Hiring technology consultants can make sense when a business decision depends on technical information you cannot yet verify. You might be deciding whether to keep a tool, connect two systems or build a missing component. The engagement should let you choose—and stop if the investment is not justified.
I would separate three approvals: accepting the assessment, selecting an option and authorising implementation. This guide covers purchasing and accepting those stages. For the role, capacity and authority of ongoing technical leadership, see the fractional CTO guide. The decision here is what evidence you need before turning a recommendation into an investment.
Start with a question the engagement can answer
“We want to modernise” gives you no clear acceptance criterion. “We need to decide whether our current system can handle order approvals without a separate manual register” defines an investigation. There is a workflow, a set of alternatives and a decision to make at the end.
In the brief, I would describe the process, its users, current behaviour and constraints that cannot be broken. I would add the available documents, access that can be provided and the person who will confirm that the process map reflects real work. If nobody measures delays or errors, ask for a bounded observation sample rather than an estimated saving presented as an achieved result.
I would not commission a broad assessment if the solution is already verified and only needs configuration. Nor would I start without access to the process and someone able to approve the decision. Scope the implementation or resolve those dependencies first.
Ask for alternatives before accepting a recommendation
For each material problem, I would ask the consultant to consider keeping the process, configuring the current tool, buying software, integrating systems and building the missing part. Not every option deserves equal investigation, but rejected options need a reason you can inspect.
- Configuration: demonstrate the required workflow on the available version and plan, including permissions and exceptions.
- Existing product: verify the requirement that determines the purchase, export options and migration work.
- Integration: establish which data can be exchanged, who owns it and which limitation requires a preliminary test.
- Custom development: identify the business rule the simpler options cannot support and who will maintain the new component.
- Wait: explain what information is missing or what would have to change to justify the investment.
If the advisor also proposes to deliver the implementation, ask them to make assumptions and rejected alternatives explicit. The recommendation should remain useful if your team or another supplier carries it out.
An acceptance matrix for the consulting engagement
Use this matrix when comparing proposals. Ask each supplier to mark items as included, optional, excluded or unresolved; assign an owner and cost; and describe the evidence they will leave. These are proposed commissioning criteria, not results of supplier testing.
| Stage | Verifiable deliverable | Condition for continuing |
|---|---|---|
| Assessment | Workflow map, findings with sources and dates, separate hypotheses and access limitations. | The process owner validates the workflow; each conclusion points to evidence or remains explicitly unresolved. |
| Alternatives | Essential requirements, options considered, reasons for rejection and confirmed or outstanding costs. | Each option's fit is visible at requirement level rather than hidden inside a combined score. |
| Decision | Recommended option, accepted risks, outstanding conditions and an approving owner. | The business chooses to proceed, request another check or close the engagement without building. |
| Bounded test | Question to resolve, test data, observed outcome and limitations. | The test answers the agreed uncertainty. A failed test triggers a decision review before more work is authorised. |
| Handover | An implementable first release, dependencies, exclusions, acceptance criteria and editable documentation. | The delivery owner can explain the scope and identify anything still preventing an estimate or start. |
Accepting a report does not establish that a solution works. Approving a limited test does not authorise a full migration. Separate approvals let you stop while retaining the useful work from the consulting engagement.
Example: an operations report nobody can reconcile
Hypothetical example: operations combines data from two applications to prepare a report. The figures disagree, so someone proposes a new dashboard. Before approving it, I would ask the consultant to follow one transaction from entry to reporting and compare identifiers, statuses and cutoff dates.
The assessment might find that one application records creation dates while the other uses confirmation dates. That finding needs anonymised examples and a rule validated by operations. If it cannot be verified, it remains a hypothesis and would not justify a build on its own.
The next decision is whether an agreed rule and report configuration are enough, an integration is required or a requirement calls for custom development. A bounded test could apply the rule to an agreed sample and identify the differences still unexplained. Acceptance would mean producing the expected classification for those cases and making exceptions visible, not promising to correct all the company's data.
If the test exposes incompatible identifiers, the next assignment may be to resolve that data issue before building the dashboard. That is a useful outcome: the assessment prevents approval of a delivery whose input requirement does not yet exist.
Write down what the engagement includes and excludes
I would specify the processes and systems included, planned interviews, samples to inspect and access required. Each finding would have a reference, potential consequence, degree of certainty and next check. Deliverables would include a decision log, an options assessment and a handover session with outstanding questions.
Implementation, cleanup of the entire data history, migrations, licence purchases, ongoing training and production support should be separate unless explicitly quoted. A technical assessment is not a legal audit, security certification or guarantee of commercial outcomes.
The business should retain editable documents and the information needed to continue. Clarify access to tests or prototypes and which parts can be reused. If a system could not be inspected, put that limitation alongside the conclusions it affects.
Compare assessment, implementation and maintenance costs
I would price the investigation, bounded test, handover and optional implementation separately. Effort depends on workflow count, data quality, people involved, access, third-party dependencies and depth of verification. I do not assign a rate to my services without understanding that scope.
For implementation, request separate items for configuration or development, migration, testing, deployment and training. For ongoing operations, identify licences, infrastructure, integration reviews, support and internal effort. Compare the same period and distinguish confirmed amounts, estimates and unresolved costs. The total-cost model helps organise them.
Agree who can authorise additional spending if a test exposes another dependency. A consulting recommendation should not automatically commit you to buying its implementation. Once you have made the decision, use the software quote guide to define the next delivery.
What I need to review your decision
Send the decision you need to make, the affected workflow, current tools and available information. Add who would accept the assessment, who could implement it and any deadline or budget constraint. A description and an example without personal data or credentials are enough to start.
This fits my technical consulting and leadership work. You can review how to work with me or share the project context. The first step is agreeing which question the assessment must answer and what evidence will let you close it.