Programmes

How to read this page

Most people open a catalogue hoping it will decide for them. This page will not. It sets out five programmes that differ in depth, in the kind of work they assume, and in how much of your own material you will need to bring. The matching happens afterwards, in writing, once we have sat with you for twenty minutes.

Foundations assumes little prior study. The practicum assumes you already try prompts at work and want them to hold still from one week to the next. Operations is for people who process other people's requests. Data sense is for anyone about to put a model onto a spreadsheet without having looked hard at the spreadsheet. Workplace writing is for people whose output is mostly prose.

After the call we may recommend one of the programmes below, a closed group described on the teams page, or a short note saying you do not need a course yet. That last answer is a normal outcome. How we teach, once you are in the room, is set out on the approach page. Practical questions about certificates, missed sittings and group size live in the FAQ.

AI Foundations for Working Adults

This is the programme we recommend when someone is curious, slightly behind their colleagues, and tired of advice that assumes they already live inside these tools. We start from the work you already do: the messages, tables, notes and searches that fill a normal Tuesday. We do not start from a tour of products.

Each evening ends with something you can use the next morning. That might be a rewritten intake note, a prompt that survives a second try, or a short list of tasks you will keep doing yourself because a model would make them worse. You will practise checking names, figures and sources, because an answer that sounds finished is still only a draft.

If, after the advisory call, it is clear that two hours of tool setup would serve you better than five evenings, we will say so. Foundations is for people who need the room, the pace, and other adults asking the same awkward questions.

• Mapping a working week: which tasks repeat, and which are poor candidates for a model.
• What a public assistant is doing when it answers, in plain language.
• Writing a first stable prompt for a task you already own.
• Checking claims, numbers, names and invented sources.
• Deciding what you will keep doing yourself.
• A Monday artefact: one process rewritten and ready to try.

Prompting & Verification Practicum

Plenty of people can get a useful answer once. The difficulty is getting a useful answer next Thursday, with a slightly different file, under a little time pressure. This practicum is for that problem. We assume you have already tried public assistants at work and have been disappointed by how quickly a clever prompt goes stale.

The two Saturdays are working sessions, not lectures. You bring tasks from your own desk, stripped of anything confidential. We look at why a result fails to repeat, how to pin a prompt to a role and a format, and how to run a short check before anything is forwarded. The second Saturday is mostly review: you test the pattern on new material and we watch where it bends.

If you have never used these tools, start with Foundations. If your question is really about a whole operations team, the six-week programme or a closed group will fit better than two intense Saturdays.

• Anatomy of a prompt you can open again next month.
• Failure modes: drift, overconfidence, and sources that were never there.
• A verification order you can run in under five minutes.
• Building a small library for one role, not a pile of one-offs.
• Pair review of live work, with the confidential parts left at the office.

AI for Operations and Admin Teams

Operations work is full of requests that look similar until they are not. Teams often buy licences first and describe the task later. This programme reverses that order. We spend the early weeks on how work actually arrives, who touches it, and where a model would quietly invent a status, a date, or a person.

The hybrid shape exists because admin teams cannot all leave the floor on the same evening. Campus sittings are for practice with other people watching the same messy example. Remote sittings are for questions that came up when you tried the method on a live queue. We work on public or dummy data in class; your own files stay on your side of the door unless you have already stripped them.

A public cohort is mixed: people from different organisations, same kinds of queues. If you need the whole team in one room, with your own examples, use the teams page and we will scope a closed group after the same advisory call.

• Intake and triage of routine requests without losing the odd case.
• Tables, logs and status notes that a colleague can still trust.
• Templates that survive a staff change.
• When a messy process needs a process fix, not a model.
• Handover notes between people and tools.
• A six-week check: what you still use, and what you quietly dropped.

Data Sense Before Automation

A great deal of so-called AI work is still spreadsheet work. If the file is incomplete, duplicated, or labelled in three different ways, a model will not rescue it. It will sound confident about the mess. These four evenings are for people who have been asked to “put AI on the data” and would like to look at the data first.

We practise reading a working file the way an auditor might: what is a record, what is a comment, what is missing, and what would be unwise to paste into a public assistant. That last point is especially sharp when a sheet holds personal data. We talk in general terms about what not to upload; we do not give legal advice, and we will not sign off a system you build afterwards.

You should come with a description of one file you use (not the file itself, if it is sensitive). We will work on public examples in the room and help you plan the version you can safely try at your desk.

• What counts as a usable record in a working file.
• Messy spreadsheets before any model sees them.
• Asking a model to summarise without inventing rows or totals.
• Personal data and public assistants, in plain workplace language.
• A before-and-after of one admin dataset you actually own.

Workplace Writing with Assistants

Much of the useful work these tools can do is still writing: a first pass at a brief, a cleaner email, notes from a meeting you recorded with permission. The risk is a page that no longer sounds like you, or a sentence that cites a paper that does not exist. This short programme stays with that problem and does not wander into model-building.

We draft on public or dummy material in class. You will practise giving the assistant a role, a reader, and a length, then taking the draft back with a pen. Tone, structure and source trails get as much time as the prompt. If your writing is bound by a house style or a public-facing clearance process, bring that constraint; we will treat it as part of the task, not as an obstacle to be talked around.

Three evenings is enough if writing is the main use you have in mind. If you also own queues, logs and spreadsheets, Foundations or Operations will cover more of the week. We can sort that on the advisory call.

• Briefs, emails and meeting notes as three different jobs.
• Giving a role, a reader and a length before you ask for prose.
• Reading a draft for tone that still sounds like you.
• Citations, source trails, and sentences that should not be trusted.
• A checklist you can share with a colleague who was not in the room.

Sittings and locations

Public programmes sit on weekday evenings and on Saturday mornings. Evening sessions run 18:45–21:15 so that people can come from work without turning the night into a second shift. Saturday practicum sittings run 09:30–14:00. We keep a short break in every sitting; the work is concentrated, and tired people write worse prompts.

The campus room is in one-north, at 1 Fusionopolis Place, #08-12, Galaxis. Closed groups for organisations can use that room or a room you provide, provided it has a table, a screen everyone can see, and enough sockets. Delivery details for those groups are on the teams page.

Cohorts stay small, usually eight to fourteen people, so that we can look at actual work rather than a slide deck. If a date does not fit, say so on the call. We would rather move you to the next sitting than fill a chair with someone who cannot attend.

Fees and how they are quoted

We quote after the advisory call. A seat in a public evening cohort is a different piece of work from a closed group of twelve in your office, and it is not useful to pretend those two things share a price. The written read-back that follows the call includes a figure, the format it belongs to, and what is in and out of that figure.

We do not publish a menu of subsidies, grants or government schemes on this site, and we will not invent one for you. If your organisation has its own training budget or an internal approval path, bring that to the conversation and we will write the quote in a form your finance colleagues can read.

The first sitting of a public programme can be treated as a trial. If the format is wrong, you may leave without paying the remainder. That offer exists because a catalogue page is a poor substitute for being in the room. Payment terms for closed groups are set in the scoping note, not here.

What we ask from you

Come having listed three tasks that already repeat in your week. They do not need to be impressive. They need to be real. That list is the material we work from; without it we would be teaching someone else's job.

Bring a laptop you can type on comfortably. Tablets are a poor substitute for this kind of drafting. Before the first sitting we name the public assistant we will use and ask you to have an account ready. We teach on generally available tools, not on a private stack, and we will not ask you to install anything unusual.

Do not bring confidential files, personal data, or anything your organisation would not want in a public model. We will give you dummy material and a method for stripping examples. If you are unsure, leave the file at work and describe the shape of it instead. More on how we handle this in class is on the approach page. If a constraint still feels unclear, ask it on the FAQ page or in the call.