Skip to main content

Raw material suggestions

Where: AI Suggestions and Review AI on Raw materials Permission: Generate Raw Materials with AI

1. Why this one is different​

Every other Inventory generator reviews its output inline and forgets it. Raw materials keep a durable queue: every suggestion ever produced is stored with a status, and the queue survives the session.

StatusMeaning
PENDINGAwaiting a decision
APPROVEDTurned into a real raw material
IGNOREDRejected, but kept on record

On the reference tenant the queue holds 272 suggestions — 148 approved, 100 ignored, 1 pending.

Figure 1 — The Review AI badge.
The badge counts PENDING only

It reads 1 on a tenant holding 272 suggestions. It is a worklist counter, not a total — which is the right design, but people read it as "the AI has only ever suggested one thing".

2. Generating​

Figure 2 — The generator. Two inputs decide everything.
FieldNotes
Target languagesEnglish required; the rest are translated from it
Business typeOne line — Cloud Kitchen - Speciality Bengali and North Indian
Finished product listFree text. This is the input that matters

The explainer states the intent: "Identify standard raw materials required for your business. The AI will suggest materials, descriptions, and standard units in your requested languages."

List the dishes, not the cuisine

The generator works backwards from outputs to inputs. Naming Sukto, Mutton Biriyani, Moglai Paratha, Hilsa, Bhetki, Luchi, Radhaballavi produces the ingredients those dishes need. Writing "Indian restaurant" produces a generic pantry.

3. Reviewing​

Figure 3 — The review queue.

Review AI opens the queue as a drawer.

Each card shows the material, the SOURCE — the prompt it came from — and its proposed UOM.

Per card: Add to Inventory or Ignore.

Across the queue: APPROVE N NEW MATERIALS, IGNORE ALL, and APPROVE DUPLICATES.

The three bulk controls​

ControlDoes
APPROVE N NEW MATERIALSApproves everything the app considers new
IGNORE ALLMarks the whole pending set IGNORED. They stay on record
APPROVE DUPLICATESSeparate, and disabled when there are none
Duplicates are separated on purpose

The generator regularly proposes materials you already have. Rather than silently skipping them, the app splits them out so "approve everything new" is safe and approving a duplicate is a deliberate, second action.

4. What an approved suggestion is missing​

Figure 4 — UOM reads `---` when the generator could not resolve a unit.
Approved materials arrive incomplete

An approved suggestion becomes a raw material with:

  • current_stock = 0
  • last_purchase_price = 0
  • reorder_level = 0
  • sometimes no unit, shown as --- on the card

So it lands flagged LOW STOCK at ₹0.00, and it will pre-fill a goods receipt line at a unit cost of zero.

Approving 148 materials, as this tenant did, creates 148 records that all need a second pass.

Fix them on the Raw materials screen: set the unit first (it is required for BOMs and receipts), then the reorder level and the purchase price.

5. Re-analysing​

The refresh control beside AI Suggestions re-runs the analysis over the pending set rather than generating from scratch. Use it when you have changed the business description and want the pending queue re-evaluated.

6. Troubleshooting​

SymptomCause
Badge shows a small number, queue is hugeThe badge counts PENDING only
APPROVE DUPLICATES is greyed outNothing pending is a duplicate
Approved material has no unitThe generator could not resolve one — --- on the card. Set it by hand
Everything reads LOW STOCK after a bulk approveExpected; stock and price default to 0
Receipt unit costs pre-fill as 0Same cause — last_purchase_price was never set
An ignored suggestion reappearsIt should not; ignored rows keep their status. Check you are not re-generating the same prompt