When should we buy rather than build?

When the task is self-contained and standard: answering from a script, booking into one calendar, replying to common questions. If a product does 80% of what you need out of the box and the missing 20% is cosmetic, buy it. Building to avoid a subscription is almost never worth it.

When is building the right call?

When the agent has to read and write in systems you already run, and follow rules specific to your business rather than generic to your industry. That is the line. Below it you are paying for flexibility you will not use; above it, a tool leaves you doing the integration by hand anyway.

Buy the component, build the connection. Use an off-the-shelf voice or chat product for the conversation and build only the integration into your systems. It is often the cheapest path, and it is what most companies of this size end up with whether or not they planned it.

How do we compare costs fairly?

Over three years, including the work a tool leaves you doing manually. A $49 monthly subscription that still needs someone to copy bookings into the CRM for ten minutes a day is not cheaper than a build — it just moves the cost somewhere nobody measures.

What is the risk of building?

Scope. A build with an open-ended scope is how budgets disappear. The mitigation is not to build less carefully but to fix the scope and the price before starting, and to keep the first one to a single process.

Deciding

Build or buy an AI agent

Buy when the task is self-contained and standard. Build when the agent has to read and write in systems you already run, following rules specific to your business. Most companies of 10 to 100 people end up somewhere between: buying the conversation and building only the connection.

By Gorden WübbeAutomation, agents and search visibilityUpdated 10 August 2026

The line between them

There is a clean test, and it is not about company size or budget:

Does the agent need to read and write in systems you already run, following rules specific to your business?

If no — the task is self-contained, the rules are generic to your industry — buy. A product will do it faster and cheaper than anything built for you, and it will keep improving without you paying for the improvements.

If yes, a tool will get you most of the way and then leave you doing the last part by hand, permanently. That last part is usually where the time was going in the first place.

What buying gets right

Live in days, not weeks. For a business with a problem this quarter, that matters more than it looks on a comparison table.
Someone else fixes it. Models change, APIs change, and a vendor absorbs that.
Cheap to abandon. If it does not work you cancel. A build you stop using is money spent.

If a product does 80% of what you need out of the box and the missing 20% is cosmetic, buy it. Building to avoid a subscription is a false economy that shows up two years later as a system nobody maintains.

Where buying quietly fails

The failure is rarely dramatic. The tool works, and someone starts doing a small amount of manual work to bridge it to reality.

Bookings arrive in the tool’s own dashboard and get copied into the CRM. The agent cannot see whether a caller is an existing customer, so everyone gets treated as new. A rule specific to how your business actually operates cannot be expressed, so there is an exception people remember to handle. Individually trivial; together, ten minutes a day forever.

That is the cost to compare against a build, and it is the number that never appears in a vendor comparison because nobody measures it.

The hybrid most companies land on

Buy the component, build the connection. Use an off-the-shelf product for the part that is genuinely standard — voice recognition, chat delivery, the conversation layer — and build only the integration into your systems and the rules specific to you.

This is usually the cheapest honest answer for a company of 10 to 100 people, and it is what most end up with regardless of what they set out to do. Planning for it deliberately is better than arriving there by accident after buying the wrong thing twice.

Comparing the cost fairly

Over three years, and including what each option leaves you doing manually:

Count this	Buy	Build
Up front	Setup, usually small	The build
Monthly	Subscription, often per seat or per minute	Model usage, hosting, monitoring
Manual work left over	Usually the largest item	Close to none if scoped properly
When you outgrow it	Migrate, and rebuild the workarounds	Extend

The real risk of building

Not technical failure. Scope. A build with an open-ended scope is precisely how budgets disappear into a platform nobody uses.

The mitigation is not building more carefully. It is fixing the scope and the price before anything starts, and keeping the first one to a single process — which is the entire design of the 30-day pilot. If a supplier will not fix both, that is the risk you are being asked to carry.

What people ask at this fork

When should we buy rather than build?: When the task is self-contained and standard: answering from a script, booking into one calendar, replying to common questions. If a product does 80% of what you need out of the box and the missing 20% is cosmetic, buy it. Building to avoid a subscription is almost never worth it.
When is building the right call?: When the agent has to read and write in systems you already run, and follow rules specific to your business rather than generic to your industry. That is the line. Below it you are paying for flexibility you will not use; above it, a tool leaves you doing the integration by hand anyway.
What is the hybrid?: Buy the component, build the connection. Use an off-the-shelf voice or chat product for the conversation and build only the integration into your systems. It is often the cheapest path, and it is what most companies of this size end up with whether or not they planned it.
How do we compare costs fairly?: Over three years, including the work a tool leaves you doing manually. A $49 monthly subscription that still needs someone to copy bookings into the CRM for ten minutes a day is not cheaper than a build — it just moves the cost somewhere nobody measures.
What is the risk of building?: Scope. A build with an open-ended scope is how budgets disappear. The mitigation is not to build less carefully but to fix the scope and the price before starting, and to keep the first one to a single process.

Want a straight recommendation?

Thirty minutes on your actual process. If the answer is "buy the tool", you will hear that — it costs us the project and saves you the money.

Book a scoping call