Use this workflow when a supplier contract is due for renewal and you need one decision pack, not a collection of emails and spreadsheets. You will produce a weighted comparison matrix, a short recommendation, and a clear handover for the budget holder.
1. Set the decision question and deadline
Write the decision question before collecting evidence. Put it at the top of a working document:
Should we renew, renegotiate, re-tender, or exit Supplier A for [service] from [date]?
Then record:
- Decision owner: the budget holder who approves or rejects the recommendation.
- Contract owner: the person responsible for the supplier relationship.
- Deadline: the latest date that allows for notice, approval and onboarding if needed.
- Options: renewal as quoted, renewal with negotiated changes, alternative supplier, or internal replacement.
- Non-negotiables: requirements that rule an option out, such as required operating hours, site coverage, security requirements or integration support.
Do not start with a preferred supplier. A matrix is only useful if each option is tested against the same criteria.
Key point
Fix the options first
Compare named options against the same requirements. Do not let a strong recent anecdote become the decision criterion.
2. Assemble one evidence pack
Create a folder called Supplier renewal - [supplier] - [year]. Add the source files and give each one a clear date.
Use these inputs:
| Input | What to extract | Owner to confirm |
|---|---|---|
| Current contract and amendments | term, notice date, service scope, price mechanism, commitments, exit duties | Contract owner |
| Renewal quote and alternative quotes | total charge, assumptions, exclusions, validity period, implementation costs | Procurement or supplier manager |
| Service records | SLA performance, incidents, missed visits, response times, corrective actions | Service delivery lead |
| Finance records | invoices, credits, unplanned charges, disputed amounts | Finance partner |
| Stakeholder requirements | must-haves, operational pain points, expected demand changes | Service users |
| Risk and security evidence | outstanding audits, incidents, insurance or assurance requirements | Risk or security owner |
Keep facts separate from opinions. For example, six missed collection windows in Q2 is evidence. unreliable supplier is an interpretation that needs evidence behind it.
Watch out
Do not compare headline prices only
A quote may exclude transition work, volume changes, travel, equipment, support or taxes. Capture the stated assumptions beside every figure.
Before you upload documents or paste internal records into the model, follow your organisation's data handling rules. Available features and document handling are version-dependent, so check the xAI documentation overview for current guidance.
3. Build the criteria and weights
Ask the people affected by the service to agree the criteria before scoring suppliers. Limit the list to five to eight criteria. More than that usually creates false precision.
A practical starting matrix is:
| Criterion | Suggested weight | What a high score means |
|---|---|---|
| Service performance | 25% | Consistently meets agreed service levels with few serious incidents |
| Total operating cost | 20% | Clear, comparable cost with limited unplanned charges |
| Contract and commercial fit | 15% | Acceptable term, notice, price changes and flexibility |
| Operational fit | 15% | Meets coverage, capacity, systems and process needs |
| Risk and assurance | 15% | Evidence meets required controls and risks are manageable |
| Transition effort | 10% | Low disruption, realistic implementation plan and clear ownership |
Change the weights to reflect the decision. If a service failure would stop operations, increase service performance and risk. Record who agreed the weights and when. Make the weights add up to 100%.
Use a five-point scoring scale:
5: exceeds the requirement with strong evidence.4: meets the requirement with minor gaps.3: meets the minimum requirement, with conditions to manage.2: material gap or weak evidence.1: does not meet the requirement.0: fails a non-negotiable requirement.
For each option, calculate weighted score = score ÷ 5 × criterion weight. Add the weighted scores. Keep the underlying score and the evidence reference visible.
Check
Test the arithmetic
The weights must total 100%. A supplier with a non-negotiable failure should be marked as ineligible, even if its numerical total is high.
4. Use the model to structure, not decide
Give the model a controlled task. Paste the agreed criteria, weights and source extracts. Identify each source, for example Service report, April to June or Renewal quote, received 14 May.
Use this prompt:
Create a supplier renewal comparison matrix from the evidence below.
Use these criteria and weights: [paste them].
Score each option from 0 to 5. For every score, cite the source name and state the evidence in one sentence.
Do not infer missing contract terms, costs or service results. Mark missing evidence as “not evidenced”. Flag any non-negotiable failure separately.
Then provide:
1. a weighted score table;
2. the three strongest reasons for each option;
3. the key assumptions and evidence gaps;
4. a draft recommendation of no more than 180 words for the budget holder.
Evidence:
[paste labelled extracts]
Do not ask the model to read between the lines of contract language. It can organise clauses and identify questions, but the contract owner and appropriate qualified reviewer must confirm the meaning and obligations.
5. Check the matrix against the source documents
This is the step that makes the recommendation defensible. Review every score of 4, 5, 0 or 1, because these scores drive the result.
Look for these common errors:
| If you see this | Check this | Do this next |
|---|---|---|
| A low price wins clearly | Costs are on the same basis | Add excluded charges, transition costs and volume assumptions |
| The current supplier scores highly | Service reports cover the whole period | Check incident logs, complaints and credits as well as SLA summaries |
| Contract flexibility is scored well | The clause is explicit | Ask the contract owner to verify notice, renewal and change terms |
| An alternative scores well with little evidence | Evidence is being assumed | Reduce the score or mark it not evidenced |
Read the recommendation without the score total. If the written reasons do not support the proposed action, the weighting, scoring or evidence is wrong. Also run a simple sensitivity check: increase the most important criterion by 10 percentage points and reduce another by 10. If the preferred option changes, tell the budget holder that the decision depends on agreed priorities.
Check
Make uncertainty visible
A useful matrix shows what is known, what is assumed and what must be confirmed before approval. It does not hide gaps behind a single total.
6. Prepare the budget-holder pack
Send a short pack with four sections:
- Decision required: the action, supplier, term or next step, and approval deadline.
- Comparison matrix: options, criteria, weights, scores, evidence references and eligibility status.
- Recommendation: one preferred option, two or three reasons, and the trade-offs.
- Conditions and next actions: negotiation points, approvals, evidence gaps, owners and dates.
State the recommendation plainly. For example: Proceed with a conditional renewal, subject to a revised service-credit clause, confirmed volume pricing and closure of the open assurance action. This is a draft operational recommendation. The budget holder makes the approval decision.
When the workflow does not work
Stop and reset the comparison if the quotes cover different scopes, the contract documents conflict, or stakeholders cannot agree the non-negotiables. Do not force a total score from incomplete evidence.
Create an evidence-gap list instead. Assign each gap to an owner, set a response date, and hold the recommendation until the gap is closed or the decision owner explicitly accepts the risk. If time is short, present the options as approve, defer pending evidence, or run a short re-tender, with the operational consequence of each.