THE PRACTICAL GUIDES

How to compare AI consulting quotes

Two quotes are comparable only when they cover the same work. A strategy workshop, a proof of concept and a maintained production system have different deliverables. Begin with a scope table rather than a headline price.

Separate the cost components

• Discovery: interviews, process mapping, data assessment and a written recommendation.
• Build: development, integration, testing and deployment.
• Adoption: training, documentation and rollout support.
• Operation: hosting, model usage, software licences, monitoring and support.
• Changes: what happens when the data, integrations or requirements change.

Ask each provider to identify which items are included, optional or excluded. Record currency, GST treatment, payment milestones and the assumptions behind any usage estimate in the written proposal.

A day rate is not a project budget

Mensor publishes a starting advisory day rate and separately scoped build work. That is useful information about an engagement model; it does not establish a typical market price or the cost of your project. The directory leaves unknown prices unconfirmed rather than estimating a range from unrelated services.

Ask what “finished” means

For a pilot, specify which workflow it will demonstrate and how results will be assessed. For a production build, clarify integrations, access controls, failure handling, user documentation and support. Ask whether source code, configuration, accounts and deployment instructions will be handed over, and whether a different supplier could maintain the result.

Compare with a shared worksheet

Use one row per deliverable and one column per proposal. Add rows for internal staff time, data preparation, recurring charges, support response and exclusions. Mark unanswered items as open questions. A fixed fee with an unclear scope can be harder to evaluate than a staged engagement with explicit decision points.

Set a decision point before expansion

Agree what evidence would justify continuing after discovery or a pilot. Avoid treating estimated time savings as realised savings. Record the original process, the proposed measurement and who will review the result. This is a practical procurement framework, not a statement of market rates.

Next: Plan a first AI pilot, or browse implementation providers.

Worked example: compare the first year, not just the build

The figures below are invented to demonstrate the method. They are not supplier quotes or Australian market rates. Both proposals are assumed to cover the same deliverable, usage level and 12 months of operation after launch, with all amounts in AUD excluding GST.

• Proposal A: $12,000 build, $300 a month operation, $2,000 training. First-year total: $12,000 + (12 × $300) + $2,000 = $17,600.
• Proposal B: $9,000 build, $650 a month operation, $1,000 training. First-year total: $9,000 + (12 × $650) + $1,000 = $17,800.

B has the lower build price but is $200 higher over this comparison period. That small difference does not establish which is better: support coverage, usage limits, internal staff time and exit costs still need to match. If monthly usage doubles, ask both suppliers to revise the estimate using the same new volume. Do not assume usage costs remain fixed.

A published price needs its unit

On 30 September 2026, Mensor's homepage displayed advisory days from AUD $2,000 and measurement audits from AUD $1,500, with builds scoped separately. Those are three distinct purchases. The page did not state GST treatment beside those prices. A day rate cannot be substituted for a complete project total. This is a checked example, not a market benchmark.

Copy this proposal worksheet

Download the blank proposal worksheet (plain text). Open it in your notes or document app and make one copy for each provider. It needs no account, and we do not receive your answers.

1. Outcome and scope: name the workflow, users, systems, included integrations and excluded cases.
2. Deliverables: list the files, working software, documentation and training you will receive. Name the person who accepts each item.
3. Evidence: record a relevant delivered project, the consultant's role, how the result was measured and whether a reference is available.
4. Costs: separate discovery, build, data preparation, training, monthly support, usage charges and optional changes. Confirm AUD, GST and the period covered.
5. Dependencies: list access, licences, data cleaning and staff time you must supply. Ask what happens if a dependency is delayed.
6. Operation: name the support contact, service hours, response targets, usage limits and owner of each account.
7. Handover and exit: record what can be exported, who holds source code and configuration, and the work needed to move suppliers.
8. Open questions: mark each answer as in the written proposal, verbally stated, or still unknown. Record the date and source.

Resolve the gaps before scoring

A missing answer is not a zero-cost item. Ask the provider to clarify it in writing. Decide your essential requirements before reading prices, then remove proposals that do not meet them or agree a revised scope. For the remaining proposals, compare the same operating period and volume. Keep any preference for a local workshop separate from evidence that the proposed team can deliver the work.

Do not award a numerical quality score merely for having more service badges. The directory records published capabilities, not audited delivery outcomes. Keep your comparison tied to the actual team, scope and evidence.

For phone reception, use the voice-agent commissioning guide to specify call permissions, human fallback, cost units and fictional acceptance scenarios.

Editorial guidance with linked sources. Reviewed 2026-09-30.

See the current directory snapshot →