Explainer · AI
AI agents vs automations vs chatbots: which one a 12-person plumbing shop should pay for
Three kinds of tool, three price models, three ways to fail. If you can draw the job as a flowchart, you want an automation, not an agent.
The short version
- What happened
Vendors now sell three different things under the word AI: chatbots that answer, automations that follow fixed rules, and agents that choose their own steps. Anthropic's engineering guide defines the split, and Zapier's October 2026 pricing page sells all three as separate products.
- What it means for a 10-to-50-person business
The label decides the bill and the risk. A chatbot (ChatGPT, Claude) answers questions and drafts text, priced per seat. An automation (Zapier, Make, n8n) moves data between apps on fixed rules, priced per task or run. An agent decides which steps to take, priced per seat plus usage. Take a 12-person plumbing company in Tulsa that gets 300 web leads a month. An automation that creates the job in Housecall Pro, texts the customer and posts to the office channel uses about 900 Zapier tasks a month. That fits Zapier's 1,500-task Professional tier at $39 a month billed annually. An agent doing the same job costs more, runs slower and will sometimes do it differently. Use agents only where the steps change from case to case.
- What to do
- Try it
List your five most repeated office tasks, automate the ones with fixed steps first, and add an AI step only where a message has to be written.
Want help deciding whether to build or buy this? J9 Systems, which publishes Back Office Beat, does that for owner-run businesses. Book a call with J9. (J9 Systems publishes this site.)
- What it costs
What it costs - Zapier Professional, 1,500 tasks (annual)
- $468/yr
- ChatGPT Business, 2 office seats × $20/mo (annual)
- $480/yr
- Building three automations, owner or contractor time
- 6-10 h
- Doing nothing
- ~4 h/week of re-keying leads and texting customers by hand
- Sources
- anthropic.com (primary) · Zapier pricing (Professional, Team, Agents add-on) · Make pricing · n8n pricing · What agentic AI means for a small business
The short version
- A chatbot answers. An automation follows your rules. An agent picks its own steps. They are priced differently and fail differently.
- Most of the hours in a small office are lost to copying data between apps. That is automation work, and it is the cheapest of the three.
- Add AI inside an automation where a message has to be written. Reach for a full agent last.
1. What each one is
Vendors put "AI" on everything, so start with what the tool does, not what it is called.
Chatbot. You type, it answers. ChatGPT, Claude and Microsoft Copilot Chat are chatbots at heart. They draft emails, summarise documents and answer questions. They do not touch your other systems unless you connect them. Priced per user seat.
Automation. A fixed recipe. "When a web form comes in, create a customer in Housecall Pro, text them a confirmation, and post the lead to the office channel." Zapier, Make and n8n sell this. It runs the same way every time. Priced by volume: Zapier counts tasks, Make counts credits, n8n counts executions.
Agent. You give it a goal and access. It decides the steps. "Go through yesterday's leads, figure out which are real, and reply to the real ones." Claude Cowork, ChatGPT workspace agents and Zapier's Agents add-on are in this group. Priced per seat plus usage.
Anthropic's engineers, who build one of the leading models, put the split this way. Workflows are "orchestrated through predefined code paths." Agents "dynamically direct their own processes and tool usage." Their advice is to find "the simplest solution possible, and only increasing complexity when needed." That is good advice from a company that sells the complex option.
2. A side-by-side for owners
| Chatbot | Automation | Agent | |
|---|---|---|---|
| What it does | Answers and drafts | Moves data on fixed rules | Chooses steps toward a goal |
| Example in a trade shop | Drafts a reply to a complaint | New lead creates a job and texts the customer | Reads every lead, decides which are spam, replies to the rest |
| Price model | Per seat, about $20-$25/user/mo | Per task, credit or run | Per seat plus usage credits |
| How it fails | Wrong answer you can see | Breaks loudly when an app changes | Quietly does something different |
| Who should own it | Each user | One person in the office | One person, with approvals on |
3. Worked example: a 12-person plumbing company
A 12-person plumbing company in Tulsa runs Housecall Pro. It gets about 300 web and Google leads a month. Today the dispatcher copies each lead into Housecall Pro, texts the customer, and tells the owner. That is 4 to 5 minutes per lead, so about 22 hours a month.
The automation version. One Zap with three actions: create the customer and job, send the text, post to the team chat. Zapier counts each successful action as a task and does not count the trigger. So 300 leads × 3 actions = 900 tasks a month. As of October 2026, Zapier's Professional plan lists 1,500 tasks for $39 a month billed annually ($58.50 monthly). That is $468 a year to save about 20 hours a month.
Adding AI where it earns its place. Some leads are spam, and some say "water heater leaking into the basement." Add one AI step that tags urgency and drafts the first line of the text. That adds tasks and maybe an AI seat for the dispatcher. It does not need an agent.
The agent version. An agent could read each lead, decide what to do, and do it. It would cost seats plus usage, be harder to audit, and occasionally text a customer something you would not. For 300 leads with the same three steps, the agent is the wrong tool.
4. When a real agent makes sense
Use an agent when the steps change case by case and a person would have to think. Examples:
- Month-end close for a small accounting firm, where each client's books need different checks.
- Reviewing 40 subcontractor insurance certificates and flagging the expired ones.
- Researching a commercial bid package and listing what is missing.
Even then, keep it draft-only at first. Have it propose; let a person approve.
5. What it costs
As of October 2026:
- Zapier Professional starts at $19.99 a month for 750 tasks billed annually, or $29.99 monthly. The Team plan, which adds shared workflows and up to 25 users, starts at $69 a month for 2,000 tasks billed annually. Agents and Chatbots are separate add-ons with their own tiers.
- Make lists Core at $12 a month for 10,000 credits, with most actions costing 1 credit. AI features can use more.
- n8n Cloud Starter is €20 a month billed annually for 2,500 executions. The Community edition is free to self-host.
- ChatGPT Business and Claude Team standard seats are each $20 per user per month billed annually, $25 monthly.
Our Zapier explainer has a full cost table for 10-to-50-person shops.
6. What to ignore
Ignore demos where an agent books a job, orders parts and invoices the customer in one go. Ask what happens when it gets step two wrong. Ignore "AI-powered" on a plain automation product; it changes nothing about how you should buy it.
7. What to do
- Write down the five tasks your office repeats most. Note minutes per task and times per week.
- Mark each one: fixed steps, or needs judgment.
- Automate the fixed ones first. Price them in tasks before you buy.
- Add an AI step only where a message must be written.
- Try an agent on one judgment-heavy job, with approvals on.
Sources and method
- Anthropic, "Building effective agents," December 19, 2024: https://www.anthropic.com/engineering/building-effective-agents
- Zapier pricing page, checked October 5, 2026 (Professional, Team, Agents and Chatbots add-ons, overage terms): https://zapier.com/pricing
- Make pricing page, checked October 5, 2026: https://www.make.com/en/pricing
- n8n pricing page, checked October 5, 2026: https://n8n.io/pricing/
- Claude Team plan pricing, Claude Help Center, checked October 5, 2026: https://support.claude.com/en/articles/9266767-what-is-the-team-plan
- ChatGPT Business seat pricing, OpenAI Help Center, checked October 5, 2026: https://help.openai.com/en/articles/8792828-what-is-chatgpt-business
- The Tulsa plumbing company is a composite worked example. Lead volume and minutes per lead are assumptions; use your own counts.
The verdict
List your five most repeated office tasks, automate the ones with fixed steps first, and add an AI step only where a message has to be written.
Questions owners ask
- What is the difference between an AI agent and an automation?
An automation follows steps you defined in advance: when a form comes in, create a job, send a text. An agent is given a goal and decides the steps itself. Automations are cheaper and predictable. Agents handle messier jobs but make different choices on different days.
- Is a chatbot the same as an AI agent?
No. A chatbot answers and drafts when you ask. An agent also takes actions in your other software, like updating a record or sending an email.
- Do I need AI to automate my small business?
Usually not for the first round. Most office time goes to copying data between apps, which plain rule-based automation handles. Add AI where a step needs reading or writing, like summarising a lead or drafting a reply.
- Is Zapier an AI agent?
Zapier's core product is rule-based automation priced by task. As of October 2026 it also sells a separate Agents add-on and a Chatbots add-on, each with its own pricing.
- Which is cheapest: chatbot, automation or agent?
For routine data moves, automation is cheapest per job. A chatbot seat is $20 to $25 per user per month on ChatGPT Business or Claude Team. Agents add usage charges on top, so they are the hardest to budget.
Three to five of these, every weekday morning.
Related
Explainer · Oct 5, 2026
Agentic AI for a 10-to-50-person business: what Claude and ChatGPT agents do in October 2026
Anthropic and OpenAI now sell agents that log into your tools and do multi step work. Here is what that…
Explainer · Oct 5, 2026
The AI for Main Street Act passed the House 395-14 in January. It has sat in a Senate committee since. Here is what it would change
H.R. 5764 would add AI guidance and training to what Small Business Development Centers already do for free.…
Explainer · Oct 5, 2026
AI phone answering for trades: the missed-call math for a 14-tech HVAC shop
Before you buy an AI answering add on, count your missed calls. For a 14 tech shop, 22 lost jobs a month at…
Explainer · Oct 5, 2026
AI receptionists cost $29 to $500 a month. A human answering service starts at $250 for 50 minutes. When each one wins
AI phone answering is now an add on in Jobber and a standalone product from Smith.ai and Goodcall. Here are…