Yash Verma

I find the one channel a B2B company can own in the US. Then I build the agents that run it.

Seven years selling into the US from India. Everything below came out of that work.

Exhibit 1Where the work happened

IIT Kharagpur
Dr. Water
Climes
Equilibrium
Schillings Air
IIT Delhi
Amazon Propel
COP28
USAID
FICCI
TEDx SRCC
Terra.do

Companies built, programmes selected for, and rooms spoken in.

Notes from my own P&L

Five things I paid to learn. Each one ends with something you can try.

A point of view is a filter

Two AI startups raised money this year on the same bet, from opposite doors. Hang Ten took $32M to replace the IT services firm. Six weeks later a startup called June took $20M to fix the part enterprises are stuck on, getting AI deployed at all. Both will sell. A point of view finds the buyers who already half-believe it.

We saw this selling air purifiers to offices. Specs and pricing got ignored. Replies came from one email: the air quality reading at the buyer's own office, that morning.

Exhibit 2Two raises, one bet

Hang Ten

$32M

"AI eats IT services"

June

$20M

"Deploying AI needs a new kind of services"

same fact: AI will do the work

Hang Ten seed round, TechCrunch, 24 Jun 2026. June pre-seed, TechCrunch, 3 Aug 2026

Try this: Write the sentence your buyer already half-believes. Put it at the top of your homepage.

Read the post

Make the technical proof the reason they buy

Consultancies sold enterprises the same carbon outcome we did, as a service. So we built the measurement platform instead. It used satellite data and field sensors to verify results, and it went at the front of every pitch.

Exhibit 3How the proof turned a service into a product

Climes, 2022 to 2024

Try this: List the proof a bigger competitor cannot copy in six months. Lead with it.

The number that kills a category is rarely in the report

We almost built an alkaline water business. We killed it in two months over one number no report carried. Distributor commissions take 30 to 40% of every Kangen machine sold. That is about $1,800 to whoever closes a household. Our bottle was going to sell for $100 to $120. You cannot pay $1,800 to close a $120 sale, so the channel that built the category was shut to us.

Exhibit 4What the channel pays vs what we could charge

Commission vs price
Item	Amount
Commission per Kangen sale	about $1,800
Our price	$100 to $120

Yash Verma, "How to check if a category is worth entering", Aug 2026

Try this: Find out who gets paid to close each sale, and how much. Do that before you size the category.

Read the full piece

Twelve email flows will not save you

The advice is to get twelve flows live. In our account, four flows made 88% of flow revenue. The real leak was two flows fighting over one trigger, where the older one wins by default. The flow earning $3.08 per recipient got 177 sends. The one earning $0.66 got 16,810.

Exhibit 5Share of flow revenue

Share of flow revenue
Group	Share
4 flows	88%
8 flows	12%

Exhibit 6Two flows, one trigger

Two flows on one trigger
Flow	Revenue per recipient	Sends
A	$3.08	177
B	$0.66	16,810

Yash Verma, "Three email lessons that go against the standard advice", Aug 2026. Dot area is proportional to sends.

Try this: Sort your flow report by revenue per recipient. Check whether two flows share a trigger. The same cap applies to B2B onboarding sequences.

Read the full piece

An agent only knows what you write down

I answered 40+ questions about myself before letting an AI write a line for me. A new hire gets six weeks to learn how you sell. An agent gets only what you put in writing. So its ceiling is how well you have examined your own work. Mine runs on a folder of plain markdown any model can read.

Exhibit 7The folder the agents read

operating-system/├── AGENTS.mdthe router: which file to read for which task├── context/│   ├── principles.mddecisions already made│   ├── voice.mdhow anything written as me must sound│   └── record.mdnumbers, with a source for each├── skills/│   ├── company-teardown/│   ├── review-mining/│   ├── backlog-routing/│   └── ... 13 more└── workspaces/one folder per live piece of work

Simplified. The real system has 16 skills.

Try this: Write the ten questions a new hire would ask in week one. Answer them before your next agent project.

Read the post

Numbers on the board

Exhibit 8Google Ads: $125K in, $280K back

Google Ads
Measure	Value
Spend	$125K
Conversion value	$280K
Return	2.2x

Google Ads account, Mar 2025 to Sep 2026

Exhibit 9Carbon credits grown in India, sold in Europe and the US

Equilibrium carbon markets

Exhibit 10Capital raised

$400K

raised from a US venture fund, pre-seed

Dr. Water pre-seed, Nov 2025

Agents I built and run

Most of these started as a task I was tired of doing twice.

Exhibit 11How the week runs

Agents
Input	Agent	Output
Idea doc	Backlog router (weekdays, 8am)	ClickUp tasks
Company name + meeting	Company teardown	Nine-section brief
Customer reviews	Review miner	Complaint themes, unmet needs

Running since 2026

Company teardown
Name a company before a meeting; get a nine-section brief with sources checked by country.
running
Review miner
Reads a category's customer reviews and returns the complaint themes and unmet needs, ranked.
running
Backlog router
Reads an idea doc every weekday morning and files each line as a task in the right project.
running
Channel whitespace finder
Give it a B2B buyer and a category; it maps where that buyer finds products and which ground nobody owns yet.
in progress, Oct 2026

I have been building since before growth was the job.

1st place, Dishathon
1st runner-up, MapmyIndia Hackathonadaptive traffic lights and accident detection on their APIs
A React dapp on Tezos, with a tutorial
A voting contract on Solidity

About

I've sold three products into three markets. Air purifiers to offices in Delhi, carbon credits from Indian farms to buyers in Europe and the US, and a hydration brand to American families. The move was the same every time: find the proof a buyer cannot ignore, then build the pitch around it.

I work in bursts. Anything I do twice becomes a system. Mornings go to training and meditation, which is why the bursts don't burn me out.

Principles

Choose the field slowly, then refuse to leave it.
A channel pays in its second year, and most teams switch in the first.
I don't invent things; I carry things across.
My best wins carried a proof from one world to another.
When you know what's coming, prepare. When you don't, position.
Launches get plans. New categories get a point of view and room to move.
Always ask what you can avoid.
Four email flows made 88% of the revenue. The work was deciding what to switch off.
Technology is an efficiency layer on the value you offer.
Agents take work off the team. The offer still has to be worth buying.
Relationships compound longer than any deal.
I would rather lose a negotiation than a person I will work with again.

Press and rooms

Exhibit 12Coverage

Linked to the original articles.

Rooms

Showcased carbon projects at COP28, Dubai, Dec 2023
Guest speaker, International Carbon Conference, Oct 2023
Judge, TEDx SRCC flagship, 2025
Seller testimonial, Amazon Seller Export Connect, 2025
Top 3 startups, FICCI
1st place, Terra Studio by Terra.do, for the measurement platform

Contact

If growth is the number you are worried about, write to me: yash@yashminimal.com.

LinkedIn GitHub