Everything a busy leader needs to start building with AI: what changed this month, the basics in plain English, real use cases by function, 30-minute recipes, prompts you can copy and a simple safety checklist. Updated monthly.
Researched, written and refreshed monthly by the club's AI operator. Every claim links its source. Updated 31 Oct 2026.
Three taps: your role, your biggest weekly pain and the time you have. You get a recipe, a prompt to copy and the tool to open.
What's new in AI: October 2026
11 changes6 min readAs of 31 Oct 2026
The eleven changes this month that matter to leaders, in two lines each. Refreshed on the first weekend of every month.
11changes that matter
11companies and sources
3Malaysia updates
31 Oct 2026last refreshed
MSMicrosoft
Microsoft Copilot gets 25+ finance skills
Copilot can now reconcile accounts, draft journal entries and prepare close packages and forecast reviews, inside Excel and its Cowork agent.
Read more
Microsoft added more than 25 finance skills and partner plugins, including Xero, QuickBooks, Zoho Books, S&P Global and Moody's. In Excel you can switch on an audit trail that records what Copilot changed and why. Microsoft says Copilot does not use your prompts, responses or documents to train its foundation models.
Google's Gemini 4 Argon is out; Carbon is announced
Gemini 4 Argon is in the Gemini app for Google AI Pro and Ultra subscribers. The stronger Gemini 4 Carbon is limited to vetted cyber defenders for now.
Read more
Google says Argon is built for deep, multi-step work in coding, finance, legal and science. Carbon, announced on 28 Oct, focuses on cybersecurity and long tasks; Google is releasing it slowly and has not given a date for general access. Developers pay an introductory API price for Argon until 30 Nov 2026.
Why it matters to you: if your company runs on Google Workspace, the newest Gemini model is already one click away on a Pro plan.
NotebookLM is now Gemini Notebook, and it makes slides
Google's research notebook (renamed in July) now walks you through turning your own documents into a slide deck or video overview, with a review step first.
Read more
Upload board papers, reports or transcripts, then ask for slides. Gemini Notebook suggests topics from your sources, lets you steer style and format, and with the Review toggle on you can edit the outline before anything renders. Answers stay grounded in the files you gave it.
BNM and the SC warn of deepfakes of their senior officers
Fake videos of central bank and Securities Commission officers are being used to sell investment schemes. Check before you trust any video, call or voice note.
Read more
Bank Negara Malaysia shows examples at bnm.gov.my/verify and asks the public to check its Financial Consumer Alert List and the SC's Investor Alert List. In September, Johor police reported a company administrator who transferred RM110,000 after a caller posing as her director gave payment instructions; police were checking whether his voice was cloned. See Deepfakes and payment fraud in the safety section for a simple team rule.
The Budget proposes tax relief for AI software subscriptions, 100,000 more free AI subscriptions for young people, a sovereign AI cloud and help for 4,000 MSMEs to adopt AI.
Read more
Key AI measures in the Prime Minister's Budget 2027 speech:
Lifestyle tax relief to be broadened to include AI software subscriptions.
Another 100,000 free AI subscriptions for young people, extended to apps such as ChatGPT.
A sovereign AI cloud is being planned so that national data stays under Malaysian control.
AI Malaysia Berhad: almost RM15 million towards a safe, ethical AI ecosystem and a target of 200,000 AI-skilled workers.
MDEC: RM30 million to help 4,000 MSMEs adopt AI and automate, and to train and certify 5,000 AI professionals.
Scam checker: the National Scam Response Centre will build an AI app to check suspicious numbers, bank accounts or links before you pay.
Rakyat Digital offers free access to more than 16,000 Coursera courses.
These are Budget proposals; confirm the detail with LHDN or your tax adviser before you claim relief.
One agent inside Gmail, Docs, Sheets, Slides, Chat and Calendar that can take an objective, plan the work and come back with something finished.
Read more
Google announced the Gemini agent at Gemini at Work 2026. Teams can also create coworker agents with their own Workspace account, email and calendar, which only see what you share with them. For finance chiefs, Google added real-time spend caps that pause an agent when a project's AI budget is reached. In our region, Google says DBS Bank runs chains of 70 to 80 specialised agents to prepare corporate credit memos, with humans kept in the loop.
ChatGPT moved to GPT-6, with answers you can click
GPT-6 now powers ChatGPT for everyone, and answers can arrive as charts, forms, buttons and small working tools instead of plain text.
Read more
OpenAI calls the new format Intelligent UI. Ask for a comparison and you may get a side-by-side view; ask for a savings calculator and ChatGPT can build one inside the chat. Paid plans (Plus, Pro, Business, Enterprise) use GPT-6 Sol, while Free and Go use GPT-6 Luna. OpenAI says more than 1.2 billion people use ChatGPT each week.
Why it matters to you: quick internal tools (a pricing calculator, a checklist, a scenario model) can now be made in one message, with no build step. Treat them as drafts and check the numbers.
Claude 5.5 family arrives, and Claude can now do the whole job
Anthropic released Opus 5.5, Sonnet 5.5 and Haiku 5.5 in three weeks, and merged its Cowork agent into the main Claude app with new Docs and Slides.
Read more
Anthropic says Opus 5.5 costs about 40% less to run than Opus 5 on typical work, and Haiku 5.5 about 75% less than Haiku 4.5. In the Claude app you can now ask a quick question or hand over a report due at noon, and Claude keeps working after you close your laptop. Claude Docs and Claude Slides are in beta on paid plans, and slides download as PowerPoint or PDF. By default Claude asks before taking an action. The unified app became generally available for Claude Enterprise on 23 Oct.
Try it: "Turn these three files into a five-slide leadership update, flag anything that slipped."
Malaysia's AI Governance Bill is coming in early 2027
The Digital Minister says the Bill will be risk-based, with stricter testing and human oversight for high-risk AI, and is expected to be tabled early next year.
Read more
Gobind Singh Deo told the Dewan Rakyat that obligations will depend on the potential harm of an AI system rather than the industry it sits in, and that the Bill will work across sectors alongside existing laws. In September he said the draft was complete and going to Cabinet, with the aim of tabling it by the first quarter of 2027.
What to do now: keep a simple list of where your company uses AI and who owns each use. That list will be the starting point for any future compliance work.
ChatGPT Work: hand over a task, get a finished file back
ChatGPT now has Chat for quick questions and Work, an agent that researches and produces documents, spreadsheets, decks, reports and simple websites.
Read more
Work can run once, on a schedule, or when something happens, such as a new Gmail message, a new Slack message or GitHub activity (eligible Plus, Pro, Business, Enterprise and Edu users; not Free or Go). It asks for your approval on important actions. On 26 Oct OpenAI added GPT-6.1 Terra, a faster, cheaper model for Work and Codex aimed at computer use and everyday office tasks.
Try it: schedule a Monday 8am task that summarises last week's sales emails into a one-page brief.
xAI: Grok 4.7 is the latest model, plus a new voice model
Grok 4.7 (21 Sep) is xAI's newest model for coding and knowledge work. Grok Voice 3.0 followed in October for natural phone-style conversations.
Read more
Grok 4.7 has a 500,000-token context window, enough for very long documents. xAI says Grok Voice 3.0 sets a record for resolving customer support calls with replies in under one second. As of xAI's release notes updated 27 Oct, there is no Grok 5 release yet, whatever social media says.
Ten short explainers, from what an AI assistant is to what an agent can do and what PDPA means for you. Two minutes each.
01
What an AI assistant is
A tool you talk to in plain language that writes, analyses, researches and now completes multi-step tasks for you.
Read more
Under the hood is a large language model: software trained on huge amounts of text that predicts useful next words. Modern assistants add tools on top: web search, reading your files, running code, making documents and using apps.
Think of it as a fast, well-read junior colleague. It is excellent at first drafts, summaries, explanations and turning a clear description into a working version. It can also be confidently wrong, so you check facts, numbers and anything that matters.
The big four you will hear about are ChatGPT (OpenAI), Claude (Anthropic), Gemini (Google) and Grok (xAI). Microsoft's Copilot sits inside Microsoft 365. Malaysia also has ILMU from YTL AI Labs, built for Bahasa Melayu and local context.
Rule of thumb: the better you explain your situation, the better it works.
Chat answers a question. An agent takes an objective, plans the steps, uses tools and comes back with the finished file. 2026 is the year agents went mainstream.
Read more
In October 2026 all three major labs offered a mode where you hand over a job rather than a question: ChatGPT Work, Claude (which absorbed Cowork) and Google's Gemini agent.
What an agent can typically do:
research across the web and your connected apps;
create documents, spreadsheets and slides;
run on a schedule (every Monday) or a trigger (a new email);
keep working after you close your laptop;
ask for your approval before important actions.
How to delegate well: give the outcome, the audience, the deadline, the sources it may use and what it must not do. Then review the result like you would a new hire's first draft.
What still needs you: judgement, relationships, approvals and anything that moves money or speaks for the company.
Put AI on a schedule or a trigger: every Monday, every new invoice, every form submission. This is where hours come back.
Read more
Two ways to automate:
Inside your assistant. ChatGPT Work can run scheduled tasks and event-triggered tasks (for example on new Gmail or Slack messages). Claude can repeat a report every Monday.
With a workflow tool such as Make, n8n or Zapier, which connect thousands of apps. Make's own guidance is useful: use plain automation for predictable, structured steps (billing runs, record updates, notifications) and an AI agent step where the input is messy and needs judgement (triaging support tickets, qualifying leads, reading invoices).
Good first automations: a weekly summary of your inbox, a form that files requests into a sheet and drafts replies, a daily digest of competitor news.
Keep a human approval step on anything that sends, pays or deletes.
AI can invent facts, numbers and quotes that sound right. Ask for sources, check the key figures, and never publish what you have not read.
Read more
The labs are improving: OpenAI reports that GPT-6.1 Terra's error rate on its hard test prompts fell from 53.6% to 16.7% at low reasoning effort. That is better, not perfect.
A three-step check:
Ask for sources and open at least the ones behind your key claims.
Re-derive the numbers that drive a decision, or ask a second model to check the first.
Ask it to argue the other side: "What would a sceptical CFO say about this?"
Be extra careful with law, tax, medical and anything about real people. Treat AI output as a strong first draft from a capable colleague who has not been to your office.
Never paste passwords, client data or anything confidential into a tool you have not approved. Under PDPA, personal data is your responsibility, wherever it goes.
Read more
While you learn: use sample files, screenshots or dummy data. Check each tool's data settings and switch off training on your chats where that option exists. Keep API keys and passwords out of prompts and code, and store them as secrets.
For your company: Malaysia's Personal Data Protection Act was strengthened in 2025. Penalties for breaching the data protection principles now reach RM1 million and/or three years' jail, biometric data counts as sensitive data, and data processors (including some software vendors) now have direct security duties. Sending personal data overseas, which most AI tools do, has its own rules.
See PDPA in plain English in the safety section for the checklist.
A national AI agency, a 2030 plan, voluntary ethics guidance and a law on the way. 38% of Malaysian businesses now use AI, but few use it deeply.
Read more
Who runs it: the National AI Office (NAIO), launched on 12 Dec 2024, became AI Malaysia Berhad on 28 Jul 2026, under the Ministry of Digital. It also hosts the Malaysia AI Safety Institute (MY-AISafe).
The plan: the National AI Action Plan 2026-2030 (AI Nation 2030) aims for Malaysia to be in the top 10 of recognised global AI indices, with up to 1.2 percentage points of extra GDP growth and up to 300,000 new jobs attributable to AI by 2030.
The rules: the National Guidelines on AI Governance and Ethics (AIGE, 2024) set seven voluntary principles: fairness; reliability, safety and control; privacy and security; inclusiveness; transparency; accountability; and human benefit. AI Malaysia has since published a voluntary AI Code of Ethics (AICE) and a Boardroom Primer for directors. A risk-based AI Governance Bill is expected in early 2027.
Where businesses are: an AWS study of 1,000 Malaysian business leaders found 38% consistently use at least one AI tool, up from 27% a year earlier. But 67% of adopters mainly use basic tools such as public chatbots, only 19% have a formal plan to scale AI across functions, and 53% name workforce skills as a barrier. That gap is exactly why this club exists.
Local models: YTL AI Labs' ILMU is a Malaysian-built model that YTL says leads global models on a Malay-language benchmark.
Pick one small, real problem, open an AI tool you already use, and ask it to help you build a first version. Ship small, then improve.
Read more
Good first problems: a repetitive task, a tracker you keep in a spreadsheet, a form you send every week, a report you rebuild every month.
Before you start: sign in to your AI tool and to GitHub, and have a sample file ready (no real client data).
Then:
Describe the problem in your own words.
Let it interview you.
Approve a simple scope.
Build it and open it on your phone.
Use it for a day and come back with three changes.
The recipes in Build your first tool in 30 minutes below give you a safe starting point. If you get stuck, bring it to the next gathering: that is what the room is for.
Use cases by function
7 roles6 min read
Concrete things leaders in each function are doing with AI today, with a prompt to try tonight.
CEO and founder
A thinking partner for strategy, a ghostwriter for your voice and a chief of staff that never sleeps.
Read more
What leaders are doing:
Board and investor prep: turn the month's numbers and notes into a board update, then ask the AI to play a tough director and list the ten hardest questions.
Strategy stress tests: "Here is our 2027 plan. Argue against it as our strongest competitor would."
Weekly brief: a scheduled agent that reads your key email threads and reports every Monday morning on what moved, what slipped and who is waiting on you.
Your voice at scale: draft LinkedIn posts, town-hall scripts and staff letters from voice notes, in Malay and English.
Market scans: a research agent that compiles competitor moves, regulation and pricing with sources.
I am the CEO of [company], [one line on what we do] in Malaysia. Here is my draft strategy for next year: [paste]. Act as a sceptical board member. List the 10 hardest questions you would ask, ranked by risk, and what evidence would answer each.
Finance
Faster month-end, clearer variance commentary and credit or investment memos drafted for review, not from scratch.
Read more
What finance teams are doing:
Close and reconciliations: Microsoft's new Copilot finance skills compare balances, flag discrepancies, draft journal entries and prepare close packages for review, with an optional audit trail in Excel.
Variance commentary: paste the P&L against budget and ask for the five drivers that explain most of the gap, in board language.
Credit and investment memos: DBS Bank uses chains of specialised Gemini agents to synthesise financial, regulatory and market information into corporate credit memos, with humans in the loop.
Invoice and receipt capture: read incoming invoices and fill in your accounting system (OpenAI cites this as a typical job for its low-cost Terra model).
Cash forecasting scenarios: build a simple 13-week cash model and ask it to stress-test receivables slipping by 30 days.
Here is our monthly P&L versus budget as a table: [paste]. Explain the variance in five bullet points a non-finance director would understand. Flag any line where the explanation is uncertain and tell me what extra data would confirm it.
Sales and business development
Account research in minutes, proposals tailored to each client and follow-ups that never fall through the cracks.
Read more
What sales leaders are doing:
Pre-meeting briefs: "Research [company] from its website, annual report and news. Give me their priorities, recent changes and three questions to ask."
Proposals and tenders: feed past winning proposals plus the client brief and get a tailored first draft in your house style.
Lead triage: an automation that reads every inbound enquiry, scores it against your ideal customer and drafts a reply for a human to send.
Call follow-ups: meeting tools such as Fireflies or Otter transcribe calls; ask the AI for next steps, owners and a follow-up email.
Pipeline review: paste the CRM export and ask which deals are stalling and why.
I sell [product] to [type of customer] in Malaysia. Here are notes from my last call with [client]: [paste]. Write a short follow-up email that confirms the next steps, then list any risks to the deal I should raise with my team.
Operations
Turn WhatsApp chaos and spreadsheets into simple trackers, checklists and dashboards your team will actually use.
Read more
What operations leaders are doing:
Replace the spreadsheet: build a small web app for site visits, maintenance requests, stock counts or delivery issues, with a form on the phone and a dashboard for managers.
SOPs from reality: record a supervisor explaining a task, then turn the transcript into a step-by-step SOP with a checklist.
Photo checks: Hitachi field technicians use image-recognition agents to verify photos of tools and wiring before and after a job, according to Google.
Exception reports: a daily digest that reads your logs or emails and lists only what needs attention.
Supplier comparisons: paste three quotations and get a like-for-like comparison table with the gaps highlighted.
Today our team reports [problem] through WhatsApp and a spreadsheet. Interview me a few questions at a time, then write the scope for a simple mobile-friendly tracker. Wait for my approval before building.
HR and people
Clearer job descriptions, faster onboarding packs and policies people read, with personal data kept out of public tools.
Read more
What HR leaders are doing:
Job descriptions and interview kits: from a short brief, draft the JD, a scorecard and structured interview questions.
Onboarding: an internal assistant that answers "how do I claim" or "who approves leave" from your handbook only.
Policy drafting: a first draft of an AI-use policy, a hybrid work policy or a code of conduct in plain English and Bahasa Melayu.
Training content: turn a long policy into a ten-question quiz or a short explainer video.
Careful: CVs, payroll and appraisals are personal data under PDPA. Use approved company tools only, and keep a human decision on hiring, pay and performance.
Draft a one-page AI acceptable-use policy for a 50-person Malaysian company. Cover approved tools, what data must never be entered, human review, and who to ask. Plain English, then a Malay version.
Marketing
Campaign ideas, content in several languages, decks and short videos, produced in hours and checked by a human.
Read more
What marketing leaders are doing:
Campaign sprints: brief once, get headlines, social posts and email copy in English, Malay and Chinese, then pick and polish.
Decks and one-pagers: Gamma, Claude Slides or Gemini Notebook turn a brief or a set of documents into a first deck.
Video and voice: tools such as HeyGen, Synthesia and ElevenLabs make presenter videos and voiceovers; label AI-generated people clearly.
Customer insight: paste reviews or survey comments and ask for themes, quotes and what to fix first.
Search visibility: check how AI assistants describe your brand and competitors, and fix what is wrong on your own site.
Malaysia note: paid endorsements, including affiliate links, must be clearly labelled as ads under the MCMC Content Code.
Here are 30 customer reviews: [paste]. Group them into themes with a count for each, quote one review per theme, and suggest one campaign message that answers the top worry.
Legal and compliance
First-pass contract reviews, plain-English summaries and obligation trackers, always signed off by a qualified lawyer.
Read more
What legal and compliance teams are doing:
Contract first pass: compare a supplier contract against your standard terms and list deviations by risk.
Plain-English summaries: turn a 40-page agreement into a one-page brief for the business owner.
Obligation trackers: extract renewal dates, notice periods and caps into a table.
Redaction: Google says law firm Cooley is building an agent that redacts personal and confidential information before filings go public.
Regulatory watch: a monthly agent that summarises new PDPA guidelines, AI governance updates and sector rules with links.
Careful: do not upload privileged or client documents to tools your firm has not approved. AI does not replace legal advice.
Compare this contract [paste] against our standard position [paste]. List every deviation in a table: clause, what it says, our position, risk (high, medium, low) and a suggested redline. Do not invent clauses.
Build your first tool in 30 minutes
6 recipes5 min read
Six step-by-step recipes using tools our members already use. Pick one, follow the steps, and bring it to the next gathering to improve it.
1
20 minbeginner
A Monday brief that writes itself
A scheduled AI task that reads the week's key emails or files and sends you a one-page brief every Monday at 8am.
5 steps
ChatGPTClaude
Read more
You need: a paid ChatGPT plan with Work, or a paid Claude plan. About 20 minutes.
Open ChatGPT and switch to Work (or open Claude).
Connect the app that holds the information, such as Gmail or Google Drive, when it asks. Give access only to what the task needs.
Paste this prompt:
Every Monday at 8am Kuala Lumpur time, review last week's emails
labelled [Label] and the files in [Folder]. Write a one-page brief:
1) what moved, 2) what slipped, 3) decisions I owe, 4) who is waiting
on me. Bullet points, most important first. Ask before sending
anything to anyone else.
Ask it to run once now so you can check the result.
Fix what is off ("Shorter." "Skip newsletters.") and confirm the schedule.
Level up: add a trigger so it also alerts you when a named client emails.
Turn a stack of reports into a ten-slide briefing with sources, then make the AI play your toughest director.
6 steps
Gemini NotebookClaudeGamma
Read more
You need: a Google account (Gemini Notebook) or a paid Claude plan. Use non-confidential documents while you learn. About 25 minutes.
Open Gemini Notebook and create a notebook, or start a Claude conversation.
Upload three to five source documents (reports, minutes, a spreadsheet).
Ask: "Create a slide deck for a board briefing. Executive summary first, then performance, risks and decisions needed."
In Gemini Notebook turn on Review and edit the outline before it renders. In Claude, ask for the outline first.
Generate the deck and download it.
Then ask:
Act as the most sceptical director on our board. List the five
questions you would ask about this deck and which slide fails to
answer each. Cite the source document for every claim.
Level up: ask for a one-page pre-read in the same style.
Describe a tracker in plain words and get a working web app with a form, a list and a simple dashboard.
6 steps
Lovable
Read more
You need: a Lovable account (Bolt.new or Replit work too). About 30 minutes.
Pick one tracker your team keeps in a spreadsheet or WhatsApp: site visits, maintenance requests, customer complaints.
Open Lovable and paste:
Build a simple mobile-friendly web app for [team] to log [thing].
Fields: [list fields]. Staff submit on their phone. Managers see a
list with filters by status and date, plus a dashboard with counts
by status this week. Use sample data only. Keep it simple.
Let it build, then open the preview on your phone.
Use it as a staff member and as a manager. Note three changes.
Ask for them one at a time: "Add a photo upload." "Make status a dropdown."
Before real use: add sign-in, check who can see the data, and test end to end.
Level up: add a daily email to the manager with new items.
Every new enquiry is read by AI, sorted, logged in a sheet and given a draft reply for a human to approve.
6 steps
Maken8nZapier
Read more
You need: a Make account (n8n or Zapier work too), a test email inbox or form, and a Google Sheet. About 30 minutes.
In Make, create a scenario. Trigger: new email in your test inbox (or a new form submission).
Add an AI agent step with this instruction:
Read the enquiry. Return: category (sales, support, partnership,
spam), urgency (high, medium, low), a one-line summary, and a polite
draft reply in the same language as the sender. If unsure, say so.
Add a Google Sheets step that appends a row with those fields.
Add a step that saves the draft reply to your drafts folder. Do not auto-send.
Send yourself five test emails, including one in Malay and one spam.
Check the sheet and the drafts, then tighten the instruction.
Why this pattern works: Make recommends AI agents for messy, judgement-based inputs like support triage and lead qualification, and plain automation for the predictable steps around them.
Record or transcribe a meeting, then get decisions, owners, deadlines and a follow-up email ready to send.
4 steps
FirefliesOtterChatGPTClaude
Read more
You need: a meeting recorder (Fireflies, Otter or your video app's transcript) and any AI assistant. About 15 minutes.
Tell everyone you are recording and why. Get consent.
Record the meeting, or download the transcript afterwards.
Paste the transcript into your assistant with:
From this transcript, list: decisions made, actions with owner and
deadline, open questions, and risks. Then draft a short follow-up
email to attendees. Flag anything where the owner or date is unclear.
Check owners and dates against your memory, then send.
Level up: ask for the actions as a table you can paste into your project tracker, or schedule a Friday check on what is still open.
6
25 minbeginner
Your own decision memo assistant
A saved project with your company context and a house template, so every proposal comes back as a clear one-page decision memo.
5 steps
ChatGPTClaude
Read more
You need: ChatGPT or Claude (Projects). About 25 minutes.
Create a new Project called "Decision memos".
Add context files: a short company profile, your strategy on a page, and one example of a memo you liked. No confidential data while you learn.
Add these project instructions:
You help me write one-page decision memos. Always use this structure:
Decision needed, Recommendation, Options considered (with cost, benefit,
risk), What we would need to believe, Next steps and owner.
UK spelling, ringgit, plain English. Ask up to five clarifying
questions first. Never invent numbers: mark gaps as [TBC].
Test it on a real but low-stakes decision, such as choosing between two software vendors.
Share the project with a colleague and compare results.
Level up: add a step where it argues for the option you did not pick.
Sixteen prompts leaders use every week. Tap Copy, fill in the [brackets], and paste into any assistant.
Interview me first
Make the AI ask the right questions before it writes or builds anything.
I want help with [task]. Before you start, interview me a few questions at a time until you are confident you understand the goal, the audience, the constraints and what good looks like. Then summarise your understanding and wait for my approval.
Read more
Tip: Use this at the start of anything important. It is the single habit that improves results most.
Write the scope for a tool
Turn a messy problem into a scope you can approve before anything is built.
I run [role] at [company]. Today [problem] is handled through [WhatsApp, spreadsheets, calls], and the biggest friction is [friction]. Write a one-page scope for a simple tool: users, what they need to do, data to capture, screens, what is out of scope, and how we will know it works. Wait for my approval.
Read more
Tip: Read the scope like a contract. Cut anything you do not understand.
Toughest board questions
Stress-test a plan, paper or pitch before the real meeting.
Act as a sceptical, experienced board director. Read [document]. List the ten hardest questions you would ask, ranked by risk to the company, and say what evidence would answer each one.
Read more
Tip: Run it twice: once as the investor director, once as the independent director.
One-page decision memo
Get a clear, structured recommendation instead of a wall of text.
Write a one-page decision memo on [decision]. Structure: decision needed, recommendation, options considered with cost, benefit and risk, what we would need to believe, next steps with owners. Mark any missing numbers as [TBC]; do not invent them.
Read more
Tip: Ask it to argue for the option it did not recommend, then decide.
Pre-mortem
Find out how a project could fail while you can still fix it.
It is one year from now and [project] has failed badly. Write the most likely story of why, in five causes ranked by likelihood. For each cause give an early warning sign we could track and one action we could take this month.
Read more
Tip: Great before approving budgets or launches.
Explain it to me simply
Understand an unfamiliar topic in five minutes.
Explain [topic] to a smart business leader with no technical background. Use one everyday analogy, then three things I must know, then two questions I should ask my team or vendor. Under 300 words.
Read more
Tip: Follow up with: "What do people usually get wrong about this?"
Summarise a long document
Get the essentials of a report, contract or paper fast, with page references.
Summarise this document for a busy CEO: five key points, three risks, two decisions needed and anything surprising. Give the page or section for each point so I can check it.
Read more
Tip: Always open the cited pages for anything you will act on.
Rewrite in my voice
Turn rough notes into an email that sounds like you on a good day.
Here are three emails I wrote that I like: [paste]. Learn my tone. Now turn these notes into an email to [person] about [topic]: [notes]. Keep it under 150 words, warm but direct, UK spelling.
Read more
Tip: Save your three example emails in a project so you never paste them again.
English and Bahasa Melayu versions
Produce a message in both languages that reads naturally, not translated.
Write this announcement for our staff in English and in natural, professional Bahasa Melayu (not a word-for-word translation): [message]. Keep both under 120 words and match the tone.
Read more
Tip: Have a native speaker check anything that goes to customers or regulators.
Explain the numbers
Turn a table of numbers into a story a non-finance director understands.
Here is [report] as a table: [paste]. Explain what changed versus last period and budget in five bullet points, naming the main drivers. Flag any number that looks inconsistent and tell me what to check.
Read more
Tip: Use dummy or summarised figures in public tools.
Competitor snapshot
A sourced one-page view of a competitor.
Research [competitor] using public sources. Give me: what they sell, pricing signals, recent news and hires, strengths, weaknesses, and how they describe themselves versus how customers describe them. Cite a source for every claim and say when information is older than six months.
Read more
Tip: Ask for the sources as a list of links so you can check them.
Prepare me for a meeting
Walk in knowing the person, the context and your three asks.
I am meeting [name, role, company] about [topic]. From public information, give me: their likely priorities, recent news, two things we have in common, three questions to ask and the one thing I want them to agree to.
Read more
Tip: Keep personal data use proportionate. Stick to professional, public information.
Draft a simple policy
A first draft of a clear, short policy your team will read.
Draft a one-page [topic] policy for a [size] Malaysian company. Plain English, numbered rules, who it applies to, what to do if unsure, and who owns it. Then list three questions our lawyer should review.
Read more
Tip: A first draft only. Have HR or legal sign it off.
Turn a task into an SOP
Capture how a job is really done, as steps and a checklist.
Here is a transcript of [person] explaining how they do [task]: [paste]. Turn it into a standard operating procedure: purpose, steps numbered in order, a checklist, common mistakes and when to escalate.
Read more
Tip: Record the expert on your phone, then use the transcript.
Check my work
Get a second pair of eyes on anything before it goes out.
Review this [document] as a careful editor and fact-checker. List: factual claims that need a source, numbers that do not add up, unclear sentences, and anything that could embarrass us. Do not rewrite it; just list the issues.
Read more
Tip: Use a different AI from the one that wrote it.
What can I automate?
Find the best AI opportunities in your own week.
Here is what my week looks like: [list recurring tasks and how long each takes]. Rank the five best candidates for AI or automation by hours saved and ease. For each, suggest the simplest tool and the first step I could take tonight.
Read more
Tip: Bring the top answer to the next gathering and build it.
Choosing the right tool
9 jobs3 min read
Start from the job you need done, not the brand. Each card links to tools in our Tools we use list.
I want to
ChatGPT
Claude
Gemini
Grok
Copilot
Also good
Ask, think and draftUse a general assistant: ChatGPT, Claude, Gemini or Grok. Pick the one your company already pays for.Read more
ChatGPT: broadest everyday assistant; Work mode for longer tasks.
Claude: strong writer and analyst; Docs and Slides built in.
Gemini: best if you live in Gmail, Docs and Sheets.
Grok: live answers from the web and X.
Copilot: best if you live in Outlook, Excel and Teams.
Try two on the same task for a week. The winner is the one your team actually uses.
I want meeting notes and follow-upsFireflies or Otter record and summarise; any assistant turns the transcript into actions.Read more
Fireflies, Otter: record, transcribe and summarise meetings.
Notion: keep notes, decisions and projects in one place with AI built in.
Always tell people you are recording.
ChatGPT: no
Claude: no
Gemini: no
Grok: no
Copilot: no
FirefliesOtterNotion
I want video, voice or imagesHeyGen or Synthesia for presenter videos, ElevenLabs for voice, Descript to edit and Runway for generated footage.Read more
HeyGen, Synthesia: avatar and presenter videos, translation.
ElevenLabs: voiceovers and dubbing.
Descript: edit video by editing the transcript.
Runway: AI video generation and editing.
Label AI-generated people and voices clearly, and never clone someone's voice without written consent.
ChatGPT: no
Claude: no
Gemini: no
Grok: no
Copilot: no
HeyGenSynthesiaElevenLabsDescriptRunwayCanva
I need Bahasa Melayu or local data residencyTry ILMU, the Malaysian-built model from YTL AI Labs, alongside the global assistants.Read more
YTL says ILMU is developed, owned and operated in Malaysia and outperforms other frontier models on a Malay-language benchmark. If where your data is processed matters, ask every vendor where inference runs and whether you can choose a region. Malaysia's Budget 2027 also plans a sovereign AI cloud.
Seven habits that let your company move fast with AI without leaking data, getting scammed or breaking the law.
PDPA in plain English
Malaysia's data law got real teeth in 2025: RM1 million fines, 72-hour breach reporting and mandatory data protection officers for many firms.
Read more
The Personal Data Protection (Amendment) Act 2024 came into force in stages during 2025. What leaders need to know:
Penalties: breaching the data protection principles can mean a fine of up to RM1,000,000 and/or up to three years' jail (from 1 Apr 2025).
Biometric data (faces, fingerprints, voiceprints) is now sensitive personal data.
Data processors, not just your company, have direct security duties. Check your AI and software vendors' contracts.
Breach notification (from 1 Jun 2025): tell the Commissioner within 72 hours of a breach that causes or is likely to cause significant harm, and affected people within seven days of that notice. Keep a breach register for at least two years.
Data protection officer: required if you process personal data of more than 20,000 people, sensitive or financial data of more than 10,000, or do regular and systematic monitoring. Register the DPO with the Commissioner within 21 days.
Portability: people can ask to move their data to another provider (from 1 Jun 2025).
Overseas transfers: most AI tools process data outside Malaysia. Transfers are allowed where the destination's law is substantially similar or protection is equivalent, or under other conditions such as consent; the 2025 guideline describes a Transfer Impact Assessment.
Your three moves this quarter: know which AI tools touch personal data, check the vendor terms, and make sure your breach plan names who calls the Commissioner.
Tell your team which tools are approved, what data never goes in, and when a human must check. Then encourage them to use it.
Read more
Most staff are already using AI, with or without permission. A short policy turns hidden risk into managed advantage. Cover:
Approved tools and accounts (company accounts, not personal ones, for work data).
Red data that never goes into AI tools: passwords, client confidential information, payroll, health data, unannounced results.
Human review before anything is sent externally, published, or used to decide about a person.
Disclosure: say when content or images are AI-generated where it matters.
Owner: who to ask, and how to propose a new tool.
AWS's 2026 Malaysia study found only 30% of businesses have clearly defined accountability for AI and 18% have documented escalation processes, so even a simple owner and escalation line puts you ahead. AI Malaysia's voluntary AI Code of Ethics (AICE) and Boardroom Primer are good free references, and the Draft a simple policy prompt in our library writes the first version for you.
A familiar voice or face on a call is no longer proof. Agree a call-back rule for any payment request today.
Read more
Real cases:
In Johor (Sept 2026) a company administrator transferred RM110,000 after a caller posing as her director told her to make payments. Police were investigating whether his voice was a deepfake.
In Hong Kong (2024) an Arup employee paid out about US$25 million after a video call with deepfaked senior colleagues.
In October 2026 Bank Negara Malaysia and the Securities Commission warned of deepfake videos of their own senior officers promoting investments.
The team rule:
No payment or bank-detail change on the strength of a call, voice note, video or WhatsApp alone.
Call back on a number you already have, or confirm in person.
Use a code word for urgent requests from the CEO or CFO.
Two people approve anything above a set amount.
If it happens: call the National Scam Response Centre on 997 immediately, then your bank.
Agents that read email and web pages can be tricked by hidden instructions. Limit their access and keep approvals on.
Read more
OWASP's 2026 list of the top risks for AI applications puts prompt injection first (hidden instructions in content the AI reads) and excessive agency third (an AI with more permissions or autonomy than it needs).
Practical rules:
Connect only the apps and folders a task needs, and remove access when the task ends.
Keep ask before acting switched on for sending, paying, deleting and posting. Claude asks before taking an action by default, and ChatGPT Work asks you to approve important actions.
Use separate, limited accounts for agents where your tools allow it. Google's coworker agents, for example, get their own identity and only see what you share.
Set spend limits on AI usage and review the bill monthly.
Read what the agent did, not just what it says it did.
Know whether your chats are used for training, how long they are kept and who in your company can see them.
Read more
For every AI tool your team uses, find out:
Training: are prompts and files used to improve the vendor's models? Switch it off where you can. Microsoft, for example, says Copilot does not use your prompts, responses or documents to train its foundation models.
Retention: how long are chats and files kept? Some business plans offer zero data retention for certain uses.
Location: where is data processed? This matters for PDPA overseas transfer rules.
Access: who can see shared projects, connected apps and agent logs?
Defaults change: for example, Google switched request storage on by default for new paid Gemini API projects in October 2026. Re-check settings when vendors announce changes.
The AI Governance Bill is expected in early 2027 and will be risk-based. A simple AI register now will save you work later.
Read more
What we know: the Digital Minister says the Bill will be cross-sector, risk-based and focused on the potential harm of each AI system, with stricter safety controls, testing and human oversight for high-risk systems. He also described a wider framework of standards, regulations and law, with a new standards centre working with SIRIM.
Start now, it takes an afternoon:
List every AI use in your company: tool, purpose, data used, owner.
Mark anything that affects people's rights, safety, money or health as high attention.
For those, write down the human check and how you would explain a decision.
Align with the seven AIGE principles: fairness; reliability, safety and control; privacy and security; inclusiveness; transparency; accountability; human benefit.
The national learning portal's free AI courses, including Agentic AI untuk Semua. Young Malaysians aged 18 to 30 can earn three months of free AI tools.
Read more
Rakyat Digital, run by MyDIGITAL under the Ministry of Digital, lists courses such as AI Fluency and Agentic AI untuk Semua. Under the AI untuk Rakyat youth programme, Malaysians aged 18 to 30 who pass six modules get a free three-month subscription to selected tools; the first partners named were ILMUchat, Alibaba's Wonderclip and MuleRun, and Google's Gemini Enterprise. Budget 2027 adds another 100,000 subscriptions, and Rakyat Digital also offers free access to more than 16,000 Coursera courses.
Great for your younger team members and your children.
AI Malaysia's voluntary guide for directors and senior management, with a board AI readiness assessment.
Read more
Four chapters: AI fundamentals, what to consider before adopting AI, governance and oversight, and a maturity assessment to find gaps. No technical background needed. Free download after a short form.
Free self-paced courses, including AI Foundations, Agents and Workflows, and a three-hour AI Leadership course.
Read more
Sign in with a ChatGPT account. Pass a course assessment with 80% or more to earn a badge; complete a pathway for a certificate of completion. OpenAI is clear that these are not certifications.
Google's practical course on using AI at work: five modules, under five hours, with a certificate. It is a paid Coursera subscription, so check Rakyat Digital's Coursera access first.
Read more
Google lists US$49 a month after a seven-day free trial in the US and Canada, and says the cost may be lower in other countries. Rakyat Digital offers free access to more than 16,000 Coursera courses; check whether this one is included before you pay.