Allergen Sequencing
Allergen families dictate run order, and every validated cleanout costs line time you never get back. Semia, an AI employee for food and beverage production planning, sequences tomorrow's runs to minimize changeovers, reserves the cleanouts you actually need, and flags unsafe transitions in the draft before sign-off. A named human approves every plan before it reaches the floor.
You do not run sesame before the allergen-free line dries, and you do not slot dairy mid-shift on a whim. Run order goes from lowest allergen load to highest, and crossing a family boundary in the wrong direction triggers a full cleanout or a risk nobody signs up for. Get the sequence wrong on paper and the floor pays for it in hours.
A validated wet clean between allergen families is not a quick wipe-down, it is line time you bought and cannot run product on. Idle capacity costs a typical plant roughly $2M a year, and avoidable changeovers are one of the steadiest contributors. Every cleanout the sequence avoids goes straight back into production.
Undeclared allergens are among the most common reasons food gets pulled from shelves, and the run sequence is one of the controls that prevents it. If a nut product runs ahead of a plain SKU without the validated clean between them, the paperwork will not save you. The sequence is food safety, not just efficiency.
Your planner knows which line cannot go straight from sesame to plain, which cleanouts QA actually validated, and which order survives a short crew. None of it is written down in the ERP, so when the planner is out, the knowledge is out with them. That is the part traditional planning tools struggle to capture and keep current, and it is exactly what Semia learns. Bakeries feel this hardest; see /industries/bakery.
Semia reads orders, recipes and ingredient declarations, inventory and lots, machine status, and attendance from SAP, Odoo, Microsoft Dynamics, Excel, or Google Sheets. Access is read-only, nothing is migrated, and nothing gets installed on the line. Granting that access takes about 30 minutes.
Each morning's draft batches like with like and runs allergen families in an order that needs the fewest full cleanouts, while still respecting shelf life, FEFO, and finite line capacity. Fewer changeovers means more hours actually producing. The mechanics behind this are classic changeover reduction; see /glossary/changeover-smed-food.
Where the sequence does cross a family boundary, the draft holds the full validated cleanout time, not an optimistic guess that turns the schedule into fiction by Wednesday. Cleanout windows are visible in the plan, so the floor knows what is coming and QA knows the clean was scheduled, not squeezed.
If a draft would put a nut run ahead of an allergen-free run without the required clean, Semia flags it in plain language and explains its reasoning, with a full audit trail behind every call. A named planner approves the plan before anything reaches the floor. Nothing self-executes.
Traditional APS tools can absolutely produce allergen-aware schedules. The catch is the deployment: they tend to be heavy, slow to roll out, and dependent on clean structured data, and they drift out of sync with the undocumented rules in the planner's head. Semia works across the messy systems you already have and keeps learning as the rules change.
A copilot helps a human who still assembles the sequence by hand, one question at a time. Semia writes the finished draft plan every morning, sequenced for allergens, cleanouts, shelf life, and capacity, and hands it to your planner for approval. The human decision stays; the assembly work goes.
Everything Semia reads stays inside your company, and the changeover logic it captures from your planner belongs to you, not to us. Onboarding is shadowing, not a migration project: Semia sits next to the planner, asks questions, and is useful in about two weeks. Pricing is at /pricing, and you can book a demo at /demo.
Runs within the same allergen family can follow each other with only a light changeover, while crossing into a new family usually triggers a full validated cleanout. Sequencing a day from lowest allergen load to highest, and batching like products together, means the line crosses fewer family boundaries and needs fewer full cleans. Semia builds tomorrow's draft this way automatically, within shelf life and capacity limits, and a planner approves it before it runs.
Yes. Kosher, organic, and halal statuses behave like allergen families for scheduling purposes: they are hard constraints on what can follow what, and on which cleanouts or line separations are required in between. Semia learns those rules from your planner and your existing records during onboarding and applies them to every draft it writes.
Semia re-plans and shows where the rush order fits without violating allergen order: which cleanout it now requires, what gets pushed, and what the change costs in line time. If the only feasible slot forces an extra cleanout or an unsafe transition, it says so in plain language so the planner sees the tradeoff before deciding. The revised plan still goes to a named human for approval before anything changes on the floor.
No. Cleanout validation and verification stay with your QA and food safety process. Semia reserves the time your validated procedure requires between allergen families and flags any draft that skips one, but it never certifies a clean and never sends a plan to the floor without human sign-off.
It onboards itself by shadowing your planner: sitting next to them, watching real days, and asking why one product cannot follow another. It also reads ingredient declarations, recipes, and past schedules from the systems you already run, with read-only access. It becomes useful in about two weeks, and the captured logic belongs to you.
Granting read-only access to your existing systems takes about 30 minutes, with nothing migrated and nothing installed on production lines. Semia then shadows your planner and becomes useful in about two weeks. You can book a walkthrough at /demo.