Procurement Automation

Procurement Automation: Which Workflow Should You Automate First?

Score purchasing workflows on volume, error cost, document regularity, data access and decision risk to pick the first worth automating.

By
PhaseVane
Published
Reading time
7 min
On This Page

Every purchasing team has five or six workflows that someone wishes were automatic: comparing supplier quotes, chasing certificates, checking acknowledgements, comparing freight, writing RFQs (requests for quotation), building the monthly spend report. Picking the wrong one first is expensive. It is usually the one that sounds most impressive rather than the one with the most retyping and the clearest rules.

This guide gives you a scoring sheet for choosing. Five criteria, a weighting, scoring anchors so two people score the same way, and four hard stops that override any score. It finishes with six workflows scored for a sample business, so you can see how the numbers fall.

In short

  • Score five things: volume, cost of errors, how regular the documents are, how easy the data is to reach, and how much risk sits on the output.
  • Weight volume and error cost highest. Together they decide whether there is anything to save.
  • Four hard stops override the score: too little volume, nobody to own the output, documents that cannot leave your network, or data that is already structured.
  • 70 or more is a first candidate; under 50, leave it manual.
  • A score is a hypothesis. Test the top workflow on your real documents before committing to anything.

The five criteria

Score each workflow from 1 to 5 on each criterion. Higher always means a better fit for automation.

Criterion Weight 1 3 5
Volume (documents a month) 25% Under 15 30 to 60 Over 100
Error cost (what a missed error costs) 25% An inconvenience Hundreds of dollars or a day's delay Thousands of dollars, a programme slip or a compliance breach
Document regularity 20% Mostly scans, handwriting or free text with no fixed fields Mixed formats but the same facts in each Same fields every time, text-based PDFs, spreadsheets or emails
Data access 20% Needs write access to the ERP or a supplier portal login Needs a read-only export from a system Only documents you already receive
Decision risk (what the output triggers) 10% Output commits money or sends a message directly Output feeds a decision with a short review Output only informs a person, who reviews before anything happens

In between, score 2 for 15 to 30 documents a month and 4 for 60 to 100. Use the same in-between logic for the other criteria.

Score = (25 × volume + 25 × error cost + 20 × regularity + 20 × data access + 10 × decision risk) ÷ 5. The result runs from 20 to 100.

Thresholds: 70 or more, a first candidate. 50 to 69, a later candidate, often after a first workflow proves the approach. Under 50, leave it manual or fix the process first.

Four hard stops

Any one of these means not yet, whatever the score:

  1. Fewer than about 15 documents a month. A good template is cheaper than any system.
  2. Nobody owns the output. A flag that nobody clears is worse than no flag.
  3. Documents cannot leave your network and there is no route to approval. Find out who signs that off before anything else.
  4. The data is already structured. If suppliers answer through an e-sourcing platform or send acknowledgements by EDI (electronic data interchange), the reading problem has been solved upstream.

Six workflows, scored

A 120-person electrical and mechanical contractor. About 15 RFQs a month drawing 60 to 70 supplier quotes; 25 active subcontractors generating 20 to 30 statements and certificates a month; about 80 purchase orders (POs) a month from an ERP that can export CSV; six import shipments a month with three forwarder quotes each (18 quotes); one monthly spend report built by hand.

Workflow Volume Error cost Regularity Data access Decision risk Score
Supplier quote comparison 4 4 3 5 4 (100 + 100 + 60 + 100 + 40) ÷ 5 = 80
Supplier document tracking 2 3 5 5 5 (50 + 75 + 100 + 100 + 50) ÷ 5 = 75
Freight quote comparison 2 3 3 5 4 (50 + 75 + 60 + 100 + 40) ÷ 5 = 65
PO acknowledgement checking 4 3 3 2 4 (100 + 75 + 60 + 40 + 40) ÷ 5 = 63
RFQ creation 2 2 2 4 3 (50 + 50 + 40 + 80 + 30) ÷ 5 = 50
Monthly spend reporting 1 2 2 1 3 (25 + 50 + 40 + 20 + 30) ÷ 5 = 33

How to read the result for this business:

  • Quote comparison (80) is the first candidate: high volume, real error cost (an expired price or a missed minimum order quantity), and it needs nothing but the quotes already arriving. See supplier quote comparison for which steps to automate and which to keep.
  • Document tracking (75) scores almost as well despite low volume, because the documents are regular and the rules are clear. For a head contractor with more subcontractors it could come first. See supplier document tracking.
  • Freight (65) is limited by volume: 18 forwarder quotes a month, only just above the hard stop. A free tool may be enough. See freight quote intake.
  • PO acknowledgement checking (63) has the volume but loses on data access, because it needs a reliable PO export. It is a natural second workflow once the export is in place. See PO acknowledgement checking.
  • RFQ creation (50) is a later module: once quotes are compared automatically, a structured RFQ that asks for the same fields every time makes the next comparison easier. PhaseVane plans RFQ Automation as a later step for this reason.
  • Spend reporting (33) stays manual for now, and one report a month trips the volume hard stop anyway. The data lives in several systems, the format changes, and the scope tends to grow.

Before you talk to anyone

Collect these for your top-scoring workflow. They turn a score into evidence, and any vendor worth talking to will ask for them:

  • 20 recent documents of the type, in the formats they really arrive in.
  • The requirement or PO they relate to, so the output can be checked.
  • A timing: how long the last three took, from first document to decision.
  • The last three mistakes that got through, and what each cost.
  • Who would own the output day to day, and who approves documents leaving the network.

Then put the time and error numbers into the time and error model. If the monthly benefit is small, the model has saved you a project.

Four scoring mistakes

  • Scoring the workflow you dislike most. Irritation is not volume. Count the documents for a normal month and a peak month.
  • Scoring error cost from memory. People remember the one disaster and forget the dozen small overpayments. Look at the last three mistakes that actually got through.
  • Giving data access a 5 because the documents are in email. If the check also needs a PO list, a subcontract register or a price file, score the hardest input, not the easiest.
  • Scoring alone. Ask the person who does the work and the person who signs the cost to score separately, then compare. A gap of two points or more on any criterion is the first thing to discuss.

Test the top score on real documents

A score tells you where to look. It does not tell you whether extraction works on your documents, how many fields a person would still correct, or how much time it would really save. That is what the Workflow Check measures.

The Workflow Check covers one workflow. You send 20 real documents and the requirement they answer. Within 5 working days PhaseVane returns the documents extracted and checked, every flag found, the fields that could not be read, the time each step took, and a plain view of whether automation is worth pursuing. The fee is fixed and stated in writing before you send any document. It is an assessment, not a commitment: the automations themselves are in development and not on sale, and if the numbers do not justify going further, the report says so.

Ask about a Workflow Check. Test your highest-scoring workflow on 20 of your real documents and get measured results instead of a score.

Not sure which workflow to test? Book a 20-minute workflow call and score them together.

Sources

No external data is used in this guide. The scoring method is PhaseVane practitioner judgement. Checked 9 October 2026.

Related Research

See all research notes

Topic: Procurement Automation

  1. Cost Driver Monitoring: Is This Supplier Price Increase Justified?

    Map each supplier cost element to a free public index, track it monthly, and test a price increase claim against the weighted move. Worked example.

    Procurement Automation
    Read More
  2. Freight Quote Comparison Automation: Can It Keep the Detail?

    How automated intake turns forwarder quotes from email, PDF and Excel into one comparison without dropping surcharges, TBA lines or validity.

    Procurement Automation
    Read More
  3. Human Review in Procurement Automation: What Stops a Misread Value?

    Ten controls that stop a wrong extracted value reaching a purchase order: source text checks, code-only maths, blocking flags and an approval log.

    Procurement Automation
    Read More