On Wednesday morning, a customer asks for pricing on 30 line items, needed by Friday at 5pm. The purchasing officer sends the RFQ to five suppliers. By Thursday afternoon, three have replied: one PDF by email, one price list pasted into the email body, one Excel file with columns that do not match the original request format. The purchasing officer opens all three, starts a new comparison spreadsheet, and begins copying.
Friday morning, a fourth supplier sends a WhatsApp photo of a printed quotation sheet. The fifth has not replied. At 11am, a revised quote arrives from the second supplier with different units and the same part numbers. The spreadsheet is now a v2. The first supplier’s pricing was in USD. The conversion cell has since been overwritten.
By Friday afternoon, the purchasing officer has spent most of the day on data entry. The comparison goes to the director at 4:30pm. The PO is raised at 6pm, just inside the customer’s deadline. Nobody has had time to negotiate. Nobody noticed that one supplier’s price on two line items was 18% above the catalogue target, because the column was not formatted the same way as the others.
This happens every week. At 25 cycles per month, it adds up to roughly 100 hours of data entry that the purchasing team is doing instead of procurement work.
The real problem is not the volume of quotes: it is how they arrive
Research on procurement in manufacturing consistently identifies the response assembly stage as the point where time is lost and errors accumulate. The cause is not volume. It is that every supplier communicates differently: PDFs with formal rows, prices pasted into email bodies, Excel files with non-matching columns, box-rate quotes when unit pricing was asked for, USD figures when the business operates in Ringgit. SME Corp Malaysia data shows that 97.4% of Malaysian businesses are SMEs, and the overwhelming majority of B2B procurement relationships run on email with no standardised document exchange.
The comparison spreadsheet is where all of this variation is supposed to be resolved. The purchasing officer reads each quote, interprets ambiguous descriptions, normalises units, converts currency, and manually keys every line. A mis-keyed value or a copy-paste error propagates invisibly until someone spots it, which is often after the PO has already gone out.
Part number mismatches compound the problem. Suppliers use trade names, brand abbreviations, or internal codes that differ from the buyer’s catalogue. Recognising that two differently-labelled lines refer to the same part is tacit knowledge that produces errors under time pressure.
Manual quote handling creates margin damage the PO cycle never recovers
Currency and unit transcription errors are the most direct source of margin damage. A USD value keyed without conversion, or a box price treated as a unit price, goes straight through to the PO if no one catches it. When a supplier revises their quote mid-cycle, the comparison built on the original figures becomes invalid and there is no authoritative version, just v2 in someone’s inbox with no record of what changed.
The audit trail gap moves this from an operational inconvenience to a compliance exposure. In August 2025, the Malaysian Parliament passed the Government Procurement Bill. Analysis of the legislation indicates it comes into force in 2026 and requires auditable procurement records from private sector suppliers seeking government contracts. A spreadsheet where every line was typed by hand, with no record of which version of which supplier’s quote each figure came from, does not meet that standard.
There is also the negotiation time that never existed. By the time the comparison sheet is assembled, the deadline is close or past. Nobody goes back to the second-cheapest supplier to ask whether there is room on the high-margin items. Two days of data entry, and the only output is a PO that went out on time.
Automation removes the data entry; it does not fix a stale catalogue or replace buyer judgement
Automation does not fix a stale parts catalogue. If the internal SKU list is incomplete or inconsistently described, the matching step will produce more unmatched flags than useful comparisons. No AI call compensates for a reference dataset that is not maintained.
Automation does not change supplier behaviour. Suppliers will still quote in non-standard units and send handwritten sheets. What it does is make handling that variation consistent rather than dependent on whoever is at the desk that day.
Automation does not replace buyer judgement. The comparison sheet still requires a human to review it, resolve flagged lines, and approve the PO. What changes is the two to three hours of data entry that currently precede that decision, and the audit trail that currently does not exist.
A free n8n template covers the core loop without a full procurement system
Cadence Innovations has published a free workflow template that covers the core RFQ automation loop, built on n8n, an open-source automation platform.
What it does. The workflow watches a dedicated email inbox. When a supplier reply arrives, it reads the email and any attachments, then passes the content to an AI model that extracts each line item as a structured record: part description, quantity, unit, price, and currency. Because supplier formats vary significantly, the extraction does not rely on fixed templates; the AI reads each document on its own terms. A second AI pass then matches each extracted line to the distributor’s internal SKU catalogue, assigning a confidence score. Lines below the threshold are flagged for review rather than passed through automatically. The workflow fetches live exchange rates, converts foreign currency to Ringgit, standardises packaging units, and writes the completed comparison to a shared Google Sheet. Each row shows the supplier’s price alongside the catalogue target price and last-paid price. Once all expected replies are in, the purchasing officer receives a notification with the count of flagged lines and a direct link to the sheet. Every cycle generates a timestamped audit log entry: lines parsed, matched, unmatched, and any warnings.
How it can be extended. WhatsApp quotes (voice notes, photos of handwritten sheets, or messages sent directly) can join the same pipeline once the WhatsApp Business API is connected; this is a Phase 2 addition. Scanned PDFs without a selectable text layer need an OCR step before extraction, which adds per-page cost and some latency. An ERP write-back step posting the approved PO line to a system such as SAP Business One or AutoCount removes the last remaining manual step. Gmail notifications can be swapped for Microsoft Teams alerts if the business runs on M365, and Google Sheets can be replaced with SharePoint Lists. The confidence threshold and over-target tolerance are both single configuration values that need calibration during the pilot, not code changes. For data residency requirements, the AI calls can be routed through AWS Bedrock in the ap-southeast-1 region.
The value is in the data that accumulates, not just the time saved
The first-month gain is time recovered. By month three, something more useful appears: a consistent, queryable record of every RFQ cycle, supplier response, and line item. At six months, the comparison history shows price drift: which suppliers have been creeping up on specific SKUs, which categories consistently arrive over target, which suppliers recover margin on low-velocity parts. This analysis does not exist for most distributors because the data sits in spreadsheets never designed to be aggregated.
At twelve months, supplier performance is measurable: average response time, missed-deadline rate, match accuracy on catalogue descriptions. The audit log that started as a compliance requirement becomes the evidence base for the next contract negotiation. A supplier over-target on 15% of line items for nine months is a different conversation when there is a timestamped record behind it.
PwC’s 2024 Digital Procurement Survey across APAC found that the secondary benefit most often cited post-implementation is visibility into supplier pricing behaviour, not just the time saving.
A well-configured system answers five procurement questions on demand
The useful frame is not features but questions the system should answer on demand: Which RFQs are open, and which suppliers have not replied? How many lines are flagged and why? What is the average supplier response time over the past 90 days? Which SKUs consistently arrive above target, and from whom? What is the full audit trail for a specific PO if it is queried?
If answering any of those requires manually compiling a spreadsheet, the system is not doing the job.
The catalogue must be current before go-live. The confidence threshold and over-target tolerance need calibration from real pilot data, not a day-one guess. The purchasing officer’s role does not change: they review the comparison, resolve flagged lines, and approve the PO.
Most RFQ automation stalls at go-live for six avoidable reasons
Importing old spreadsheet data with inconsistent units and currencies. Clean and standardise historical data before importing, or start the live catalogue fresh and treat old records as archived reference.
Setting the over-target threshold before reading actual historical spreads. A 5% tolerance set before anyone knows the typical spread on certain categories is 12% generates too many flags to be useful. Run a few cycles first, then set the threshold.
Underestimating change management. The most common failure is the purchasing officer running the manual spreadsheet in parallel after go-live. A clear cutover date matters as much as the technical build.
Skipping the late-reply chaser. The workflow flags a missing reply; it cannot chase a silent supplier. A follow-up process for non-responsive suppliers needs to be in place from day one.
Building the catalogue retrospectively. If the internal SKU list does not exist in consistent digital form before implementation, the matching step produces low-confidence results throughout. The catalogue work comes first.
Treating go-live as the finish line. The confidence threshold, over-target tolerance, and flagged-line review process all need tuning after the first month. Assign ownership of recalibration before go-live.
Cadence Innovations starts with a catalogue audit, not a workflow build
The starting point is always discovery: a catalogue audit, a supplier channel mapping exercise to identify which formats and channels are in active use, and a review of any contract or compliance obligations that set the audit-trail requirements. The picture has to be clear before anything is built.
From discovery, Cadence Innovations scopes a proof of concept covering one inbox, one set of document formats, and one cohort of suppliers. This produces something the purchasing team can use and react to within four to six weeks, not a presentation. PoC engagements are priced at SGD 15,000 to 60,000 before grant co-funding. Malaysian SMEs can offset a significant portion of that cost through the MDEC MDAG-AI grant (up to 70% co-funding, capped at RM2 million) or, for smaller businesses, the Geran Digital PMKS Madani (up to 50% co-funding, capped at RM5,000). Full rollout follows once the PoC is validated: production build, full supplier cohort, procurement-team onboarding, and documented handover, with periodic recalibration after go-live.
The compliance pressure makes this urgent; the grant makes it affordable
The RFQ response problem is not a technology problem. It is a process problem: the absence of a consistent record from the moment supplier quotes arrive to the moment a PO is raised. Getting this right requires a clean catalogue, a process the purchasing team will follow, and tooling that enforces consistency. The data that accumulates from three months of structured quote capture is more useful than anything most businesses have had before: price drift visibility, supplier response benchmarks, and an audit log that satisfies the Government Procurement Bill.
Procurement teams face pressure from buyers who want faster turnaround and auditable records, and from finance teams who want margin defended. The gap between those expectations and what a spreadsheet-based process can deliver is only going one direction.