The AI Sessions · 22 modules · 60 min — 2 days
This is the whole library. Each module has been run in front of a real room and revised after it. A session is assembled from these to fit the people in front of me — a founder room and a finance team get different days out of the same catalogue.
§ 01 — Foundations
No matter who is in the room, they need shared words and a working mental model before anything else lands.
Open & intros
6–18 min
A structured round of introductions that surfaces what each person does, what they’d want AI to solve, and the one document their role depends on.
Best for
Every session. It’s also the discovery call you didn’t have to schedule.
Vocabulary
25 min
Model, context, prompt, agent, API, connector, skill, context graph, harness — each defined in plain language with a mental model attached, so nothing later in the day sounds like jargon.
Best for
Any room where people have used a chat window but never built anything.
How code changed
5 min
Software used to be about describing the world’s structure in advance. Now you give a model enough scaffolding to find that structure itself — which is why the whole toolkit suddenly looks different.
Best for
The bridge between vocabulary and everything after. Skip it and the rest floats.
Agent to employee
18 min
Building an agent is onboarding a smart new hire: reading, access, rules, and a month of corrections. The module climbs from prompt to agent to code, one rung at a time.
Best for
Non-technical rooms and executives who already think in org charts.
§ 02 — Frames
What AI actually changes about how a company works — and the objections that block the room until they’re answered.
Companies in the AI age
35 min
A two-thousand-year arc from span of control through Taylorism to now, landing on the line the whole day turns on: AI doesn’t augment a company, it reveals what it actually understands.
Best for
Founders and leadership. The most-discussed module in the library.
Institutional knowledge
12 min
The tacit expertise your best people carry is an asset you never booked, and it leaks three ways. This is the case for the dashboard as company memory — so senior judgment survives the person who held it.
Best for
Family businesses and any room with a succession problem.
Moats & Thiel
10 min
Software used to be the moat. When building gets cheap, the defensible things become distribution, taste, and genuinely going from zero to one — ending on the contrarian question worth holding through the build.
Best for
Rooms worried AI erases their advantage. It reframes the fear as a question.
Privacy for enterprises
12 min
The objection that quietly blocks the room: what the encryption actually is, why API data doesn’t train the model, how the tiers differ — and why banning AI creates more risk than sanctioning it.
Best for
Every corporate and regulated room. Answer it early or lose them.
§ 03 — Building
The practical modules. Each one leaves people able to build or operate something specific, not just describe it.
Dashboards
15 min
Five stages — input, view, analyze, insight, action — that turn a reporting screen into an intelligence layer: something that makes the company legible to its people and to AI, and then acts. Most dashboards stop at stage three.
Best for
The spine of most corporate sessions. Where a dashboard stops describing and starts doing.
Building your agent
20 min
Purpose, context, tools, boundaries, memory — one at a time, in build order, with read-only tools first as the discipline that keeps early agents safe.
Best for
Immediately before a build session. It’s what makes the build productive instead of frantic.
Firebase setup
10 min
A six-step checklist for getting a real backend running, plus the prompt that hands the entire setup to Claude Code.
Best for
Build days with people willing to install things. Skip for pure operator rooms.
Voice agents
15 min
Speech-to-text, reasoning, text-to-speech — three swappable pieces. The India case: vernacular voice isn’t a feature, it’s the front door for the next few hundred million users.
Best for
Any room with field operations, or an India-market product.
Mastering Cowork
25 min
For people who’ll never open a terminal: pointing an agent at your actual files and folders, the four jobs worth delegating first, how to brief it, and where it still fails you.
Best for
Operators and functional heads. The most-requested module of the year.
Skills — what to capture
15 min
How to spot the expertise worth writing down, and how to capture it so an agent can reuse it rather than being re-taught every time. The third time you correct it the same way, that’s the skill.
Best for
Rooms with deep institutional knowledge. Pairs with the succession frame.
§ 04 — Applied
Deep modules for a single business function, built from real engagements and adaptable to the room’s own industry.
AI in marketing
35–55 min
Content nobody wanted that anyone could have made — a systems failure, not an AI one. The fix is a brand ontology, built in seven steps, that any agent can then work from.
Best for
Marketing teams and founder-led brands. A real-estate variant exists.
AI for accounting
20 min
Built for Indian finance functions: Tally and Zoho workflows, GST reconciliation, and the capture-classify-reconcile work that quietly consumes a finance team’s month.
Best for
Finance heads and CFOs, and any business where the books are the bottleneck.
Live demo
10 min
A walkthrough of a working intelligence dashboard — built for a real supply-chain business — so the room sees the finished shape before being asked to build toward it.
Best for
Swappable for a walkthrough of the host company’s own systems.
AI executives
5 min
What it looks like when each seat at the leadership table has an agent reading everything underneath it and surfacing the three things that matter today.
Best for
Leadership rooms, immediately after the demo.
§ 05 — Mechanics
The structural modules that make a session hands-on — and the closing that makes people leave changed rather than informed.
Build session
30–90 min
A framed, timed working block where people build something real for their own work — in pairs or alone, with help in the room. A hackathon, not a pitch.
Best for
The centrepiece of any hands-on day. Protect this time; cut philosophy first.
Show & tell
10–25 min
A four-question format for sharing what got built, which surfaces the good ideas and quietly teaches the room what was possible.
Best for
Immediately after any build. It’s where the peer learning happens.
Next steps
4 min
Connect a database, bring in outside data, add reasoning, deploy it — plus how to hand each of those to Claude Code without knowing how to write it.
Best for
The last practical slide of a build day.
Why this matters
6–10 min
The closing arc — on unevenly distributed futures, on why checklists beat brilliance, and on what it means to build rather than only consume this technology.
Best for
The ending. It’s what makes a session land as a turning point, not a talk.
Common session shapes
| Session | What it's built from |
|---|---|
| 60 min — talk | Vocabulary · Companies (lite) · Dashboards · Closing |
| 90 min — leadership | Vocabulary · Companies · Privacy · Dashboards · Closing |
| 3 hrs — intro | Foundations · Frames · Dashboards · Build · Show & tell · Closing |
| 3 hrs — build day | Foundations · Dashboards · Agent · Firebase · 75-min build |
| 3 hrs — operators | Foundations · Privacy · Cowork · Skills · Build · Closing |
| 2 days | Day one, the full arc. Day two, the build. |
The three-hour intro is the ₹1,00,000 format. The six-hour customised day is ₹3,00,000, or ₹5,00,000 with a demo dashboard built for your company beforehand.
To build a session
Who's in it, what they do, and how long you have. I'll come back with the modules I'd run and why.