How to assess AI skills in an offshore team member

“Good with AI” has become an easy line to put on a CV. The useful question is what happens after the first answer appears. Can the person spot a mistake, protect information, finish the task and hand it over clearly?

For a business owner, the goal is not the most impressive prompt. It is work you can use without having to redo it yourself. That applies whether your team member is in the Philippines or sitting next to you.

Start with the role, then add the tools

A social media team member might use an approved AI tool to draft caption options from a supplied brief. An EA might structure meeting notes into actions. A video editor might use transcription to find useful moments in footage.

Those are different workflows. Someone who can generate captions has not automatically shown they can reconcile accounts, build automations or manage customer complaints. Assess the underlying role skill as well as the tool use.

List the tools your business actually uses, such as ChatGPT, Claude, Canva, Google Workspace or your CRM. Then decide which tasks are appropriate for assistance and which require your specialist judgement or approval.

Try this small hiring exercise

Use a fictional business brief and agree a small, paid assessment. This example suits content or admin support; adapt the output to the actual role.

1. Supply a short description of a fictional service, its audience and a few approved facts.
2. Ask for a customer email draft and a short internal handover note.
3. Allow the candidate to use an agreed AI tool, with the fictional inputs only.
4. Ask them to identify missing facts instead of filling gaps with guesses.
5. Have them show their draft, corrections and final output.
6. Change one requirement and ask how they would update the work.

Include a deliberate ambiguity, such as an unspecified delivery date. A useful answer should flag that the date needs confirmation. A confident invented date is a problem, even if the email reads beautifully.

Score what matters

Ask what the tool got wrong and what they changed. An honest explanation of a correction tells you more than a claim that AI did everything perfectly. Keep the same criteria for each candidate.

Set the data rules before the first live task

Decide which accounts and tools are approved, what information they may receive, who can access outputs and where work is stored. Account settings and provider terms matter; do not assume every product or subscription handles data the same way.

The OAIC recommends, as a matter of best practice, that organisations avoid entering personal information, especially sensitive information, into publicly available generative AI tools. Its guidance on commercially available AI products explains the privacy considerations. Use fictional sample data during assessment and get suitable advice for your business’s live information.

Do not treat removing a customer’s name as a guarantee that a document is safe to share. Addresses, account numbers, unusual circumstances and other details can still identify someone.

Make approval boundaries explicit

A draft is not permission to act. Define who approves customer messages, public content, financial changes and anything that spends money. Start new workflows with review before those actions happen.

For recurring work, keep a short record of the input, the finished output, the checks and the owner. Avoid turning this into a reporting job that takes longer than the original task. The point is that somebody can understand and verify what happened.

Improve one workflow before adding five more

Choose a repeated task, agree what good looks like and compare the results with your current process. Track usable output, corrections and review time. Fast drafts are not a win if you spend the afternoon repairing them.

When you brief Team Hatch on a role, include your tools and the work you expect them to support. We can make those requirements part of the hiring conversation. For a clear starting brief, use our first offshore hire guide.