Conversational AI platform vendor or full-service implementation firm

Buy a platform when your process looks like everyone else's in your industry. Hire an implementation firm when the part that matters is specific to your company and exists nowhere else. The deciding question is not capability or price, it is whose process you are automating. If the answer is "the industry's", a platform will be cheaper and faster, and we will tell you so.

Both options answer the same request and arrive at very different bills, which is why the comparison usually gets framed as build versus buy. That framing is close but slightly wrong, because a platform is not "buy" and a firm is not "build". A platform is a product with a configuration layer. A firm is people writing software for your case.

The useful distinction is what happens when your process does not fit the configuration layer. With a platform, you adapt your process to the product, and often that is fine. With a firm, the product adapts to you, and you pay for that adaptation twice: once to build it, once every year to keep it running.

Neither is a mistake. Choosing without knowing which one you are doing is.

You are not comparing two vendors, you are comparing two purchases

A platform sells you access to something that already exists. The vendor's economics depend on many customers using roughly the same thing, so the roadmap belongs to them, the integrations are the ones they chose, and support means help using the product as designed.

An implementation firm sells you hours applied to your situation. The economics depend on the work being specific, so the scope belongs to you, the integrations are the ones your business needs, and support means changing the thing when your business changes.

These differences show up late. Both start with a discovery call, a demo and a proposal, and at that stage they look like competing quotes for the same job. The divergence appears at the first requirement that was not in the original scope, because that requirement is either a configuration setting or a small project, and which one it is was decided when you signed.

Ask that question during evaluation. Take the least standard thing your business does and ask how it would be handled.

The question underneath: whose process is this

Take the workflow you want automated and try to describe it to someone in the same industry. If they nod along, you are describing an industry process. If you find yourself explaining exceptions, you are describing yours.

Industry processes are where platforms are strong and where custom work is close to indefensible. Booking appointments, answering catalogue questions, routing tickets by category, collecting contact details: thousands of companies do these almost identically, a product has already solved them, and paying someone to write that from scratch means paying for a solved problem.

Company-specific processes are the opposite. They usually involve your own rules about what happens in edge cases, your own sequence of approvals, or a connection to a system your industry does not standardly use. They resist configuration because they were never in the product's model of the world.

Most businesses have both, which is why the honest answer is often "a platform for the standard part, and a decision about whether the specific part is worth building at all".

Where platforms stop

In one of our projects a system answered questions, presented services and booked meetings. It worked, and the client was satisfied. Then the scope grew by something that looked like more of the same: the system was to complete transactions inside a partner company's tool, under a commission arrangement.

It was not more of the same. The flow was specific to that business and did not assemble from ready-made parts, so each step had to be written for that one case. We said it was no longer a configuration job and proposed software written for the process instead.

We could have assembled it from components. It would have demonstrated well. It would also have produced something that completes the operation correctly most of the time and almost correctly the rest of the time, with nobody able to tell which run was which. On a note field that is survivable. On a commission-bearing transaction it is not.

The boundary is not difficulty. It is the moment a system stops talking and starts changing state in a system of record, especially one you do not control.

What you own when it ends

Platforms end by cancellation. When you stop paying, you stop having the capability, and what remains is whatever data you exported. This is a normal trade and for standard processes it is a good one, since you also stop paying for maintenance, hosting and everyone else's roadmap.

Custom implementations end differently. You keep the code, the infrastructure accounts and the integrations, and you also keep responsibility for them. That responsibility has a price whether or not anyone is currently working on it, because providers change how their systems connect, your own tools change, and each change needs a fix.

The mistake we see most often is not choosing wrongly between the two. It is choosing custom for its ownership benefits and then not budgeting the year that follows, which produces a system that works beautifully for eight months and quietly rots in the ninth.

Ask any firm what the second year costs before you compare it with a subscription. A quote that covers only the build is not a comparison, it is half of one.

We are not the cheap option here

Our own position in this market is mid-priced, not low, and pretending otherwise would be a promise we could not keep. Published pricing across ten providers serving small and mid-sized businesses in this segment, as listed in August 2026, runs from around 1 000 USD for a basic build to 250 000 USD at the enterprise end, with monthly retainers between roughly 200 and 5 000 USD. Our build starts around 9 000 USD and our monthly plan starts around 1 000 USD. That places our retainer above most of that set. These are published starting prices from a market scan, not quotes, and they move, so treat them as a shape rather than a table.

There is one position in that market nobody should try to beat on price. At least one provider sells a dedicated developer at 25 USD an hour on a monthly arrangement. If cost per hour is your deciding factor, that is the rational choice and we will not pretend otherwise.

What we are useful for is narrower. It is the case where the specific part of your process is the part that matters, where a wrong write into a system of record is expensive, and where you want the person who built the thing to still be reachable when it needs changing.

If that is not your situation, buy the platform. It will be cheaper and it will be faster.

Before you get in touch

We do not promise a payback period and we will not put one in a first meeting. If your process matches how other companies in your industry work, we will tell you to buy the product rather than commission one.

If the part that matters is yours alone, the first step is a paid review with a document as its output, not an implementation. You can act on it with us or without us.

See what that first step covers: AI implementation review.