Most AI training ends with a deck and a shrug. Mine ends with your team holding something they built themselves - an assistant, an automation, a workflow that runs on Monday morning without me in the room.
"We rolled out AI to 200 people" is not a result. Hours returned, cost removed, revenue moved - those are results. If a board slide counts licences instead of outcomes, the programme is already drifting.
Bolting a chatbot onto a broken process gives you a faster broken process. I map the actual workflow first, find where the time leaks, then redesign it - and only then decide which tool earns a place in it.
Literacy is knowing what AI is. Fluency is reaching for it without thinking, the way you reach for a spreadsheet. Literacy comes from a keynote. Fluency only comes from building something yourself, badly, and then fixing it.
This one came from my wife, CA Jagariti Mathur, describing AI in her accounting practice: the AI drafts, the CA signs. In regulated work that distinction is the whole ballgame - and it is what lets a cautious profession adopt AI at speed.
Most people are still writing better prompts. The compounding move is building a context architecture you own - who you are, how you work, your frameworks, your voice - in plain files you can load into any model. I run my whole business on one. When a better model arrives, I move in a day instead of starting over.
A caution I repeat in every session: the stack is the least durable part of this page. Tools will churn. The workflow design and the context architecture underneath them will not.
Built in Claude Code for R Accounting Solutions, an award-winning practice serving 200+ clients across APAC. Client journey tracking, document intake, a compliance calendar that auto-calculates statutory dates from financial year end, and a two-legged receivable/payable ledger the off-the-shelf accounting software could not hold.
Positioning, proposition, design, copy, the build itself, the domain on GoDaddy, hosting on Netlify, Google Search Console, Google Analytics and Microsoft Clarity. No agency, no developer, no team. Claude Code in the terminal, NotebookLM and n8n. Eighteen years running marketing departments, and I had never done a single one of those things with my own hands.
An 18-field Notion CRM - schema, views, and data - designed and built through Claude Code rather than by hand, then populated with roughly 356 contacts across recruiters, former LEGO and P&G colleagues, and client relationships. It tells me who to reach out to this week and drafts a personalised hook for each one.
Identity, voice, frameworks and live project state held in plain markdown files that load into any model - plus reusable skills for the writing I do most often. This is belief number five, made concrete. It is also the single thing I am most often asked to help leaders replicate.
A smaller case, but the one that convinced a sceptical professional. CA Jagariti Mathur used NotebookLM to prepare for her CSP examination - source material in, generated Q&A and audio summaries out.
Prior speaking and workshop clients across all topics include Citibank, UBS, Grab, Deloitte, Sephora, PropertyGuru, Johnson & Johnson, Common Purpose, and the Ministry of Home Affairs, Singapore. See the full speaking page →
AI Spark Session, Singapore, May 2026 · verbatim answers from the written feedback form · session rated 4.5 / 5.
The most common request in the "what could we improve?" column was more use cases and more case studies. Nobody asked for less hands-on time.
- My read of the same feedback formA no-fluff tour of what is already working, what is hype, and what your people can use on Monday morning. Designed to convert AI anxiety into AI energy.
The signature workshop. Hands on keyboards from the first hour. Nobody leaves without a working AI assistant and an automation they built themselves. No technical background required.
I embed part-time and drive real adoption - mapping workflows, building the automations, and coaching your team until AI becomes an operating habit rather than a one-time event.
Tell me what your team does all day and I will tell you, honestly, whether AI moves the needle on it. If it does not, I will say so.
Straight into my calendar. No form, no back-and-forth on times. Bring one workflow that is eating your team's week.
Pick a time →What I am building, what is working, and what turned out to be hype. Roughly monthly. No sequences, no pitch.