate.
- List the inputs: demand, supply, capacity, inventory, quality, labor, logistics, and customer rules.
- Define the planning unit: SKU, lot, batch, field, line, pack size, customer, site, or delivery window.
- Build the current view: show the plan, constraints, assumptions, update times, and owners.
- Add exception logic: missing data, short stock, quality hold, capacity conflict, late delivery, short shelf life, demand change.
- Connect decisions: capture what changed, who approved it, and what downstream plan or customer was affected.
- Review weekly: compare plan vs actual, track rework, improve definitions, and add the next input only when needed.
Common mistakes
The first mistake is treating the forecast as the plan. Forecasts are inputs. The plan has to respect capacity, inventory, quality, labor, logistics, and customer priorities.
The second mistake is ignoring shelf life and hold status. Food inventory can look available while being practically unusable for the order in front of you.
The third mistake is planning too far from execution. If the people who know the field, line, warehouse, or customer constraints cannot update the plan, the plan gets stale quickly.
The fourth mistake is keeping exceptions outside the system. When every change is handled in chat or calls, the organization loses the reason for the change and repeats the same discussion next week.
The fifth mistake is adding AI before the planning units are defined. AI cannot help much if the team has not agreed what counts as a lot, batch, field, capacity unit, committed stock, or planning horizon.
The sixth mistake is building an elegant dashboard with no decision rights. A planning workflow needs owners and approvals, not just charts.
How Ubisar would approach it
Ubisar would start with the planning decision that creates the most visible rework or margin risk: what to harvest, what to produce, what to pack, how to allocate limited supply, or how to respond when the plan changes.
We would map the current planning path across commercial, operations, farm or supplier, production, inventory, quality, logistics, and leadership teams. Then we would identify the few fields and decisions that matter most to the first version.
From there, we would build the operating layer: planning data model, update rules, exception queue, review view, decision log, inventory and batch links, and AI support for summarizing changes, classifying exceptions, extracting document facts, and drafting review notes where it helps.
The goal is not to build a perfect planning suite. It is to make the first planning workflow faster, clearer, and more resilient, then expand into inventory, batch visibility, quality, procurement, demand allocation, traceability, and sustainability reporting.
This workflow connects closely to inventory and batch visibility, quality and compliance workflows, supplier and procurement operations, demand and allocation planning, and traceability reporting. For the broader operating model, see our food and agriculture workflow page or the AI, Data & Tech Implementation Retainer.