Every vendor calls their thing an AI agent now, but most are still decision-tree chatbots with a nicer font. The difference matters, because one deflects tickets and frustrates customers, while the other actually resolves them. If you're paying for support automation, it's worth knowing which you're buying.
The chatbot ceiling
A scripted chatbot can only answer what someone anticipated and wrote a branch for. The moment a customer phrases things differently or asks something off-script, it falls back to 'let me connect you to an agent.' It feels like automation but it's really just a slower contact form.
What an agent does differently
- Reads your real knowledge: docs, past tickets, and policies, instead of a fixed script.
- Takes action: looks up an order, checks a status, drafts a refund, not just talk.
- Knows its limits: escalates with full context when it genuinely can't help.
- Improves: gaps in its answers become the next thing you document.
Which one you actually need
If 80% of your tickets are the same handful of questions, even a simple assistant pays for itself. If your support involves looking things up and taking action across systems, that's where a real agent, wired into your tools, turns a cost center into something close to self-service. Start by counting your top ten ticket types; the answer is usually obvious once you do.
Related reading
AI-ready data: four tests before you add an AI step
Gartner expects most AI projects without AI-ready data to be abandoned through 2026. The model is rarely the reason. Here is what AI-ready data means for a small business, the workflow that sits under an AI step that works, and four tests you can run on your own data this week.
Sources → Pipeline → One source → AI step → Checks → Person
Everyone sells agents now. Most of what you need is not an agent.
Agentic AI is the loudest pitch of the year. But for most back-office work, a pipeline with one well-placed model step beats an autonomous agent on cost, reliability, and sleep. Here are the three questions that tell you which one you actually need.
Fetch → Extract → Check → Write → Checkpoint
What AI agents can't do yet, and how to design around it
The gap between what agent demos show and what agents reliably do in production is real, specific, and worth designing for instead of ignoring.
Scope autonomy → Short stages → Checkpoints → Ask, not guess → Human confirms
Continues · Got a workflow like this?