01 — How I think
The question I can’t stop asking
For a long time I thought I was confused about what I wanted to do. Marketing, maybe. Sales. Startups. Tech. AI. I kept starting things to find out.
Then I noticed the same thing kept happening in everything I built. I’d run into something slow, repetitive or stuck on one person, and I couldn’t leave it alone.
And I’d ask the same questions, pretty much in the same order.
- 01 Why are we doing it this way?
- 02 Can this be simpler?
- 03 Can this be automated?
- 04 Can we build a system around it?
- 05 Can it work without me?
That’s basically it. The rest of this page is me doing that in real businesses.
02 — Proof
Three problems I couldn’t leave alone
Each of these started because something annoyed me and I couldn’t stop thinking about why it worked that way. I’ve written them up the same way: the problem, what I did, what happened, and what I’d tell you in an interview.
-
The agency that ran on my hours
The problem
I started Knox on my own, building Shopify stores for Indian D2C brands. The more clients we took on, the more everything depended on my time and attention, and I was running out of both. Honestly, I also didn’t enjoy the repetitive delivery work, so I really wanted to stop doing it.
What I did and what happened
What I did
- Turned delivery into a repeatable pipeline. Set the same steps for scoping, build, QA and handoff, so quality didn’t depend on who picked up the job.
- Cut the service menu to two things. Shopify builds and D2C growth. With fewer kinds of work, a proper delivery system finally became possible.
- Hired six people with clear ownership. Delivery, design and client success, each person owning one thing. I hired for the setup we needed next, instead of whatever was piling up that week.
- Documented before delegating. If you hand over a process nobody has written down, you mostly hand over stress. So I wrote things down first, then let go of them properly.
What happened
- 21 → 9 days Average project turnaround
- 45+ D2C brands, incl. boAt, FNP, Miduty
- ₹1Cr+ ARR, from zero
- 95%+ Client satisfaction, held while turnaround halved
- 20–30 min Of my day it needs now. Sometimes less.
What I’d tell you in the interview
Most of our growth came from referrals. I’m not good at sales and I won’t pretend I am. Where I’m useful is after the client says yes.
-
The founder was the operating system
The problem
Every escalation, pricing call and delivery problem went through the founder, so the business couldn’t grow faster than one person’s calendar. Cash came in slowly, and nobody could say which role was responsible for what.
What I did and what happened
What I did
- Split the business into four systems. Sales, onboarding, delivery and account management, with clear handoffs between them. A lot of the chaos turned out to be in the handoffs.
- Wrote SOPs and one owned number for each of seven roles. Once every role had a written process and one number it owned, people stopped having to ask who was responsible for what.
- Put billing on a fixed cycle. Invoices went out on a schedule, receivables got checked every week, and chasing payment became a step in the process instead of something someone had to remember.
- Built a second line and pulled the founder out of eight workflows. Other people took over decisions the founder used to make. By the end, eight recurring workflows ran without them.
What happened
- 9 → 3 days Client onboarding
- 34 → 10 days Cash collection cycle
- −62% Overdue receivables
- +18% Client retention
- 8 Recurring workflows off the founder
What I’d tell you in the interview
This was the first time I’d done this inside someone else’s company, and it’s where I realised the thing I’d been doing for years had a name: operations. I want a lot more of it.
-
The bookkeeping I refused to keep doing
The problem
Running an agency meant the same finance admin every month: categorising, reconciling, GST, ITR. The books, the compliance and the actual work sat in three places that didn’t talk to each other. It had been annoying me for three years.
What I did and what happened
What I did
- Built the automation for myself first. Python scripts and workflow automation for categorising, reconciling and reporting, so I didn’t have to do the monthly clean-up by hand any more.
- Realised every agency and freelancer had the same problem. So what I’d built to fix my own problem turned into a product.
- Partnered with CA firms for the parts that need humans. Bookkeeping, GST and ITR are done with CA firms, because software on its own can’t handle Indian compliance properly.
What happened
- 7 months Concept to launch
- Python I built it myself
- CA firms Compliance partnerships
Launched. It’s early days and I’m still learning.
What I’d tell you in the interview
I’m not a hardcore programmer. I know enough to build what I want, and AI tools have made that a lot easier than it was two years ago.
03 — An experiment
Fix the agency
A small game about the elevraa problem. You get the business as it was in February 2026, twelve weeks and twelve possible moves. Six are what I actually did. The other six are things we were tempted to do instead. See if your quarter beats mine.
The short version: a LinkedIn outreach agency where the founder was the operating system. In one quarter, onboarding went 9 → 3 days, the collection cycle went 34 → 10 days, overdue receivables fell 62%, retention rose 18%, and eight recurring workflows left the founder's calendar for good.
The moves, in order: Map the business as four systems → SOPs and one owned KPI per role → Install billing discipline → Build the second line of leadership → Weekly reporting cadence.
Read the full case studyWeeks left
12
-
Client onboarding Onboarding 9 days
How long from signed to running.
-
Cash collection Collection 34 days
Invoice sent to money in the account.
-
Founder load Founder 8 flows
Recurring workflows that still route through one person.
-
Client retention Retention 72 %
Clients still there after six months.
Your moves
Pick a move. Each one costs weeks and you can’t afford all of them, so you have to choose.
Your score
0
You vs. what happened
| Meter | Start | You | Actual |
|---|
What each move actually did
The line I played
- Map the business as four systems
- SOPs and one owned KPI per role
- Install billing discipline
- Build the second line of leadership
- Weekly reporting cadence
Twelve weeks exactly. The order mattered more than the moves. SOPs written before the systems are mapped just write down the mess, and handing over a process nobody wrote down mostly hands over stress.
04 — What I notice
Things that get my attention
Tick the ones that sound like your company. Honestly, these are the things I look for in my first week anywhere.
And, honestly, what I’m not
- I don’t enjoy selling, cold outreach, or convincing people again and again.
- I don’t enjoy doing the same task every day, or being handed a checklist every morning.
- I’m lazy in a specific way. If I have to do something every day for a year, I’d rather spend a week getting rid of it.
Tools I reach for
- n8n
- Zapier
- Python
- Claude Code
- spreadsheets
- Notion
- ClickUp
- Shopify
I use whatever fits the problem.
0 of 8 sound like your company.
05 — Lab
Small tools I build for fun.
Mostly for me and the people around me: something is annoying, a spreadsheet will not cut it, so I build the thing.
-
Business tool
Knox Finance
Knox's money lived in a bank account, an invoicing tool and a Wise account, and none of them could tell me how long the cash would last. Now one page does, and it keeps itself current from my phone.
-
Personal tool
Expense Pulse
Most of what I spend is the same few meals. Logging a ₹30 breakfast shouldn't take a sentence to Claude, so now it takes two taps on my phone.
-
Business tool
Knox Project Management System
Running an agency means quoting, chasing, tracking and billing every single day. I built one place that does all four, so I stopped doing them by hand.
06 — What’s next