AI fluency training · Singapore

We teach your team to work with AI, using your own work.

We are a Singapore training collective. We run cohort programmes and in-house workshops where people bring the tasks already sitting on their desks. Participants leave with working prompts, checked outputs and a written routine the team can keep running without us.

See the programmes How we work

What we actually do

We work as a small teaching pair, not as a stage act. One of us leads the room; the other sits with people at their screens, unblocking the moment a prompt fails or a spreadsheet export refuses to cooperate. That ratio is how adult learners stay in the work instead of watching a demonstration they will not repeat on Monday.

Every session is built around tasks the participants brought that week: a customer reply that keeps going out with the wrong tone, a monthly commentary that eats two afternoons, a policy question that should be answerable from the handbook. We do not invent sample files to make the tools look tidy. If a task is a poor fit for a general assistant, we say so in the room and put it on the “do not automate” list.

When the programme ends we document what worked — the instruction patterns, the review checks, the places where a human still has to decide. The point is that your team can run the same routine the following month without booking us again. We are a training collective, not a retained operations desk, and the handover is written with that in mind.

Who joins us

Most groups come from operations, customer support, marketing, legal-ops, finance and HR. Typical size is eight to sixteen people who already share a queue, a document set or a reporting cycle. We also work with public-sector communications units and education groups across Singapore and the wider region. Companies of roughly twenty to six hundred people are the usual fit: large enough to have recurring work, small enough that a written routine can actually travel.

What you take away

You leave with a prompt set adapted to your own documents, not a generic pack copied from a vendor blog. You also leave with a review checklist the team agrees to use before anything is sent, filed or published, and a short internal guide written during the programme in your language. The thirty-day follow-up is part of the cohort and the sprint so we can see which parts of the routine stuck.

What we don’t do

We do not sell or resell tools, and we do not take commission from vendors. We do not build software, integrations or custom models for you. We will not run a session that has no named business task attached to it, because that is a product demo wearing a training badge. If you need implementation engineering, we will say so plainly and step aside.

Selected work

The write-ups below are anonymised. Figures describe programme parameters — group size, hours, number of workflows documented — not client-verified returns. Full case blocks live on the work page.

Regional customer support, logistics

Engagement type: Six-Week Cohort. Sector: logistics. Format: online, 12 people. What we produced: a 22-prompt library tied to their macro categories and a review checklist for tone and escalation notes.

The team already used a general assistant in pockets, which meant replies drifted in tone and sometimes invented policy. We spent the first week on a task audit, then practised grounded drafting against their macros. By handover they had a shared library and an explicit rule that no reply left the queue unreviewed.

Finance operations, manufacturing

Engagement type: Embedded Sprint. Sector: manufacturing. Format: on-site, one team of 7. What we produced: four documented workflows and one explicit “do not automate” list.

Monthly commentary and variance narratives were being written from spreadsheet exports by hand. We sat with the function for two weeks, rebuilt the drafting path with source checks, and marked the judgements that still belong to a person. The pack is theirs to maintain.

All case blocks

How a programme runs

The sequence is the same whether you book a clinic, a cohort or a sprint. Depth changes; the order does not. We would rather spend an extra hour on the task audit than discover in week three that half the group brought work that should never meet a general tool.

Scoping conversation. A thirty-minute call to hear the function, the current tools, the blockers and whether a hands-on format is even the right next step.
Task audit. We collect ten to twenty recurring pieces of work and sort them into three piles: safe to hand over with review, worth assisting but not delegating, and not suitable for AI at all.
Teaching sessions. Short explanations, then supervised practice on the files people actually use. Two facilitators in the room so nobody waits twenty minutes for help.
Supervised practice. Participants run the same task more than once, with a source check written into the steps, until the routine is boring in a useful way.
Handover pack. Adapted prompts, a review checklist, the team’s written routine, a list of tasks we recommend leaving alone, and a thirty-day follow-up session.

Questions we hear first

Do we need to buy a particular AI tool before we start?

No. We teach with whatever your organisation already licences: general assistants, document-grounded search, spreadsheet helpers and meeting tools. We do not resell software and we do not take referral arrangements from vendors. If a licence is missing for a task you care about, we say so in the scoping conversation rather than inventing a stack. Practical AI fluency for teams that already have work to do starts with the tools on the desk, not a shopping list.

How large should a group be for a hands-on programme?

Cohorts work best with eight to fourteen people who share similar work. Clinics can take up to sixteen. Hands-on sessions above twenty-five people stop being practice and become a lecture, which we will not run. Leadership briefings are the exception: they are ninety minutes of framing for managers, not supervised work at the keyboard. If two functions need different document sets, we split the groups rather than force a mixed room.

How do you handle confidential documents during training?

We start from your existing policy. Anything that must not leave the building stays out of general tools; we practise with redacted samples or with tools your security team has already approved. We do not retain client documents after an engagement. Material we bring into the room is our own. Verification — checking numbers, quotations and citations against source — is taught as a habit, not as a warning slide at the end of the day.