Prebuilt or custom AI agents: how to choose

The short version

If your need is common, choose prebuilt. If your process is specific, or it has to work with internal systems, choose custom. If you are unsure, start with the discovery call: the first thing we decide is which route fits.

When prebuilt fits

• The job is common: enquiries, follow-up, support triage, inbox, quotes, reporting
• Your tools are common ones such as email, calendars and popular CRMs
• You want something working quickly and extendable later
• You want predictable scope

When custom fits

• The process has steps that no ready-made agent covers
• The agent must use internal systems or private data
• You need specific approval and audit rules
• The work spans several tools and must run as one flow

What changes between the two

Both follow the same safe foundation: limited access, approval points, logging and testing. The difference is the starting point. A prebuilt agent starts from a design that already exists and is tailored to you. A custom agent starts from your workflow and a written scope.

Questions that decide it

1. Could a new employee do this task from a standard checklist?
2. Do most businesses like mine do this task the same way?
3. Do I need the agent to touch systems that are not mainstream?
4. Do I have special rules about who must approve what?
5. Do I need to prove to someone else what the agent did?

Answer yes to the first two and no to the rest, and prebuilt is likely enough. Answer yes to the last three, and plan for custom.

Many customers start prebuilt and later commission a custom agent for the one process that is unique to them.

Cost and time

Prebuilt agents usually involve less discovery and a more predictable scope. Custom agents involve a written scope and staged delivery. In both cases you should receive a written quote before work starts.

Next step

Book a 30-minute call. Bring one task you would like off your plate, and we will tell you which route fits.