Do you need AI training, an audit or a build?
These solve different problems. Choose by the uncertainty you need to remove and the outcome you need to keep, rather than the name of the package.
Indexify practical guide. Updated 11 September 2026.
The short answer
Choose training when people need practice, an audit when the opportunity or constraints are unclear, and implementation when a defined workflow is ready to build and test. An existing pilot may need adoption support rather than more development.
Choose the work that resolves the real uncertainty
Choose a starting point
Training, audit or implementation?
Choose the statement that best matches the current problem. You can change direction as you learn more.
Your choices stay in this page. No sign-up required.
Start with practical team training
The tool is approved and the task is known, but people need a repeatable method. Practise prompting, source checking and safe handling using a real role-specific exercise.
Bring: participant roles, approved accounts and a suitable sample task.
Explore this route →Assumptions and how to use this tool
This is a decision aid, not a diagnosis. A build still needs approved access, ownership, acceptance tests and a support plan.
Compare what you should have afterwards
| Route | Best starting problem | Useful output | Decision afterwards |
|---|---|---|---|
| Practical training | People struggle to use an approved tool consistently. | Practised tasks, reusable prompts and a quality checklist. | Can the team perform and review the work? |
| Readiness audit | The opportunity, data or constraints are uncertain. | Evidence, prioritised use cases, gaps and a scoped recommendation. | Is a pilot worth doing, and under what conditions? |
| Focused implementation | A defined workflow needs to be built. | A tested workflow, operating instructions and acceptance evidence. | Does it meet the agreed criteria? |
| Adoption support | A pilot exists but is not becoming routine. | Role-specific practice, ownership, repeated measurement and support. | Can the team sustain useful, controlled use? |
Three situations that lead to different starting points
“Everyone has ChatGPT, but the drafts need too much fixing.”
Inspect a representative task and its source material. If accounts and data use are already approved, practical training can improve the brief, required output and review method. Assess whether participants can catch a planted mistake, not just whether they enjoyed the workshop.
Use ChatGPT training for that environment, or Claude training for teams working in Claude. The corporate AI training hub covers transferable methods across approved tools.
“We want to use AI, but there are ten possible projects.”
An AI readiness audit can compare candidates against value, inputs, risk and ownership. A useful conclusion may be to fix a process or measurement problem first. It should not assume that every opportunity deserves a build.
“One person spends every month rebuilding the same report.”
That may be a candidate for reporting automation. First confirm stable definitions, accessible sources and a known reviewer. If those are missing, scope the discovery needed to resolve them before quoting a production workflow.
Ask for boundaries in the proposal
Training does not automatically include integrations. An audit does not automatically include a build. A working pilot does not automatically include ongoing operation. Make these boundaries explicit, with named outputs, responsibilities and exclusions.
- What question will this stage answer?
- What will we receive and be able to reuse?
- Which access, examples and decisions must we supply?
- How will success and failure be tested?
- Who owns review, support and changes afterwards?
- What evidence would make us stop rather than expand?
Compare total scope, then cost
Use the AI consultancy cost guide and editable planner to separate a one-off engagement from software, support and internal review. Different deliverables cannot be compared fairly using a headline day rate alone.
A proportionate first engagement leaves you with a decision you can act on. If the next step depends on evidence you do not yet have, buy the work that produces that evidence first.
Questions people ask
Should we book AI training before an audit?
If the approved task and tools are clear and the gap is practical skill, training can be the right first step. If the opportunity, access or constraints are unclear, an audit can establish what is viable before choosing training or implementation.
Can training and implementation be combined?
Yes. A focused implementation can include role-specific practice and handover. The scope should distinguish building the workflow from teaching people to operate, review and maintain it.
What should we have before commissioning a build?
A defined problem, named owner, suitable inputs, approved access, baseline, acceptance tests and a support decision. Unresolved prerequisites should be part of discovery rather than hidden assumptions in an implementation quote.
A practical first conversation
Bring the work that is causing the problem.
Tell us what the team is trying to improve, what it has already tested and where confidence breaks down. We will suggest a proportionate next step.
Talk it through with Jon