Skip to content

Procurement decision model / 2.3

Procedure or policy? Cost the same purchase two ways.

Both alternatives work within the same boundary: company policy and, where it applies, Polish public procurement law (PZP). Compare time and cost and keep the result together with its assumptions and limits.

Start a comparisonCheck the model and evidence

The buyer after automation

Four questions about procurement work once systems and AI agents take over routine tasks: who does what, who sets the rules, procedure or policy, what to measure.

3Question 3

Procedure or policy

Workflow drawing

Two ways of working. The same boundaries.One path and a prescribed sequence of steps. Several routes branch and rejoin within the same boundaries. The same rules on both sides. Mandatory legal waits remain locked.TunnelFormal sequentialVerificationApprovalAuthorisationmandatory wait3FieldAdaptive compliant∂Φauthcompetitiondocsethicsthe system handlesthe standarda person judgesthe exceptionmandatory wait124

One path or many routes within the same policy. The system handles the standard; a person judges the exception. The same rules on both sides. Mandatory legal waits remain locked.

∂Φ = {auth, competition, ethics, docs}

A workflow schematic, not a time scale. Numbers link to the four questions below.

Model and shared boundaries

1Question 1

Where will the recovered time go?

Automation
  • Check budget, authorisations and contract compliance
  • Retrieve approved master data
  • Match order, receipt and invoice
  • Route a deviation to its owner
  • Keep an audit trail
Human
  • Non-standard need
  • Market analysis
  • Negotiating risk allocation
  • Exception decision
Implementation readinessArticle (in Polish)

2Question 2

Who changes the rules?

  1. 01Owner
  2. 02Rationale
  3. 03Effective date
  4. 04Event triggering a review

A change is justified before checking whom it helps.

Method and assumptions register

Questions 3 and 4

The result depends on the cost of a day's delay

Reference scenario

Delay

PLN 100,000

Central differences (PLN): formal sequential minus adaptive compliant.

Role effort
PLN 2,186
Non-labour costs
PLN 2,000
Delay
PLN 100,000
Contract
PLN 0
Total difference
PLN 104,186

Fleet. Central differences (PLN): formal sequential minus adaptive compliant. Role effort: PLN 2,186; Non-labour costs: PLN 2,000; Delay: PLN 100,000; Contract: PLN 0; Total difference: PLN 104,186.

The delay row is the elapsed-day difference multiplied by a declared daily cost that the model cannot verify. This is arithmetic, not an empirical finding.

Starting points are illustrative. Their TCO and amendment differences are zero. Informal process bypass remains non-monetised. Full ranges and assumptions are in the calculator.

Full record in the calculator

4Question 4

What to measure beyond cycle time

  • Valid offers and price against an ex-ante benchmark
  • Escalations to a human, with the reason
  • Share of agent decisions reversed or corrected by the buyer
  • Time from decision to record in the transactional system
10 measures in the article (in Polish)

Question 3

Where tunnel and field give the same result

Catalogue call-off
PLN 0
MRP release
PLN 0

Both controls have identical alternative maps. The model assigns no benefit to an alternative's name alone.

Reference scenarios

Questions 1 to 4

Questions for a procurement team discussion

  1. Where did the time recovered through automation go in your team, and how do you know?
  2. Which competency will buyers lack in three years if we change nothing in recruitment and performance assessment today?
  3. Who in your organisation can change a rule used by a system or agent? How long does that change take?
  4. Which rule in your workflow reproduces a historical sequence rather than a policy requirement?
  5. Where is a single-path procedure a better choice than a bounded field?
  6. Do you know the cost of a day's delay for your most important purchases? Who estimates it?
  7. Who signs off a decision prepared by an agent, and what must they see before doing so?
  8. Which measure beyond cycle time revealed a problem that cycle time did not?
Article (in Polish)

01 / Choose your goal

Compare a purchase or review the calculation method

I have a procurement case

Create a comparison

Move through the case, two workflows, costs and the result record.

Start the comparison

I am reviewing the model

Open the research centre

Review the formulas, assumptions, data sources and steps to reproduce the results.

Go to research

02 / What the result contains

You get the result, its assumptions and what the model does not price

The record contains the costs of both alternatives, the assumptions used and the factors the model leaves unpriced.

  1. 01Alternative names and workflow types
  2. 02Central result
  3. 03Declared scenario range
  4. 04Main cost drivers
  5. 05Assumptions and limitations

03 / Practitioner path

From case to record in four steps

Choose a similar purchase, adjust the activities and enter your costs. Then download the result with its assumptions.

  1. 01

    Case

    Choose a starting point and check the shared legal boundary.

  2. 02

    Workflows

    Name the alternatives and review two connected maps of work.

  3. 03

    Costs

    Enter the essential values and open detail only when it is needed.

  4. 04

    Record

    Read the difference, range, drivers and result boundary.

04 / How the comparison works

The model does not assume which alternative will be cheaper. Mandatory legal periods are locked on both sides, and the draft is saved in your browser only with your consent.

Do you have a procurement to compare?

Start from the closest scenario. You can return and revise the assumptions at any point.

Create a comparison record

Visitor measurement

With your consent, Google Analytics stores cookies and measures page visits. We do not send calculator values or custom comparison names. Declining does not limit the service. You can withdraw consent in the analytics settings at the bottom of the page.

Google privacy policy