Most automation advice assumes a business two steps ahead of yours

Start with capture, not with integration

A great deal of what is written about business automation assumes that the data already exists somewhere structured, and that the job is to move it between systems. Plenty of established Peshawar businesses are not in that position, and there is nothing wrong with that. The orders are real, the money is real, and the record is a ledger and a set of chat threads.

Selling that business a full platform is how money gets wasted here. The system asks for master data that has never been written down, the staff keep using the ledger because it is faster, and within two months there are two records that disagree. What gets blamed afterwards is the technology, when the actual mistake was starting three steps up the ladder.

The right first project captures one thing properly. Usually it is the enquiry: who asked, for what, when, and what was quoted. That single record, built from the messages and calls that already arrive, is enough to answer questions nobody in the business can answer today. How many enquiries came in last month. How many were never replied to. Which customers ask repeatedly and never buy.

Once that exists, the next step is obvious and the business can decide it for itself. Building in that order also means the first thing a supplier delivers is something you can check within weeks, which is the only reliable protection against a project that talks well and delivers slowly.

• One record of enquiries, built from the channels you already use
• One record of what was quoted, with the date it was quoted
• A reply to every enquiry, including the ones that arrive after hours
• A number for how many enquiries went unanswered last month, which nobody has today
• No requirement to abandon the ledger until the new record has proved itself

Consignments, transit and delays that are not yours to fix

Peshawar's trading and transport businesses move goods along routes where the timeline is often decided by something outside the company. A border crossing closes, a queue builds, a clearance takes longer than anyone planned. No system fixes that, and any supplier suggesting otherwise should be shown the door.

What a system can change is what the customer knows and when. In this trade, the damage from a delay is usually smaller than the damage from silence, because a customer who is told on day two makes different arrangements from one who finds out on day six. Most businesses here already know this and still fail at it, because telling forty customers something is a full day of somebody's time.

The record to build is per consignment rather than per customer: what was booked, what documents exist, where it currently is, and what is known about the hold up. When that record changes, the people waiting on it are told. When nothing has changed for longer than usual, that itself is worth surfacing rather than waiting for the customer to ask.

The documents follow the same logic. Copies arrive by photograph and by email from clearing agents, transporters and customers, and they are then hunted for later. Filed against a consignment number on arrival, they stop being hunted for.

Wholesale, where the rate is a phone call and the deal is a relationship

The wholesale trade around Peshawar runs on personal relationships, on rates that move, and to a large extent on cash. This is not a market that wants to be reorganised by software, and a supplier who arrives with a plan to formalise everything is going to be politely ignored.

The parts that automation improves are narrow and specific. An enquiry that arrives when the shop is closed gets an accurate answer instead of waiting until morning. A rate is quoted with the date attached, so a customer cannot come back next week holding an old screenshot. An order is confirmed in writing before it is loaded, which settles most disputes before they start. A running balance can be sent to a customer on a schedule instead of becoming an awkward conversation once a quarter.

Everything else, meaning who gets which rate and who gets extended credit, stays exactly where it is, with the person who has always made those calls. Automation here is administrative support for a trader, not a replacement for their judgement.

Design for voice, for Pashto and Urdu, and for people who do not type

A system that expects English text typed into form fields will not be used by half the people who hold the information in a business here. Staff and customers alike send voice notes, write in Roman script, and switch between languages inside a single message. This is not a problem to be corrected. It is the input format, and the system is what has to adapt.

In practice that means a voice note is a first class input rather than an exception, replies go out in the language the customer wrote in, and any step asking a member of staff for information asks for as little as possible. Where something is ambiguous, the system should ask a short clarifying question in the same channel rather than guessing or dropping the message.

Connectivity deserves the same treatment. Anything that depends on a stable connection at the moment of capture should tolerate a delay, and anything that silently fails when the network drops will lose records nobody knows are missing.

Choosing a supplier when you cannot walk into their office

Wobble is an AI automation company based in Karachi, working with Peshawar businesses remotely. There is no office here and no local team. Saying so plainly matters, because a lot of businesses in this city have already paid for a website or a system from a supplier who promised presence and then stopped answering.

The protection against that outcome is not proximity, it is structure. Ask for a first deliverable you can see working within weeks rather than a long build. Insist that platform accounts are in your own company name with administrator access held by you, so ending the relationship costs you nothing beyond the agreed notice. Get the response commitment in writing. Ask what happens to the system if you stop paying, and listen to how quickly the answer arrives.

A supplier who agrees to all four is safer at a distance than a local one who agrees to none of them.

When capture has to come first, and integration much later

If your business has no written record of prices, customers or stock, and no appetite to create one, most of this is premature. Start with the enquiry record described above and nothing else, and revisit in six months.

If your trade genuinely depends on relationships and negotiation for every transaction, automated quoting is not for you. Capture the enquiry and put it in front of the person who prices. Anything more will produce numbers your customers will hold you to.

If the delays that hurt your business are caused by borders, queues or clearances, understand what you are buying. Better communication, not faster movement. That is often worth paying for, and it is not what most vendors will describe when they are selling.

And if the person who would run the system is the same person who runs everything else, keep the scope to a single workflow. A system with nobody watching it degrades quietly and takes the trust with it.

What the first record costs, how soon you can check it, and who still answers

No price appears anywhere on this site, and on a first capture project the honest thing to say is that it ought to be small. What moves the number is how many channels the enquiries arrive on and whether anything has to be read out of a photograph, rather than how large the business is. The audit takes the first week and the enquiry record is live inside a fortnight. That is deliberate. The point of starting here is that you can check whether a supplier delivered inside a month, instead of believing a roadmap for a quarter.

The status update is where this either builds trust or spends it. A system says what is known, when it was last confirmed and what is still uncertain, and where nothing has changed for longer than usual it hands it to a person rather than sending a reassuring sentence nobody can stand behind. Any rate, any credit decision and any conversation with a customer who is already unhappy need human approval before they go anywhere, because a wholesale trade running on relationships is not improved by a machine being confident on your behalf.

Ownership matters more here than elsewhere, because the whole argument is that you keep the ledger until the new record has earned its place. The enquiry record, the consignment file and the documents filed against it sit in accounts under your own logins from the first day, so you own the system and nothing depends on a supplier still being engaged. The first version is also something your own team can build in-house with a shared sheet and a rule about writing every enquiry down. If that holds for a month, you have already answered the question the rest of this page is asking.

The published work behind this is the ordinary kind rather than the impressive kind: three ERP systems for RM Gulistan Engineers, an engineering and contracting business in Karachi, and a delivery line rebuilt as one audited automation for Quillon. There is no Peshawar trading house on the work page, which is worth saying plainly rather than implying otherwise. Wobble works from Karachi and remotely, bills month to month, across 25 engagements in six countries.

Common questions

Our business is not computerised at all. Can we still automate anything?

Yes, and the sensible starting point is different from what most vendors propose. Rather than installing a platform that expects master data you have never written down, build one record of enquiries from the channels you already use: who asked, for what, when, and what was quoted. That answers questions nobody in the business can answer today, and it can be checked within weeks.

Can a system handle Pashto and Urdu voice notes?

It has to, because that is the input format in this market rather than an exception to it. Voice notes are treated as a first class input, replies go out in the language the customer wrote in, and staff facing steps ask for as little information as possible. Where a message is ambiguous the system should ask a short clarifying question in the same channel rather than guessing.

Can automation speed up a consignment held at a border?

No, and any supplier who suggests otherwise is selling something they cannot deliver. What changes is what your customers know and when. A per consignment record with documents filed against it, updates sent when something genuinely changes, and a flag when nothing has moved for longer than usual, all reduce the damage from a delay without pretending to shorten it.

Do you have an office in Peshawar?

No. Wobble is based in Karachi and works with Peshawar businesses remotely, with no local team. Given how many businesses here have been let down by a supplier who promised a local presence, the useful protections are structural: a first deliverable visible within weeks, accounts in your name with administrator access held by you, a written response commitment, and a clear answer about what happens if you stop paying.

How do we avoid paying for something that never gets used?

Judge the first project by whether staff use it without being told to. If people keep using the ledger or the old chat thread, the system asked too much of them. Keep the first build small enough to fail cheaply, watch what actually happens for a month, and expand only into the places where the first piece is holding.

What does it cost to start?

It depends on scope, and no honest figure fits every business, which is why none is printed here. What you should insist on is a scoped number before work begins, split into an installation cost and a monthly cost, and a first phase small enough that you can judge the supplier on delivered work rather than on a proposal.