AI training without the slideware

Most "AI transformation" programmes start with a capability map. Start with a calendar instead.

Where do senior people still paste, summarise, reformat, and chase? That is the value chain. That is where a two-day AI-for-teams sprint pays for itself. If AI lives in a lab, it is a demo. If it lives in the Tuesday standup, it is an operating system.

This piece is an AI Training piece. Keep the slug if you want. Change the behaviour. Collective Campus does not teach a model zoo. We teach a team to take hours back from a workflow they already own, with guardrails a risk lead will sign.

The companies worth watching this year are not the ones with the loudest keynote. They are the ones staffing enablement. DBS in Singapore. HSBC in Hong Kong. Telstra in Australia. Canva, which already lives inside the tools it sells. And a pattern we now treat as a signal, not a press release: a new Chief AI Officer, CHRO, or Head of L&D who funds literacy in the first 30-120 days.

The appointment is the tell

Watch the job, not the white paper.

When a bank in Singapore or Hong Kong stands up a Head of AI Enablement, a CAIO with a training mandate, or an L&D lead with AI literacy in the first-year objectives, that is a programme. DBS has spent years making digital work a P&L conversation. The current chapter is making generative tools a desk conversation without turning every officer into a shadow IT shop. HSBC Hong Kong has the same physics: volume, regulation, and a workforce that will not be replaced by a chatbot demo in a town hall.

Telstra has run digital and AI programmes at a scale that fills conferences. The useful tell is not the model announcement. The useful tell is whether frontline and corporate teams get a sequenced enablement (guardrails, use cases, practice) or a portal and a hope. Canva is the product-company rhyme: AI is in the product, and the internal question is still "which workflows get hours back this month?"

In the US, UK, and Canada the same appointment pattern is firing. A new CAIO at a bank or a telco. A CHRO who just inherited "AI readiness." An L&D director whose budget suddenly has a line that says literacy. First 30-120 days, they will either fund a two-day sprint on a hated workflow or they will fund a capability map. We write for the first group.

Do not invent a quote from those people. You do not need one. The org chart is the quote.

Why slideware wins (and how it loses)

Slideware wins because it is easy to govern. A 40-page "AI strategy" can be approved without changing a Tuesday. It names pillars. It names risks. It names a centre of excellence. It does not name the team whose week will change.

Slideware loses the moment a regulator, a journalist, or a CFO asks "what did we stop doing?" If the answer is "we stood up a guild," you bought theatre with a new acronym.

DBS-style programmes that work treat enablement as operations: a use case, an owner, a control, a metric. HSBC Hong Kong-style programmes that work put the control next to the prompt, not in a PDF six clicks away. Telstra-style programmes that work pick a handful of workflows (care, field, corporate knowledge, code, customer comms) and run them as products. Canva-style programmes that work dogfood the tools and kill the ones that do not return time.

The losing pattern is identical in every geo we cover (AU, NZ, HK, SG, US, UK, CA): a vendor day, a licence dump, a Slack channel, a "champion network," and a steering pack that reports adoption as logins.

Logins are not hours.

If AI lives in a lab, it is a demo. If it lives in the Tuesday standup, it is an operating system.

A sequence for AI literacy that leaves a residue

This is the sequence we run. Steal it. Rename it if your legal team needs a local noun.

1. Pick a workflow with a clock on it

Not "knowledge management." "The weekly regulatory pack that takes a manager six hours to collapse." Not "customer service." "The second-line email that is 80 percent identical and still written from scratch." Not "coding." "The test boilerplate one squad recopies every sprint."

Telstra care and field workflows have clocks. Bank credit memos at DBS-scale and HSBC Hong Kong-scale have clocks. Canva design-ops and support workflows have clocks. If you cannot point at a calendar block, you do not have a use case. You have a theme.

2. Pair a domain lead with a builder

The domain lead owns the work and the risk. The builder (analyst, engineer, ops designer, or a sharp BA) owns the tooling path. A centre of excellence can coach. It cannot own. If the CoE presents the demo and the desk was not in the room, you have a roadshow.

3. Write the guardrails before the prompts

What data never goes in. What must be checked by a human. What the model is not allowed to decide. What the audit looks like. HSBC Hong Kong and DBS both live inside regimes where "we tried a prompt" is not a control. Telstra and any AU Authorised Deposit-taking Institution rhyme. Canva has customer-content risk of a different flavour. Write the one-pager. Put it in the same folder as the play.

4. Practice on live work, not a toy

Two days. Real documents. Real tickets. Real messy folders. The first afternoon will be ugly. Good. Ugly is how you find the step the slide skipped. We time-box: morning on framing and guardrails, afternoon on first passes, next morning on tightening the play, next afternoon on the decision (keep, kill, or constrain) and the Tuesday handover.

5. Leave a play the team will run without you

A one-page play: trigger, inputs, prompts or tools, checks, output, owner, metric. If it needs a 40-page manual, it will not run on Tuesday. Canva-style product teams already think this way. Banks and telcos can too. They just have more checks in the play.

6. Measure hours returned

Baseline the clock. Re-measure at day 14 and day 30. Publish the number, including when it did not move. Kill the play that did not pay. Scale the one that did. Models deployed is vanity. Hours is a P&L line, or at least an FTE line a CHRO understands.

What CHRO, CAIO, and L&D each owe the other

The new appointment pattern only works if the three roles do not run parallel theatres.

CAIO (or CDO / Head of AI). Owns the guardrails, the approved tool set, and the "what we will not automate" list. Funds the first two sprints. Does not fund a second capability map.

CHRO. Owns the incentive and the job design. If the bonus still rewards old throughput, literacy will be performative. Funds the time to practice. Protects maker hours.

L&D / capability. Owns the sequence and the cohort design. Not a content library. A repeating two-day format with residue. Reports hours and plays adopted, not completions.

When these three appoint into a bank in HK or SG, a telco in AU, or a product company in Sydney or Singapore, the first 120 days decide whether AI training is an operating rhythm or a town-hall video.

Don'ts, because someone will try them anyway

Don't start with a model bake-off. Start with a workflow. The model is downstream.

Don't train 400 people and change zero desktop standards. You will get 400 shadow workflows and a headache for risk.

Don't let a vendor run "AI 101" as the programme. Awareness is a Tuesday lunch. Enablement is a sprint.

Don't measure success as licences. Canva does not measure success as "people opened the app." Neither should a bank.

Don't hide the failures. A killed play is a result. DBS-style and GDS-style cultures already know this. Copy them.

A briefing for the new CAIO

If you are 40 days into the job at a Telstra-scale telco, a DBS-scale or HSBC Hong Kong-scale bank, or a Canva-scale product org, send this to your CHRO and your L&D lead.

1. Name three workflows with clocks. Pick one.
2. Name the domain lead and the builder. Book them for two days.
3. Write the one-page guardrail with risk.
4. Run the sprint. Live work only.
5. Leave a play. Measure hours at day 14 and day 30.
6. Kill or scale. Then pick workflow two.

That is AI training without the slideware. The rest is a keynote.

What "hours returned" looks like on a real desk

A credit officer at a DBS-scale or HSBC Hong Kong-scale bank collapses a memo. Baseline: four hours. After the play: ninety minutes plus a 20-minute human check the guardrail requires. Net: more than two hours, if the check is real and the memo still passes. A Telstra care writer stops rebuilding the same explanation from a blank page. A Canva-adjacent operator stops reformatting the same internal brief. Publish those numbers, including when they do not move. A play that saves six minutes will not survive contact with a CRO. Kill it. The literacy is in the killing as much as in the prompting.

CHRO-shaped owners should care because job design will change. If the hours come back and you fill them with more of the old work, you did not buy productivity. You bought capacity for more theatre. Put a rule on the returned hours: maker time, customer time, or a deleted room. Atlassian-style and Canva-style cultures already argue this. Banks can steal the argument.

A 90-day academy that is not an LMS

If you must have a programme name for the board, make it three sprints, not a content library.

• Sprint 1: one workflow, one team, hours at day 30.
• Sprint 2: a second workflow, teach-back so L&D can facilitate.
• Sprint 3: the first team runs without the supplier. You measure whether the play still runs on a wet Tuesday.

DBS-style academies that matter have a path to the desk. HSBC Hong Kong enablement that matters has a control next to the path. Telstra-scale programmes that matter pick a few families of work. Canva-scale programmes that matter dogfood and delete. The LMS can hold the guardrail and the play template. It cannot hold the residue.

When a new CAIO funds this in the first 120 days, they have bought training. When they fund a 40-page map and a vendor day, they have bought a delay. We will keep naming the difference.

A two-day run of show, hour by hour

This is the sequence we actually run. Steal the clock. Rename the nouns if legal needs a local word. Do not add a third day. A third day is how a sprint becomes a course.

Day one, 09

to 10

. Frame the hated workflow

The domain lead walks a live piece of work: a credit memo at a DBS-scale or HSBC Hong Kong-scale desk, a care explanation at Telstra, a weekly pack at NAB, a support brief at Canva. Not a persona. A file. Write the clock on the wall: how long this took last time, who touched it, where the paste-and-reformat lives. If you cannot point at a calendar block, stop. You have a theme, not a use case.

Risk or the control owner sits in for the last 20 minutes of this block. They hear the workflow before they hear the tool. That order matters in MAS-shaped and HKMA-shaped rooms, and it matters at any AU ADI.

Day one, 10

to 12

. Guardrails before prompts

One page. What data never goes in. What a human must check. What the model is not allowed to decide. What the audit looks like. HSBC Hong Kong and DBS both live inside regimes where "we tried a prompt" is not a control. Telstra and NAB rhyme. Canva has customer-content risk of a different flavour. Write the page. Put it in the same folder as the play.

Do not debate the model zoo here. The model is downstream.

Day one, 13

to 16

. First passes on live work

Ugly on purpose. Real documents, real tickets, real messy folders. The builder drives the tooling path. The domain lead owns truth. L&D watches for the step the slide would have skipped. Time-box each pass. Capture failures in the open. A failed pass is how you find the check the guardrail missed.

Day two, 09

to 12

. Tighten the play

Turn the mess into a one-page play: trigger, inputs, prompts or tools, checks, output, owner, metric. If it needs a 40-page manual, it will not run on Tuesday. Walk it once more on a fresh piece of live work. If it breaks, fix the play, not the story.

Day two, 13

to 15

. Decide and hand over

Keep, kill, or constrain. Named owner for Tuesday. Baseline hours written down. Day-14 and day-30 measure booked on an existing heartbeat, not a new steering pack. Residue circulated before people stand up. If the CAIO or the CHRO sent a delegate who cannot kill, you scheduled applause.

Atlassian-style and Canva-style product teams can run this inside a cadence they already have. Banks and telcos can too. They just have more checks in the play.

A practical invite line: two days, live work only, domain lead plus builder plus control owner. No capability map. No vendor keynote.

How L&D measures hours, not completions

Learning and Development is trained to report completions. Completions make a dashboard green. They do not make a Tuesday shorter. When a CAIO funds AI training, L&D's job is to change the unit of progress.

Measure four things. Publish them, including when they do not move.

1. Hours on the named workflow. Baseline before the sprint. Re-measure at day 14 and day 30 on the same class of work. A credit officer collapsing a memo. A Telstra care writer rebuilding an explanation. A Canva-adjacent operator reformatting a brief. If the number did not move, kill the play. The literacy is in the killing as much as in the prompting.
2. Plays still running on a wet Tuesday. Not "people who attended." Plays the desk ran without the supplier. L&D can audit this in 20 minutes by sitting with three people. Do that. Do not send a survey.
3. Human checks that actually happened. A guardrail that is not used is a liability. Count the checks, not the policy PDFs.
4. Rooms deleted to pay for practice. If the programme added a guild and deleted nothing, you bought a calendar parasite. See the productivity piece.

Completions can live in the LMS as a hygiene line. They cannot be the board line. CHRO-shaped owners should refuse a pack that leads with certificates. CAIO-shaped owners should refuse a second cohort until sprint one has hours. DBS-style academies that matter have a path to the desk. HSBC Hong Kong enablement that matters has a control next to the path. Telstra-scale programmes that matter pick a few families of work. Copy the path, not the brand.

A practical L&D report sentence:

Cohort one: one workflow, one team. Hours at day 30: [baseline] to [now], including no-move. Plays still running: [n]. Checks sampled: [n]. Rooms deleted to pay for practice: [n]. Completions are in the appendix. We will not book cohort two until the play survives a Tuesday without us.