Md Ayaan Ali · Business Operating Systems
Why are we doing it this way?
01
Nobody ever had a good answer.
02
Most companies answer it by buying software.
New licence. Same shape of business. The process survives, now with a renewal date.
03
I redesign the business first. Then I build the system that runs it.
04
Technology is never the hero. AI goes in only where it genuinely improves the system, and comes back out when it doesn't. What I show you is the transformation, not the stack.
↓
Four of those redesigns became products.
Method
No discovery deck. No transformation programme. I sit inside the job until the friction is obvious, then design it out.
Do the job myself. Watch where the hour actually goes — not where the process diagram says it goes.
Separate the real constraint from the habit. Most ceilings turn out to be a misread specification.
Change the workflow before writing a line of code. Automating a bad process only makes it faster.
Machines take the labour. Never the judgement. The human checkpoint sits where an error is most expensive.
Every number traceable to its source. A figure nobody can source is a decision nobody should make.
Every system here is still running, and still being cut back. Removing a feature counts as shipping.
back to observe
Four products
Long buying cycles, high-value orders, and a sale that closes because a designer trusts a person. Every category leader in outbound software is built for the opposite of that. So three of these were built for it instead, and the fourth keeps the promise the sale made.
More conversations. Not more people.
The business operating layer. CRMs record deals. Engagement platforms send messages. Nothing owned the space between — identity, intent, memory and pipeline. Weft is that layer.
Scale the accounts. Not the headcount.
The LinkedIn operating system. The platform's real limit is per account, not per person — which means outbound capacity was never a hiring problem. Nine identities, nine markets, one operator.
Every email personal. Every one remembered.
The email operating system. Thousands of individually written emails from one click — sent from your own name, your own servers, and stopped the instant someone replies.
Know it's late before it's late.
The order visibility platform. A months-long production pipeline, watched by five functions through five different windows. One page now, scoped to each person, mirrored from the ERP every thirty minutes, and flagging the orders that are behind by stage before they are behind by date.
Other systems
Smaller in scope, same method. Each one began because something broke in front of me and nobody else was going to fix it.
01 · Operations
Orders arrive across five marketplaces at once. The system identifies the exact physical piece in the warehouse, routes packing instructions, builds the customer file and retrieves the legal invoice — but only after a human has seen the product photograph.
9,400 ordersprocessed without a single packing instruction leaving before a person confirmed it. That gate is the architecture, not a limitation.
02 · Applied AI
One command becomes a researched, deduplicated, individually written outreach list. A structural gate discards contacts without a reachable address before any model is paid to think about them — so cost scales with results, not with attempts.
12,400 contactsresearched once and shared across the team, replacing 20+ hours a week of manual work per person.
03 · Computer vision
The warehouse already photographed every piece against a printed tape. The measurement was in the picture the whole time. The tool reads the tape, calibrates the scale, reads the tag, and writes verified centimetres straight into the audit record.
Runs offlineEntirely on the warehouse machine. Every uncertain reading is labelled unconfirmed rather than guessed.
04 · Merchandising
Merchandisers were hunting product imagery through nested network drives and pasting the wrong colourway, because one shade's name contains another's. The fix wasn't a smarter algorithm — it was encoding a naming convention the business already followed.
Hours → one clickFull-quality imagery, correct colourway, no recompression, no duplicates.
About
That is the only unusual thing here — and it's what makes everything else make sense.
I am not a software engineer. I'm not a designer, and I'm certainly not an AI consultant. I sat inside a sales and operations function, and every time something broke or wasted an afternoon, I designed the thing that removed it.
The lesson that shaped everything came early, and it wasn't technical. The company wanted more outbound, so it was going to hire — a salary for every additional thirty or forty approaches a day. That is how the whole industry does it. But the platform we were selling through publishes its limit plainly, and the limit is per account. Not per person.
Which meant every one of those salaries was really buying an account. A hiring plan, several years long, was a configuration problem nobody had read carefully enough to notice.
Once you have seen that once, you cannot stop seeing it. Personal and scalable were never a trade-off in email — that was a tooling assumption. Volume didn't have to cost relationships — that was a memory problem wearing a headcount costume. Almost every ceiling I have been asked to raise turned out to be a habit that had hardened into a budget line.
So the method stayed constant as the ambition grew. Read the constraint before you buy your way around it. Automate the labour, never the judgement. Never scale the sending without scaling the memory. Never sum two things that aren't the same thing to get a tidier number. Remove the feature rather than ship a fifth of it.
Over a year that went from automating a single lookup to designing the operating system for how an entire outbound revenue function works. The engineering got harder. The instincts stayed the ones I learned from doing the job — what an operator actually does at nine in the morning, what a warehouse auditor will tolerate, and what a designer's inbox does to a message that was obviously written by a machine.
I still start every project the same way. I watch. I ask why it's done that way. And nobody ever has a good answer.
Contact