Mike Cheong presenting a ChatGPT session to a seated audience of participants

Planning AI upskilling across departments: where to start

Short answer: Start with one or two departments where slow, document-heavy work is already visible to everyone, agree what good output looks like, and expand only after that works. An AI upskilling plan fails when it buys licences before it decides what people should do differently.

In Malaysian boardrooms this year, one question keeps reaching HR and L&D leaders: what is the company’s plan for AI? The pressure is real, and the instinct is to answer with a purchase. Buy seats, schedule a training day, announce the initiative, and the question feels answered for a quarter.

An AI upskilling plan that starts with software ends with shelfware.

The companies getting real value did something less exciting first. They looked at the work.

Why most AI upskilling plans stall

The common pattern is not laziness. It is sequence. A leadership offsite produces a mandate, procurement turns the mandate into licences, and L&D is told to train everyone.

Six months later, usage dashboards show a spike in week one and a slow decline from week two.

You cannot train your way out of a workflow problem.

The stall happens because nobody answered three plain questions before the purchase. Which departments feel the most pain from slow writing, slow analysis or slow replies? What does good output look like in each of them?

Who reviews quality after the training ends?

Without those answers, the AI upskilling plan becomes a tour of features, and feature tours do not survive contact with a busy Tuesday.

Pick the first departments by pain, not enthusiasm

The best first department for an AI upskilling plan is the one whose pain is already visible.

Enthusiasm is not a selection criterion. The team asking loudest for AI is often the team that wants to experiment, not the team with the most to gain. Look for signals in the work instead:

  • Document volume: the team produces proposals, reports, letters or summaries every day, and writing eats their week.
  • Visible backlog: month-end closes late, tender responses go out at midnight, inboxes carry a queue.
  • Rework: drafts bounce between people because quality is inconsistent, and seniors spend their time fixing juniors’ documents.

Two teams that use AI well beat twenty teams with logins.

Three signals for choosing the first department in an AI upskilling plan

In many Malaysian organisations the shortlist for an AI upskilling plan writes itself. HR drowns in letters, policies and first drafts. Finance loses days to month-end narratives and reconciliation notes.

Sales lives and dies by proposal turnaround. Operations carries SOPs nobody reads. Pick one, at most two, and be honest about why they were chosen.

Different departments need different first wins

A licence is not a skill, and a skill is not the same skill everywhere.

HR’s first win is usually speed on routine documents: offer letters, policy summaries, FAQ answers drawn from the employee handbook. Finance gets value from drafting variance explanations and turning raw numbers into a readable first narrative that an accountant then verifies line by line.

Sales wins on proposal structure and first drafts of standard sections, which frees time for the client-specific thinking that actually closes deals. Customer service gains from reply drafts that agents edit for tone and facts before sending.

Notice what these have in common. The first win is never “let the AI do the job.” It is a draft, a summary or a structure that a competent person then checks. That checking habit is the actual skill your plan needs to build.

Malaysian colleagues reviewing AI-drafted work together in an AI upskilling plan rollout

Set the output standard before the rollout

If nobody reviews the output, the tool gets the blame.

Before any department starts, its manager needs a working definition of acceptable AI-assisted output. Facts verified against source documents. Tone that sounds like the company, not a chatbot.

Numbers traced back to the spreadsheet they came from. Confidential client data kept out of tools the company has not approved.

This standard does not need to be a forty-page policy. One page per department, written by the manager who owns the quality, is enough to start. The page exists so that “the AI got it wrong” stops being an excuse, because the person’s job was always to catch it.

Managers set the ceiling on any AI upskilling plan.

When the manager reviews AI-assisted work with the same care as any other work, standards hold. When the manager waves it through, the whole department learns that wave-through is the real standard.

Grow by evidence, not by calendar

Sequence beats scale.

Expand the AI upskilling plan to the next department when the first one shows results you can point to: turnaround time on a document type, error rates in drafts, hours returned to client work. That evidence does two jobs. It justifies the next wave internally, and it teaches you what to repeat.

Start small enough to be honest about what worked.

A pilot that is too big to fail is also too big to learn from. Give the first department eight to twelve weeks, watch what actually happens, and let the second wave inherit the lessons instead of the slogans.

Common mistakes to avoid

  • Buying licences before use cases. Seats purchased ahead of purpose become a renewal line nobody can defend.
  • Measuring attendance instead of output. A full training room is not evidence that work improved.
  • One curriculum for every department. An AI upskilling plan that teaches HR, finance and sales the same half-day course serves none of them well.
  • Treating upskilling as an IT project. The tool rollout is the small part; the working-habit change is the real project, and it belongs to line managers.
  • Banning AI quietly while staff use it anyway. Ungoverned use does not disappear when ignored; it just stays invisible.
  • Announcing a transformation instead of fixing a workflow. Staff hear the slogan, then go back to the same slow Tuesday.

Attendance is not a result.

Five tools for an AI upskilling plan

Most teams already have access to three of these. Start with what you have before buying anything.

Tool Tier What it is good for
ChatGPT Free Drafting, summarising and reworking text; the free tier covers daily use for most staff
Microsoft Copilot Free Drafting and analysis inside the Microsoft apps many offices already run
Google Gemini Free Drafting, research summaries and brainstorming in the Google ecosystem
Claude Paid Long-document work: policy drafts, training materials and careful rewrites
Perplexity Paid Research with cited sources, handy for benchmarking and fact-checking

The tool matters less than the standard it is held to.

FAQ

Which department should start with AI upskilling first?

Start the AI upskilling plan where the pain is visible: high document volume, a backlog everyone can see, or heavy rework. In many Malaysian companies that is HR, finance or sales. Enthusiasm alone is a weak reason to choose a department.

How long should the first phase run before expanding?

Eight to twelve weeks is a practical window for the first phase of an AI upskilling plan. It is long enough to see real turnaround and quality changes, short enough to adjust before habits harden. Expand when you can point at evidence, not when the calendar says so.

Do we need to hire AI specialists to run this?

Not at the start of the AI upskilling plan. The first phase needs managers who can define good output and review work against it, plus ordinary staff training on the tools you already pay for. Specialists help later, when specific technical workflows justify them.

How do we stop staff pasting confidential data into AI tools?

Write a one-page rule per department that names what stays out, approve specific tools for company data, and make managers responsible for the standard. A ban without an approved alternative pushes usage underground instead of stopping it.

An AI upskilling plan on paper is not the hard part. The ones that work will look boring from the outside: one department, one workflow, one visible improvement at a time.

Mike Cheong helps Malaysian organisations build an AI upskilling plan that works in practice for HR, finance, sales and operations teams, built around real work instead of feature tours. To discuss what a sensible first phase looks like for your team, message him on WhatsApp at wa.me/60164207727.

Similar Posts

Leave a Reply

Your email address will not be published. Required fields are marked *