Buying a product, building a system, or renting the output

The rule, stated first

Buy a product when your process is close enough to standard that an existing tool already does it, and you are willing to change your process to match the tool. Build when the process is specific to how your business makes money and changing it to fit somebody's software would cost you the thing that makes you worth buying from.

Rent the output when you need the result soon, the work is not going to become part of how the company operates, and you are content for the capability to leave when the invoices stop.

Cost is the wrong first question, because the three options are cheapest at different points on the calendar. A product is cheapest in month one. Renting is cheapest in the first quarter. A build is usually most expensive at the start and stops charging you for the same thing every month afterwards. Which of those matters depends on how long the work will exist, and that is the question worth arguing about.

One more thing to settle before the comparison. Most businesses do not face a clean choice between the three. The common shape is a bought product doing the standard parts, a small built layer holding the parts that are specific to you, and somebody renting you attention while it settles.

Buying a product, and when the product is enough

A product is other people's decisions, already made. Somebody has chosen the fields, the workflow and the edge cases, and thousands of businesses are pressure testing it for you. Where your process is genuinely ordinary, that is an excellent deal, and building your own version of a well made product is an expensive way to end up with a worse one.

The test is whether you can name the part of your process that no product handles. If you cannot, buy the product. If you can, ask a harder question: is that part the reason customers choose you, or is it just a habit nobody has questioned? Plenty of what feels distinctive about a business is an accumulation of workarounds, and adopting the product's way of doing it is an improvement disguised as a compromise.

Where products struggle here is at the edges of this market. Many are built around email and web forms rather than around a WhatsApp thread containing a voice note and a payment screenshot. Some cannot hold prices that are negotiated per customer. Some assume an address format that does not survive contact with how people here actually describe where they live. Those gaps are where the built layer usually ends up sitting.

Building, and the costs that are missing from the quote

Building buys you fit and control, and it bills you for both twice: once to construct it and again, quietly, forever. The construction cost is the one in the proposal. The running cost is the one that catches out owners who priced a one off project.

Model usage is charged by volume, so a busy month costs more than a quiet one. The platform it runs on carries a monthly cost. Neither is usually the large one. The real ongoing line is attention. Templates get rejected, an interface changes without notice, a workflow starts failing silently and produces nothing rather than an error. Something has to notice within the week rather than the quarter, and most systems that fell over were built correctly and then left alone.

There is one factual difference worth knowing before you choose a platform, because it is difficult to reverse later. n8n can be self-hosted, so the instance and its credentials can sit inside infrastructure you control. Make and Zapier cannot be. If keeping credentials and customer data inside your own environment matters to you, that decision is made at the start or not at all.

• Model usage, which rises with volume rather than sitting flat
• The platform subscription, whoever's name it is in
• Someone noticing within the week when a workflow fails quietly
• Changes when your process changes, which it will
• The first month of unglamorous work cleaning records the system will inherit

Renting the output, which is a real option nobody names

The third option gets discussed as if it were a failure of nerve, and it is a legitimate purchase. An agency keeps the machinery and sends you results. You get the outcome quickly, you carry no maintenance, and when you stop paying the capability goes home with them.

That is the right trade in specific circumstances. When you are testing whether a market exists at all. When the work is seasonal and the machine would sit idle for eight months. When nobody inside the business has the time to own anything and pretending otherwise would produce an expensive shelf. Renting fails when the work turns out to be permanent and central, because you have then been paying a subscription for years to avoid owning something you use every day.

The way to keep the option honest is to write down, at the start, what would make you stop renting. A volume, a date, a moment when the work becomes routine enough to bring inside. Rentals that nobody agreed an ending for tend not to have one.

If the value disappears the month you stop paying, you rented it. That is fine, as long as you decided to.

A test you can run this week

Take the process you are thinking about and answer four questions on one page. How long will this work exist in roughly its current form: a season, a year, or as long as the business does? Is the way we do it the reason customers pick us, or an accident of history? Who inside the business will own this once it runs, by name? And what does it cost us today, measured in the units we already use?

The answers usually decide it without an argument. Short life and ordinary process means buy or rent. Long life and a process that is genuinely yours means build, and build the specific part only. Long life, ordinary process, and nobody to own it means buy the product and spend the saved effort on getting people to actually use it, which is where that money goes anyway.

If you cannot answer the fourth question, stop and measure for a fortnight before deciding anything. A comparison between three options is worthless when the current cost of the fourth option, carrying on as you are, is unknown.

The three options side by side

Read the last two rows first. What you are left holding when the payments stop is the row that separates these three, and it is the one least often argued about before signing.

Criterion	Buy a product	Build a system	Rent the output
Choose it when	Your process is close enough to standard, and you will change it to match the tool	The process is specific to how your business makes money	You need the result soon and the work will not become part of how the company operates
Cheapest at	Month one	Later, once it stops charging you for the same thing every month	The first quarter
What you are buying	Other people's decisions, already made and pressure tested by thousands of businesses	Fit and control	The outcome quickly, carrying no maintenance
What it costs after launch	The subscription	Model usage, the platform, and attention, which is the real ongoing line	The invoice, for as long as you want the result
Where it fails	At the edges of this market: a WhatsApp thread, a negotiated price, an address nobody formats	Built correctly and then left alone	When the work turns out to be permanent and central
When you stop paying	The tool goes and your process stays changed	It keeps running	The capability goes home with the agency

Cases where neither answer is the right one

Sometimes the honest recommendation is that the process should be deleted rather than automated. Weekly reports nobody reads, approvals that have never once been refused, checks that duplicate a check done downstream. Removing a task beats buying anything, and no vendor is incentivised to notice it for you.

There are also two timing conditions that override the whole comparison. If you are part way through moving to a new CRM or accounting system, finish the move before deciding, because building or buying on top of something you are about to replace means paying for the same work twice. And if your busiest season starts within a few weeks, map the process now and install after the peak rather than rebuilding the machine while it is running.

Finally, if the real blocker is that two departments disagree about who owns a piece of work, none of the three options addresses it. A product will not settle it, a build will encode one side of it, and renting will hide it for a while. The disagreement is the project, and it is a conversation rather than a purchase.

The question the one page test is missing, and who is behind each option

There is a fifth question for the one page test and it applies to all three options. Who is on the other end when the thing stops? Money, pricing, anything published in your name and any serious complaint have to wait for human approval whether you bought a product, built a system or rented the output, and a case outside scope has to hand it to a person rather than resolve itself quietly. A product answers this with a support queue. A build answers it with somebody inside your business. Renting answers it with the agency, which is the strongest argument for renting and the one nobody making it seems to reach for.

The building option also has a version the article does not separate out, which is building it with your own team rather than buying the build. Where the process is specific, the volume is real, and somebody in-house already automates things and will still be there next year, that is frequently the cheapest of the four and it is rarely recommended by anybody selling. The running cost this page names is the same either way. What changes is who pays it and whether that person can be reassigned to something else.

Wobble sells three of the four and says so. Moiz Khan owns automation architecture and decides what gets built and how it runs. Haad owns growth and client solutions and does the diagnosis that decides which option you are. Ibrahim owns build and workflows. Ali owns marketing and sales. The company is based in Karachi, bills month to month and is answerable for what it operates across 25 engagements in six countries, which is what makes renting the output a purchase rather than a dependency.

Common questions

Should I build a custom AI system or buy an existing product?

Buy when your process is close enough to standard that a product already does it and you are willing to adapt to the tool. Build when the specific way you work is the reason customers choose you. The deciding question is whether you can name the part of your process no product handles, and whether that part actually matters.

Is building always more expensive than buying?

It is usually more expensive to start and cheaper to keep, because a build stops charging you a per seat fee every month while a product does not. Which is cheaper overall depends entirely on how long the work will exist. Over one season a product wins comfortably. Over several years the arithmetic often reverses.

What ongoing costs does a custom AI system have?

Model usage that rises with volume, a platform subscription, and attention. Attention is the one that gets missed. Templates get rejected, interfaces change, and workflows start failing silently, so somebody has to notice within the week. Most systems that fell over were built correctly and then left alone.

Does it matter which automation platform we build on?

It matters when credentials or customer data need to stay inside your own environment. n8n can be self-hosted, so the instance and its credentials can sit in infrastructure you control. Make and Zapier cannot be. That is a factual difference between the tools, and it is difficult to reverse once a build exists.

When is renting from an agency the right choice?

When you are testing whether a market exists, when the work is seasonal and a system would sit idle, or when nobody inside the business can own anything yet. Write down at the start what would make you stop renting, because rentals with no agreed ending rarely find one.

How do I compare the options fairly?

Measure what the process costs you today first, in the units your business already uses, over a period long enough to include a bad week. Without that number every comparison is between three quotes and a feeling, and the option that wins is whichever supplier presented most confidently.