How to Build an AI Upskilling Plan for Your Team

"We should do something about AI" is where most team upskilling starts and, without a plan, where it ends. Here's a framework that produces a decision rather than a discussion.

Step 1: Measure Before You Design

You cannot plan for a gap you haven't measured. Most leaders dramatically misjudge their team's baseline in both directions — assuming younger staff are fluent (often they use AI casually but unsystematically) and older staff are lost (often the most disciplined adopters).

Assess three things:

• Current usage. Who uses what, how often, for what. Ask directly and without judgment, or you'll get flattering answers.
• Capability level. Can they brief well? Do they verify? Have they built workflows?
• Appetite. Who's curious, who's indifferent, who's anxious. This determines sequencing more than skill does.

Our free AI Readiness Score handles the first two at individual level and gives you a team picture you can re-measure later.

Step 2: Decide What You're Actually Optimising For

Different goals produce completely different plans. Pick one primary:

• Time recovery — free capacity from routine work. Focus on workflow automation for high-volume tasks.
• Quality and consistency — raise the floor across the team. Focus on shared prompts, templates and standards.
• Capacity without headcount — do more with the same team. Focus on the specific bottleneck function.
• Retention and morale — people want to develop. Focus on breadth and visible investment.

Trying to optimise all four produces a plan that achieves none.

Step 3: Sequence by Willingness, Not Hierarchy

The instinct is to train everyone at once, or to start with managers. Both are usually wrong.

Better: start with a small group of willing people across different functions. Six to eight is ideal. They produce results, those results become internal proof, and their enthusiasm is contagious in a way that a mandate never is.

Then go wide, using their examples as the teaching material. "Here's what Nishani in finance built" lands better than any external case study.

Reluctant staff should be last, not first, and should be shown outcomes rather than tools. Someone anxious about AI is not persuaded by a feature tour; they're persuaded by a colleague saying it made their Thursday easier.

Step 4: Settle the Boring Things First

Before any training happens, decide and document:

• Which tools are approved, and who pays
• What data must never go into a public tool
• Whether AI-assisted work needs disclosure, and to whom
• Who to ask when someone's unsure

This takes an afternoon and prevents the most common failure, where trained people return to their desks and hit an access wall.

Step 5: Choose the Format Honestly

Match format to what your team can realistically sustain:

• Weekly sessions over 4–6 weeks — best retention and application. Requires protected time.
• Full-day intensive — good for kickoff energy, poor for retention alone. Needs follow-up.
• Self-paced with check-ins — cheapest, works only with genuinely self-directed people and a real accountability rhythm.
• Embedded in existing meetings — fifteen minutes of a weekly team meeting, one person sharing one thing. Slow but remarkably durable.

That last option is underrated and costs nothing.

Step 6: Build the Internal Owner

External training creates a spike. Internal ownership creates a slope.

Identify one person — not necessarily senior — who will own AI capability after the training. Give them explicit time for it, not just the expectation. Their job is maintaining the shared prompt library, running a monthly session, and being the person people ask.

This role is the highest-return investment in the whole plan. It's what our AI Champion programme is designed to produce.

Step 7: Measure at 90 Days

Set the measurement date at the start, and measure the things you named in step 2:

• Re-run the readiness assessment and compare
• Count workflows: how many people have at least two they use weekly
• Time recovered on the specific processes you targeted
• Tool usage data if your tools provide it

If nothing moved, the problem is almost always design — no owner, no protected time, or blocked tools — rather than the training content. Diagnose before you spend again.

A Realistic 90-Day Shape

1. Weeks 1–2: Assess baseline. Settle tools and data policy. Pick your goal.
2. Weeks 3–6: Pilot cohort of 6–8 willing people. Weekly sessions, real deliverables.
3. Week 7: Pilot group presents what they built to the wider team. This is your internal marketing.
4. Weeks 8–11: Wider rollout using pilot examples. Internal owner leads.
5. Week 12: Re-measure. Decide what to invest in next.

If you want help designing this for your organisation, our corporate AI training in Sri Lanka is built around exactly this sequence — and we'll tell you honestly if a full programme isn't what you need yet.