Academy

Your team has AI tools. Is the next step training, access or process support?

Your company has paid for AI tools. A few people use them regularly, others avoid them and a manager still rewrites much of the output. It is tempting to book another workshop or add a more capable tool before identifying where the work breaks down.

Begin with an example the team can attempt today. Check whether people can access the right tool, whether the necessary information exists and whether anyone has defined an acceptable result. Then look at the skills needed to produce and review it.

Training can address a learning need. A missing permission, an unreliable source or an unowned process needs a different response, sometimes alongside training.

Observe one attempt before choosing the remedy

Ask someone to demonstrate a normal piece of work using an approved, non-sensitive example. Watch the whole task, including finding sources and checking the result. Avoid evaluating only the moment an AI answer appears.

Suppose a coordinator needs to prepare a weekly customer update. The difficulty might arise because they cannot open the approved order record. They might have access but find several versions with different dates. They might produce an accurate draft that waits for approval because nobody owns that decision.

Those situations call for different action. Only some are likely to improve through teaching the coordinator better instructions.

Write down the observed difficulty in ordinary language. “Cannot locate the current delivery date” gives the technical or process owner something to investigate. “The team needs AI adoption” leaves too much unresolved.

Separate the main constraints

The table below is a discussion aid. It is not a formal assessment or a diagnosis of your systems.

Observed symptom Check first Next conversation
A person cannot open the required tool or source Their account, approved access and relevant configuration The responsible administrator or technical support route
Different records give different answers Which source is authoritative, current and complete The information or process owner
A draft is prepared but nobody can approve it Who owns the decision and what they need to check The manager responsible for the workflow
People can access good sources but struggle to instruct, check or explain the output A representative attempt against an agreed standard A suitable training or practice discussion
The work needs systems to exchange information or act automatically Integration, permissions, failure handling and operating responsibility A separately scoped implementation discussion

Several constraints may appear together. A team can need both an access correction and practice. Agree who will resolve each part instead of asking one workshop to carry the whole project.

The OECD’s 2026 SME paper discusses skills and implementation conditions alongside tool uptake. Its findings are context for asking these questions, not proof that training alone will solve a particular company’s problem.

Recognise a learning need you can act on

A useful learning objective describes something a participant should be able to demonstrate. “Understand AI” is difficult to check. “Prepare a source-supported customer update and identify any missing commitment” is clearer.

Look for a specific gap. People may need to provide better context, separate instructions from source material, check factual details or explain why they accepted an answer. A manager may need to set the review standard before asking others to improve their output.

Use a task with enough information to complete it. If the source is deliberately incomplete, the exercise should include recognising the gap. That is a judgment skill worth practising. It should not be treated as a failure to make the AI answer longer.

Review the intended audience and prerequisites of the programme you are considering. A general business-user course and a technical data programme serve different needs. Neither should be selected simply because its title sounds more advanced.

Check whether the process is ready for practice

A good exercise needs an output, a usable source and a person able to judge the result. When one is missing, do a small amount of preparation before booking.

Choose an approved source version. Define what the output must contain and what it must avoid. Assign the person who will answer questions about the work. These decisions make the training requirement easier to explain without requiring a complete process redesign.

Keep operational deployment separate. A participant learning to draft a reply is different from a system sending replies to customers. The latter needs an agreed implementation scope, including what happens when information is missing or an action fails.

Our existing guide on moving an AI pilot into a reviewed workflow covers that implementation discussion. The question here is which people need to learn what, using the access and process available to them.

Compare another tool with better use of the current one

Before adding software, name the capability that the current setup cannot provide. Is it a genuine product limitation, an access restriction or something the team has not learnt to do? Ask the appropriate owner to verify that distinction.

For example, staff using Google Workspace may need an access check or help with an existing feature before considering another application. A frontline request process in Lark may need clearer ownership or onboarding. Those possibilities do not establish a fault or imply that either platform includes a training package.

A cloud infrastructure project, application build or ongoing support requirement also needs its own technical and commercial discussion. General AI training should not be used to imply that such work is included.

A supplier should be able to explain the scope you are buying. Ask what people will practise, what access they need and which parts of the business problem remain outside the booking.

Prepare a short gap statement

Use this structure for an internal discussion or a provider enquiry:

The people doing this work are [roles]. They currently use [approved tools and sources]. They need to produce [output]. In a representative attempt, the difficulty was [observed gap]. The person reviewing the result is [role]. We want to discuss [learning, access, process or implementation scope].

Use roles and sanitised descriptions in the initial enquiry. You do not need to share account access or private customer records to explain the problem.

This statement gives the conversation a useful direction. If you are unsure which category applies, describe the observed gap and ask for help identifying the next discussion. Avoid choosing a complex programme to compensate for an unclear requirement.

Choose the next step that fits the gap

When the gap is people’s ability to use and judge AI outputs, explore team training with MatrixC Academy. The learning-path comparison explains the existing AI and data routes so you can discuss a suitable starting point.

For teams whose main difficulty is reporting, the next article works through checking an AI-assisted management report. It shows why source checking and explaining the evidence belong in the learning discussion together.

seantay

Recent Posts

How to brief an AI training provider: a checklist for your team

A request for “AI training for our staff” leaves a provider guessing about the people,…

2 minutes ago

AI and data storytelling: check a management report before presenting the recommendation

An AI-generated report can read smoothly while giving the wrong impression of the figures. A…

2 minutes ago

AI training for Malaysian SMEs: choose the work your team should practise first

Image above: AI-generated illustration of a Malaysian team planning a learning task. A team can…

2 minutes ago

What is Qwen, and when should a Malaysian business consider it?

Assess Qwen for a defined business application. Understand model access, task fit, evaluation, costs and…

1 day ago

Qwen API or self-hosted model: which route fits your application?

Compare managed Qwen access and selected model weights. Check serving costs, model terms, control, failures…

1 day ago

What does Qwen really cost? A token and operating-cost worksheet

Estimate Qwen input and output charges, then add retries, review and operations. Use a dated…

1 day ago