If AI prepares the first draft, where will the next expert learn?

An experienced account manager can spot a weak proposal quickly. They notice the missing constraint, the condition that should not have been promised and the question the customer has not answered.

Now imagine giving a junior colleague an assistant that prepares the proposal. The document is complete, fluent and formatted. The junior can move through the task without performing the comparisons that once helped people learn it.

The draft may be useful. The learning path still needs attention.

A business that automates routine work must decide how less experienced employees will acquire the judgment needed to supervise that work later.

Identify the decisions hidden inside the old task

Take a task an expert performs well and ask them to work through an example aloud. Avoid asking for a general explanation of their expertise. Ask why they accept one condition, query another and stop at a particular missing fact.

In a proposal, the valuable decisions might concern delivery feasibility, service exclusions or compatibility with the customer's current setup. Formatting the document is a separate skill. Faster formatting does not remove the need to understand the commitment being made.

This exercise produces a more useful training plan than “learn to use the assistant”. It identifies what a person must recognise, what they may delegate and what evidence should make them pause.

Let learners form a view before seeing the answer

For selected training cases, ask the junior to identify the open questions and propose a next step before revealing the generated draft. Then compare both with an expert's review.

The aim is not to recreate every manual step forever. It is to preserve deliberate practice on decisions that matter.

Use a range of cases. Straightforward examples build fluency, while ambiguous ones show where information is insufficient. Include a well-written answer with a substantive flaw. Otherwise, learners can start confusing presentation quality with sound judgment.

Keep this work separate from unreviewed customer commitments. An exercise can use appropriately sanitised historical material or an expressly constructed scenario. Live decisions remain subject to the organisation's actual authority rules.

Ask for the reason behind a correction

When the expert edits a draft, capture the business reason where it is useful. “Changed the wording” teaches very little. “Removed the delivery promise because the installation slot is unconfirmed” explains what the learner should check next time.

An assistant can group recurring corrections and prepare exercises from them. A responsible person needs to review those exercises. Otherwise, the training material may reproduce an outdated rule or generalise an exception beyond its scope.

Be selective. Recording every keystroke creates noise and may raise privacy concerns without improving learning. Preserve decisions and evidence that help another person understand the work.

Assess supervision, including the ability to stop

A person who completes the tool's training module has demonstrated attendance or a particular exercise. That does not establish readiness for every live case.

Assess whether they can identify missing information, check the relevant source, explain a correction and recognise when a decision exceeds their authority. Ask what they would do if two sources disagree. “I would ask the model again” is not enough where the disagreement needs a business owner.

Increase responsibility gradually, using observed work and appropriate supervision. Expertise in the tool and expertise in the domain may develop at different speeds. The most technically comfortable employee may still need help with the commercial judgment.

Make teaching part of the expert's workload

The senior colleague needs time to review reasoning, not merely approve the final document while attending another meeting. If every minute saved is immediately converted into more throughput, the apprenticeship can disappear unnoticed.

Allocate a manageable set of training cases, a regular review and a clear progression of responsibilities. Check whether learners become more capable of explaining decisions without the assistant. Also check whether the assistant is removing clerical work that previously consumed the expert's teaching time.

This is a practical workforce decision. The business needs today's output and tomorrow's competent operators. Both should appear in the implementation plan.

Bring one role whose routine work is changing to a Strategy Session. We can map the decisions that still need practice and design a workflow that leaves room to learn them.