Tender pricing and supplier RFQs with AI: a BOM costing guide

What tender pricing is

Tender pricing is how a manufacturer works out what a bid will cost before it is submitted. The tender comes with a bill of materials — an Excel file that can run to hundreds of line items — and the bid has to state a price for each production quantity the customer asks about. For every line, someone has to find who can supply the part, ask for a quote, read what comes back and decide where to buy it. The result has to be competitive, because the bid is compared with others, and it has to be right, because the company is bound by it if it wins.

The scale is easy to underestimate. As an illustration: 200 line items, five candidate suppliers per line and three production scenarios add up to about 3,000 pricing decisions — all before a tender deadline.

What the work involves

• A BOM with hundreds of line items. Part numbers, descriptions, manufacturers and quantities per unit, in the customer's format, not yours.
• Several scenarios. The tender may ask for prices at different production quantities, so each line is costed more than once.
• Supplier search. For each line: which of your known suppliers can offer it, and what the distributor catalogs list.
• RFQ rounds by e-mail. Requests for quotation go out to several suppliers; replies come back on different days, and some lines need a second round.
• Replies in mixed formats. A quote can be a few lines in an e-mail body, an Excel sheet or a scanned PDF, each with its own layout.
• MOQs, package quantities and price breaks. A minimum order quantity (MOQ), a reel or box size that sets the steps you can buy in, and unit prices that change at volume thresholds. Together they decide what a line really costs at a given quantity.
• Splitting lines. Sometimes the cheapest way to cover a quantity is to buy part of it from one supplier and the rest from another.
• Lead times. When each supplier can deliver, set against the delivery dates in the tender.
• An audit trail. Who chose which quote, when, and why — something a customer, an auditor or your own team may ask about later.

Why traditional methods miss it

Without a dedicated tool, a tender is priced with what is at hand: a master spreadsheet, an inbox and experienced people. Those people know the parts and the suppliers. The method is what lets errors through.

• Manual spreadsheets. One large workbook, copied per scenario, with formulas written under time pressure. A formula that points at the wrong row or a scenario column that was not updated produces a price that looks plausible and is wrong.
• Copy-paste from quotes. Prices are retyped from e-mails and PDFs into the sheet. A digit, a decimal separator or a price quoted per 100 pieces instead of per piece is easy to carry over unnoticed.
• Missed MOQ and package effects. The cheapest unit price is not the cheapest line if the MOQ or the reel size forces you to buy far more than you need — and the reverse holds at a higher quantity. Checking this by hand for every line and every scenario is hard to do thoroughly.
• Missed price breaks and splits. A volume price break just above the required quantity, or a split between two suppliers, can make a line cheaper. Finding these takes calculations nobody has time for line by line.
• RFQ e-mail by hand. Writing a request per supplier, keeping track of who was asked about what and matching each reply to its line takes time that is short before a deadline.
• Deadline pressure. Late replies arrive close to the submission date, and the last changes go into the sheet in a hurry.
• No record of decisions. The final sheet shows a price, not which quote it came from or why one supplier was chosen over another. When someone asks months later, the answer is in someone's memory or an old inbox.

How AI agents reading documents make it solvable

Primarch builds AI agents for work like this: software that does not just answer a question, but carries out a sequence of steps — finding suppliers, drafting requests, reading replies, computing a plan — and stops for a person where the decision belongs to a person. Tender pricing fits that shape well: it combines a great deal of reading and bookkeeping with a smaller number of real decisions.

• It reads documents the way a buyer does. A quote can arrive as an e-mail body, an Excel sheet or a scanned PDF. Models that read text and page images together can pick out the part number, the unit price, the price-break table, the MOQ and the package quantity, whatever the layout.
• It keeps each answer tied to its line. What a supplier quotes is attached to the BOM line it answers, so prices are not retyped from one file into another.
• It leaves the arithmetic to calculation. Reading is the model's job. Choosing the cheapest valid combination of MOQs, package quantities, price breaks and splits for each scenario is a calculation, and it is done as one.
• It stops for approval. An agent that sends e-mails and fills in prices needs a clear boundary: e-mails go out and prices are written only after a person approves them.
• It writes down why. Each decision can be recorded with who made it, when and why, so the bid can be explained later.

The useful question is not "how fast can we fill in the spreadsheet?" It is "for every line and every scenario, can we show which quote we used and why?"

An agent reads what suppliers send and computes from it. Whether a part is technically acceptable, which suppliers are approved for a program and the final bid price stay with your engineers and your commercial team.

What to look for in a solution

Whichever vendor you talk to, these are fair questions to ask about an AI tool for tender pricing:

• Takes your BOM as it is. Does it work from the tender's Excel file as you receive it, or does the BOM have to be reformatted first?
• Searches where you buy. Does it look at your own supplier list as well as distributor catalogs?
• Asks before it sends. Does any e-mail reach a supplier without a person approving it?
• Reads real replies. Can it read quotes in the formats your suppliers actually send — e-mail text, Excel and scanned PDF — and can you check what it read against the original?
• Handles the purchasing rules. Are MOQs, package quantities, volume price breaks and line splits part of the calculation, separately for each scenario?
• Explains each choice. For any line, can you see which quote was chosen and why it was cheaper than the alternatives?
• Leaves the price to people. Is any price written into the bid without approval?
• Keeps a record. Is there a who/when/why record of each decision that you could show to an auditor or a customer?
• Returns your own file. Do the results come back in the same Excel, or in a new format your team has to copy from?
• Covers what matters to your bid. If lead times, currencies or approved-supplier rules matter in your tenders, ask how each one is handled.
• Is proven on your tender. A trial on one of your own tenders, not a demo data set.

Example use cases

The walk-throughs below are illustrative examples of how a tender line gets priced. They are not customer cases, the prices are illustrative, and they contain no performance figures.

Example 1: a 200-line tender with three scenarios

1. 01

   Situation

   An electronics manufacturer receives a tender with a 200-line BOM in Excel. The bid needs prices for three production scenarios, and the deadline is close.

2. 02

   What goes in

   The tender's BOM Excel, as received. The line items are recognized from the file.

3. 03

   What the agent does

   It searches the company's supplier directory and the DigiKey and Mouser catalogs for each line, and drafts the RFQ e-mails.

4. 04

   What the engineer does

   Reviews the drafts and approves each send. As replies come in, they are read and attached to their lines.

5. 05

   The result

   A cheapest valid plan for each of the three scenarios, with MOQs and package quantities taken into account, proposed line by line for approval.

6. 06

   What comes back

   The prices approved by the engineer are written back into the same Excel, ready for the bid.

Example 2: the same line, a different choice per scenario

1. 01

   Situation

   One capacitor line needs 120 pieces in scenario A, 600 in scenario B and 2,400 in scenario C.

2. 02

   The quotes

   Supplier X sells cut tape at $0.40 per piece in any quantity. Supplier Y sells only full reels of 2,500 at $0.12 per piece, so $300 per reel.

3. 03

   The calculation

   Scenario A: $48 from X against $300 for a reel from Y. Scenario B: $240 from X against $300. Scenario C: $960 from X against $300 for one reel from Y — 100 pieces more than needed, and still the cheaper option.

4. 04

   The proposal

   X for scenarios A and B, Y for scenario C. The lowest unit price wins only where the reel size does not force too large a purchase.

5. 05

   The decision

   The engineer approves the line for each scenario; the decision is recorded with who, when and why.

Example 3: splitting a line to reach the cheapest valid plan

1. 01

   Situation

   A connector line needs 1,500 pieces in one scenario.

2. 02

   The quotes

   Supplier A sells in packs of 1,000 at $0.30 per piece. Supplier B has an MOQ of 100, sells in multiples of 100 and quotes $0.42 per piece below 500 and $0.36 from 500 pieces.

3. 03

   The calculation

   All from A: two packs, 2,000 pieces, $600. All from B: 1,500 pieces at $0.36, $540. Split: one pack of 1,000 from A for $300 and 500 from B at the $0.36 break for $180 — $480.

4. 04

   The proposal

   Split the line: 1,000 from A, 500 from B. The split lands exactly on B's price break.

5. 05

   The decision

   The engineer approves the split; both orders are tied to the same BOM line in the record.

Example 4: three reply formats, one record

1. 01

   Situation

   For a line of voltage regulators, three suppliers reply to the same RFQ.

2. 02

   The replies

   One writes the price and MOQ in the e-mail body. One attaches an Excel sheet with a price-break table. One sends a scanned PDF quote with the package quantity in a footnote.

3. 03

   What the agent does

   All three replies are read and attached to the line, and the cheapest valid option is proposed for each scenario.

4. 04

   The decision

   The engineer checks the proposal and approves it. Who approved it, when and why is recorded.

5. 05

   Later

   Months after the bid, a colleague asks why this line was priced with that supplier. The record answers it: which quote, which scenario, who approved it and when.

How Primarch's Tedarik AI helps

This section is about our product. Tedarik AI is an agentic product that turns a tender's bill of materials into the cheapest valid procurement plan, line by line — the tool proposes, the engineer approves. It was built on the real tender pricing process of a leading defense electronics manufacturer. The flow:

1. 01

   01 · Upload the list

   Drop the tender's BOM Excel; line items are recognized automatically.

2. 02

   02 · Find suppliers

   Your supplier directory and the DigiKey and Mouser catalogs are scanned.

3. 03

   03 · Send RFQs

   Quote request e-mails are drafted; every send goes out with your approval.

4. 04

   04 · Parse the replies

   Incoming quotes are read — e-mail, Excel or scanned PDF.

5. 05

   05 · Compute the plan

   The cheapest valid plan per scenario, with MOQ, package quantities and volume price breaks taken into account; when it pays off, a line is split in two.

6. 06

   06 · Approve

   Prices are written back into the same Excel; every row returns approved.

• No price is written automatically. No black box: the engineer approves, and every decision carries a who/when/why record — traceability ready for the audit reality of defense procurement.
• Your process stays the same. Whatever your tender Excel is, that is what goes in, and the same file comes back with prices written and every row approved.
• Reads the lead time each supplier quotes and takes it into account in the plan, so an offer that cannot arrive in time is not proposed.
• Brings quotes in different currencies into one currency so they can be compared line by line.
• Tracks which suppliers have replied to each RFQ and drafts reminders for those who have not, each sent with your approval.
• Suggests alternative or equivalent parts when a line cannot be sourced as specified, for the engineer to approve.

Putting it to work

The simplest start is a demo on one of your own tenders: we price it in front of you, from the BOM Excel you received to the approved file. After the order is placed, Teyit AI reads supplier invoices and matches them line by line with your MRP/ERP plan.