AI customer service: build the agent, buy a bot platform, or implement

To put AI into customer service, there are three paths. Buy a bot platform. Build the agent internally. Hire someone who implements.

The question “which bot is best” is the wrong one. The right question is who answers for the customer when the agent errs, and who operates the channel a year and a half from now.

The three paths, as cost shapes

The difference is not in the feature list. It is in the shape of the cost and the responsibility that remains.

Buy a bot platform. Monthly license by service volume, predictable. Take Blip, Zendesk, Intercom, and Zenvia compete for that pocket. It turns on fast. The agent is generic by design, tuned for the platform’s average client, not for your operation.

Build internally. Cost in an engineering and product team. Variable, slow, permanent responsibility. It makes sense when service is a differentiator and crosses stock and order in a way no off-the-shelf bot covers.

Hire an implementer. Project cost, fixed by scope. The internal team keeps the agent and the runbook. It builds against your stock and your flow, with a defined action limit, and leaves.

The decision tree

The deciding condition is operational: who answers for the customer at 2am, 18 months from now?

• If the answer is “the bot vendor”, buy the platform and accept generic service.
• If it is “the consultant”, you did not finish buying.
• If it is “my team, with criterion and limits written before the code”, you want an implementer.

The trap in each path

The bot platform fails when generic service becomes the problem, not the solution. That is what Klarna lived: the AI delivered generic answers and the company began rehiring humans. The build fails when it becomes an ownerless project. The implementer fails when the service-quality criterion stays vague and the work stretches.

The defense, on any path, is the criterion: what the agent can promise, when it hands off to a human, and the acceptable resolution rate. Without it, any of the three delivers volume and not trust.

What the choice decides

Take the service decision on your desk. Ask the team which path you assumed last time. Almost always nobody chose; you bought the bot the competitor used.

Tell me the shape of the problem, service volume and the kind of question that arrives most. In one hour I will send back which path fits, with no bias. If it is a platform, I will point you to the names that show up in service RFPs. If it is an internal team, the profile to hire. If it is an implementer, the next conversation is the two-week Diagnóstico, with criterion, timeline, and cost signed before the code.