What an AI agency needs from a client, and what you get back

Why this is published before the first call

Almost every engagement that goes badly went badly for a reason both sides could have seen in week one. The terms were fine. The way of working was not what somebody expected, and by the time that surfaced there was a build in flight and a monthly invoice attached to it.

So the conditions are here, in public, before anybody books anything. Read them and you will either think that is exactly how I want to work, or that is not for me. Both of those are good outcomes and one of them saves you a great deal of money.

None of this is a complaint about anyone. It is a description of a method that produces consistent output, and consistency has requirements on both sides of the table.

The four conditions this works under

Everything below is checkable before you commit to anything. If all four are true, the work tends to go quickly and the output tends to be good. If one is not true yet, that is usually a thing you can fix rather than a permanent no.

• The decision maker is you, or the person who decides is in the room. A build generates a decision roughly every other day. When those queue behind somebody who was not in the audit, the project stops moving and nobody notices for a fortnight.
• Payment lands on the agreed date without being chased. Everything we run costs money the day it runs, which is why late payment stops the output that month rather than being absorbed quietly. That is stated up front rather than discovered.
• Your team will follow a written procedure instead of doing it their own way. Every workflow has a documented procedure, a named owner and a check before it goes live. A team that works around it and then holds the system responsible for the result will get a bad result.
• You will answer honestly in the audit, including the numbers you are not proud of. Vague answers about a vague problem produce a generic system. This is the one part of the work we genuinely cannot do for you.

What you get from us in return

A page that only lists obligations in one direction is a page worth ignoring. Here is the other column, and it is enforceable in the same way, because most of it arrives in writing.

The first call costs nothing and always will. The audit produces a written roadmap that is yours whether or not you build with us, which means you can take it to another supplier if you want to.

• A written diagnosis, the bottlenecks in priority order, and an integration report saying what your tools will and will not permit, before you spend money rather than after.
• No number before a diagnosis. A price quoted before anyone has looked at your data and your tools is a guess, and we would rather make you wait a week than revise it in month two.
• A plain statement of what a system cannot do, said before we take money for it. If a tool genuinely will not let an outside system write a record, you hear that in the audit.
• Every workflow labelled automated or augmented, so nobody is guessing whether a person was meant to be involved.
• Training for the people who will use it, delivered as each piece lands rather than in one long session at the end, plus written procedures and recorded walkthroughs.
• Admin access in your name, and a document listing every account, key and connection the system depends on. On every model except renting the output, that is yours to keep.

Why the written procedure matters more than it sounds

Every workflow gets a documented procedure, a named owner and a check before it goes live. That sounds like paperwork and it is the reason the output is the same in month nine as it was in month one.

It also has a cost, and the cost falls on your team. Somebody has to read the procedure, follow it, and say so when it stops matching reality. In return, a new starter is useful in days rather than months, and the knowledge stops living inside one person who might leave.

Businesses that run on people remembering things find this genuinely uncomfortable at first. Most of them find, a few months in, that the discomfort was the point.

The output is consistent month after month because the procedure is written down, and that is also why we suit some businesses badly.

The audit is where this is won or lost

Everything good in an engagement depends on how openly the audit goes, and the difference between the two versions of it is stark enough to be worth naming.

Generic answers in looks like this. Half an hour squeezed between meetings. The problem described in one vague sentence. Nobody wants to show the messy spreadsheet. Numbers guessed rather than looked up. What comes out is a generic system, and then it is nobody's fault and everybody's disappointment.

Real detail in looks like this. Proper sessions with the people who actually do the work. The awkward parts, the workarounds and the arguments. Access to see how the tools are really being used. Real numbers, including the embarrassing ones. What comes out is built around your actual business, and your team uses it because it matches how they already work.

So we will say this plainly. If there is no time to sit through a proper audit, we are probably not the right fit yet. Better to hear that now than three months into a build.

When we are the wrong company for you

There is a version of a business this genuinely does not suit, and describing it honestly is more useful than pretending otherwise. If several of these are true today, close the tab with our blessing. The door stays open when that changes.

You want a system but you are not willing to change a single habit. You expect AI to fix a broken process, when what it will actually do is perform the broken thing faster and more often. Nobody internally will own it and everyone assumes somebody else is handling it.

You want the price before the diagnosis and you will judge us on the number. You pay late and call it cash flow, which means the output stops in the months where it matters most. Or you are hoping this replaces people rather than freeing them up, which is not what we build and not what we would enjoy building.

Being talked out of a purchase costs you an afternoon. Being talked into the wrong one costs a great deal more, and we would rather lose the deal here than in month four.

Who is on this side of the table, and what you can hold us to

The four conditions above are ours to ask, so it is fair to say who is asking. Moiz Khan owns automation architecture and decides what gets built and how it runs. Haad owns growth and client solutions and takes the audit conversation, which is where the honest answers are needed. Ibrahim owns build and workflows and trains your team as each piece lands. Ali owns marketing and sales. None of the four arrived from a consultancy. The company works from Karachi, bills month to month rather than annually, and is answerable for what it operates across 25 engagements in six countries, which is the arrangement that makes conditions in one direction reciprocal rather than one sided.

Two things belong in the return column that are not there yet. There is no price on this site, and that is a condition of ours as much as a policy, because the figure moves on your volume, your tool count, how much of your process exists in writing and who maintains it afterwards, and none of those can be guessed from an industry name. And the timing is published rather than negotiated: the audit takes the first week, one named system is live inside a fortnight, the second and third land across month two. Where a build will take longer than that, you hear it in the audit rather than in month three.

The labelling condition cuts both ways, so here is our half of it plainly. Anything published in your name, any change to advertising spend, any pricing decision and any serious complaint stop and wait for human approval. An automated workflow meeting something outside its scope hands it to a person with the record attached rather than guessing at it. You see the label on every workflow before you approve the workflow. If a supplier tells you all of it can run unattended, the useful question is which part they would put their own name on.

Ownership is the last of it and it runs in your favour deliberately. Admin access is in your name, the accounts are yours, and the document listing every key and connection is handed over rather than held back. On every model except renting the output you own the system, which means you can take it in-house or to another supplier without asking permission. That is arranged on purpose, because a client who cannot leave stops being asked whether the work is still worth buying, and this only holds together while that question is live every month.

Common questions

What does an AI agency actually need from a client?

Four things, in practice. Someone with authority to decide, available through the build. Payment on the agreed date, because the systems cost money the day they run. A team that will follow the written procedure rather than working around it. And honest answers during the audit, including the numbers and workarounds nobody enjoys showing. Access to the tools is the easy part.

How much of my team's time does a build actually take?

The heaviest week is the audit, which is sessions with the people who do the work. After that it is light: approving access, making a decision when one is put in front of you, and short training as each piece lands. What we do ask for permanently is one person willing to own the thing internally, even if that is an hour a week.

What happens if we pay late?

On the managed and rent models the output stops that month. That is not a penalty applied in anger, it is arithmetic. Everything running costs us money on the day it runs, so an unpaid month is a month we are funding. It is written into the agreement and stated before anyone signs so it is never a surprise.

We are not technical. Is that a problem?

No. Building, testing and connecting is entirely our side and nobody on your team touches code or manages infrastructure. What is a problem is nobody being willing to own it internally, because a system with no named owner slowly stops being used no matter how well it was built.

Why will you not give a price before the audit?

Because two businesses buying the same system correctly pay very different amounts, depending on how many tools need connecting, how open those tools are, how many people need training and the state of the data on day one. A number given before any of that is looked at is a guess, and a guess that has to be corrected later damages trust more than the wait does.

What if we do not fit the conditions on this page yet?

Then the honest answer is not yet rather than no. Most of these are fixable. Getting the decision maker into the audit is a calendar problem. Naming an internal owner is a management decision. If the process underneath is genuinely broken, the first useful piece of work is fixing or deleting the process, and sometimes that is all the advice you need from us.