By Md Abdur Rouf Mia | Updated 21 October 2026 | Tidio and Intercom prices checked 21 October 2026
AI agent vs chatbot for small business sounds like a vocabulary question, but it is really a buying question. Vendors use both words for products that range from about $24 a month to several hundred dollars a month, and the label on the box rarely tells you which one you are paying for. This guide explains the difference in plain English, compares the two on seven points, prices each option at 1,000 conversations a month, and ends with a decision flowchart and three starter stacks.
The short answer: a chatbot talks, and an agent acts. A chatbot answers questions inside a conversation. An AI agent is given a goal, chooses its own steps and uses tools, such as your calendar, CRM or helpdesk, to change things in other systems. For most small businesses the right first purchase is a chatbot or a simple rules-based flow. An agent earns its place when the job ends with an action in another system and a wrong action is cheap to undo.
Here is the cost picture before the detail, for a business with 1,000 customer conversations a month:
- A rules-based chat flow starts at $24.17 a month (Tidio Flows, annual billing).
- A managed AI support agent billed per outcome (Intercom Fin, $0.99 per outcome) comes to about $326 to $722 a month, one seat included, if it resolves 30% to 70% of chats.
- A do-it-yourself workflow agent costs about $12 to $80 a month in model and server fees in the Beeznez Tech cost model, before build time, platform plans and review time.
A note on evidence. Tidio and Intercom prices were read from their official pricing pages on 21 October 2026. Voice-agent and model-cost figures come from two earlier Beeznez Tech cost guides, linked where they are used. The Fin and do-it-yourself totals are arithmetic on published prices, with every assumption written out so you can change it. This guide did not involve hands-on testing of these products, and where a number comes from a vendor claim or a third-party review rather than an official page, the text says so. There are no affiliate links in this article.
AI Agent vs Chatbot for Small Business: The 60-Second Definitions
Two words hide three different products, and the middle one is where most of the confusion lives.
| Type | What it does | Who decides the next step | Typical example |
|---|---|---|---|
| Rules-based chatbot | Follows a script of buttons and fixed replies | You, in advance | A flow that asks three questions, then collects an email address |
| AI chatbot | Understands a free-text question and writes an answer from your content | The model chooses the wording; you chose the content | A chat widget trained on your help pages that hands off to a person when unsure |
| AI agent | Works toward a goal, picks steps and calls tools in a loop | The model, inside limits you set | A support agent that looks up an order, checks your refund policy, drafts the refund and asks you to approve it |
The most useful definition comes from Anthropic’s engineering team. It separates workflows, where models and tools run along predefined code paths, from agents, where the model steers how a task gets done and which tools it uses. The difference is who holds the steering wheel: you, or the model. It is one respected definition rather than an industry standard, but it is the clearest one available.
Between the AI chatbot and the true agent sits a fourth option that many small businesses actually want: an automation workflow with one AI step in it. A form arrives, an AI step classifies it, and fixed rules route it to the right person. The path is fixed and only one step uses a model, so it is cheaper, more predictable and easier to audit than an agent. Anthropic’s own advice is to find the simplest solution that works and add complexity only when needed. The guides to AI automation for small business and the no-code AI agent pipeline cover this middle ground.
One analogy to keep. A rules-based chatbot is a printed menu. An AI chatbot is a waiter who has memorised the menu and can answer questions about it. An agent is a waiter who can also walk into the kitchen, change your order, apply a voucher and come back with the bill, which is exactly why you want to be sure what they are allowed to do.
Why the Labels Are So Confusing (and What Agent Washing Means)
Gartner gave the problem a name in June 2025. It said many vendors were guilty of agent washing, which it described as rebranding existing assistants, robotic process automation and chatbots without substantial agentic capabilities, and it estimated that only about 130 of the thousands of agentic AI vendors were real. In the same release it predicted that more than 40% of agentic AI projects will be cancelled by the end of 2027 because of rising costs, unclear business value or weak risk controls.
The vendor pages show the blur in practice. Tidio’s pricing FAQ describes its Lyro AI Agent as answering from your support content and contrasts it with Flows, which follow rules you set. On paid quotas Lyro can also run a limited number of Actions. Intercom lists the ability to take action on external systems among the features of its Fin AI Agent. Both are sold as agents, and both sit on a spectrum: the core job is answering support questions from your content, with optional actions layered on top. The lesson is to judge a product by what it is allowed to do, not by what it is called.
Three questions cut through the marketing. Ask them of any vendor before you pay:
- What can it change in my systems without a person approving? Reading a record is different from writing one. Refunds, bookings, emails and deletions are writes.
- What is the billable unit, and who decides when something is “resolved”? A conversation, a resolved conversation and an action are three different meters.
- What does it do when it is not sure, and can I read a log of every step? A good answer includes a handoff to a person and a step-by-step record.
The 7-Point Comparison: AI Agent vs Chatbot
The table adds the AI chatbot as a middle column, because that is the product most small businesses meet first. The cost row uses figures worked out in the next section.
| Point | Rules-based chatbot | AI chatbot | AI agent |
|---|---|---|---|
| 1. Autonomy | None. It follows the script you wrote | Chooses its wording, not its actions | Chooses its steps and which tools to use to reach a goal |
| 2. Memory | Only the answers collected in this session | The current conversation plus your uploaded content; most start fresh on the next visit | The conversation plus records it can look up; longer-term memory has to be built or switched on |
| 3. Integrations | Forms, email, sometimes a CRM field | Your knowledge base and a handoff to a human inbox | Read and write access to calendar, CRM, helpdesk, billing or email |
| 4. Cost | From $24.17 a month (Tidio Flows) | From $32.50 a month for 50 AI conversations (Tidio Lyro add-on) | $0.99 per outcome (Intercom Fin), or about $0.012 to $0.08 per conversation in model fees if you build it yourself |
| 5. Human in the loop | A handoff button | Handoff when the answer is unclear | Approval before anything that spends money, messages a customer, deletes or shares data |
| 6. Errors | Visible dead ends (“Sorry, I did not understand”) | A confident wrong answer you can read | A wrong action several steps deep, where mistakes can compound |
| 7. Security | Low exposure: collects form data and makes no decisions | A hostile visitor can try to manipulate answers, and your content passes to a model provider | Highest exposure: hostile text in an email, page or chat can try to steer an action, so limit access to what the task needs |
Two rows matter most for a small business: human in the loop and errors. A chatbot’s mistake is visible because you can read the wrong answer. An agent’s mistake can sit several steps deep in a record you never open. Anthropic’s guidance notes that autonomy brings higher costs and the potential for compounding errors, which is why every use case in the 10 real-world AI agent use cases guide is tagged with an autonomy level and an approval point. The AI agent governance guide turns the same idea into a short checklist, and the AI agents trust gap statistics show why buyers are right to be cautious.
Chatbot vs AI Agent Cost: What Each Option Charges at 1,000 Conversations a Month
Cost is where the two terms stop being interchangeable. The table prices one 1,000-conversation month five ways. All prices are in US dollars, exclude tax and use annual billing where a vendor offers it.
| Option | What you get | Published price | Approx. monthly cost at 1,000 conversations |
|---|---|---|---|
| Rules-based chatbot (Tidio Flows) | Scripted paths and buttons | From $24.17 a month for 2,000 visitors reached | About $24. Billed on visitors reached, not conversations |
| AI chatbot (Tidio Lyro) | Answers in plain language from your content | From $32.50 a month for 50 AI conversations; the first 50 are free, once | Not shown on the page; Tidio’s calculator covers up to 1,000 AI conversations a month |
| Managed support agent (Intercom Fin) | Resolves chats and can act on other systems | $0.99 per outcome; seats from $29 a month | About $326 to $722 at 30% to 70% resolution |
| DIY workflow agent | You build it with a no-code tool and a language model | Model tokens plus a small server | About $12 to $80, before platform plan, build time and review time |
| Voice agent (3-minute calls) | Answers the phone and can book | About $0.08 to $0.14 per minute | About $240 to $420 per 1,000 calls |
The Lyro row needs a caution. Third-party reviews put Lyro near $0.70 per conversation at the higher tiers on monthly billing, which would mean roughly $700 for 1,000 AI conversations, but that figure could not be confirmed on Tidio’s own page, where prices appear only in an interactive calculator. Use the calculator with your own volume before you decide.
How the Intercom Fin numbers are calculated
Fin is billed per outcome, so the monthly bill depends on how many of your 1,000 chats it resolves. Intercom’s page shows $0.99 per Fin outcome, $29 per seat a month on the Essential plan, and a version of Fin with no seats for teams that already have a helpdesk. Vendor-published resolution rates vary: an Intercom customer quote on its pricing page cites 50%, and Tidio says its customers average 75% with Lyro. Both are vendor-selected numbers, so the table below brackets them with a low case.
| Resolution rate | Outcomes out of 1,000 chats | Fin charge at $0.99 | With one Essential seat ($29) |
|---|---|---|---|
| 30% | 300 | $297.00 | $326.00 |
| 50% | 500 | $495.00 | $524.00 |
| 70% | 700 | $693.00 | $722.00 |
Assumption: every one of the 1,000 conversations starts with the AI, and one person on the Essential plan handles the rest.
What the headline price leaves out
- The billing unit. Tidio counts a Lyro conversation as any chat with at least one AI reply, however many replies follow. Fin charges only for outcomes. Third-party reviews of Intercom’s billing report that a chat can count as resolved when the customer simply stops replying, and that standalone use carries a monthly minimum of 50 outcomes. Confirm both in Intercom’s own definitions before signing. The outcome-based pricing guide lists four questions to ask.
- Retries and failed runs. The do-it-yourself figures include a 20% allowance for failed runs and an $8 server. Using a top-tier model for the same 1,000 conversations takes the bill to about $366, so the model you choose matters more than the platform.
- Review time. The person who approves drafts and audits logs is the largest unpriced line in every agent budget. The AI agent ROI guide shows how to count it.
- The automation platform. A workflow agent still runs on something. The Zapier vs Make vs n8n guide prices that layer at 10K, 50K and 500K tasks.
A better yardstick: cost per solved problem
A per-outcome price is only as honest as the definition of “outcome”. Pull a random sample of 50 chats your tool marked as resolved, read them, and count how many genuinely solved the customer’s problem. Then divide the bill by the problems actually solved.
Worked example, with illustrative numbers: at 50% resolution Fin bills $495 for 500 outcomes. If 40 of your 50 sampled chats were truly solved (80%), about 400 problems were solved and the real cost is $495 ÷ 400 = $1.24 per solved problem, not $0.99. The same check works for a per-conversation plan, a per-minute voice bill and a do-it-yourself agent, which makes it the fairest way to compare them.
What about voice? Vapi, Retell, ElevenLabs and Bland
Voice is where the chatbot-or-agent question is easiest to answer. A phone bot that only reads FAQ answers is a chatbot with a voice. One that checks your calendar and books an appointment is an agent, because it calls a tool and changes a record. The per-minute rates below come from the AI voice agent cost per minute guide, which read each vendor’s pricing page on 16 October 2026.
| Platform | What the headline rate covers | Published rate |
|---|---|---|
| Vapi | Hosting and orchestration only; model, transcriber and voice at provider cost | $0.05 per minute, or about $0.082 to $0.129 all-in on Vapi’s own calculator at 1,000 minutes |
| Retell AI | Every component priced separately | $0.07 to $0.31 per minute all-in |
| ElevenLabs Agents | ElevenLabs hosting; model tokens and telephony billed separately | $0.08 per minute |
| Bland AI | One bundled rate covering model, speech and telephony | $0.14, $0.12 or $0.11 per minute, depending on plan |
Most small-business setups in that guide landed between $0.08 and $0.14 a minute. For 1,000 three-minute calls, which is 3,000 minutes, that is $240 at $0.08, $330 at $0.11 and $420 at $0.14, before extras. Voice only makes sense when missed calls cost you real money. The AI answering service cost guide prices that decision at 100, 500 and 2,000 calls a month, and the self-hosted voice agent guide shows what the same job costs on a $5 server.
Use a Chatbot When… Use an Agent When…: A Decision Flowchart
Work through the four questions from the top. They are ordered so that you stop at the cheapest tool that does the job.
Figure: four yes-or-no questions lead to one of five answers. A text version follows for readers using a screen reader.
- Is the job only answering questions from information you already have? Yes: an answer-only chatbot. Use a rules-based flow if you have few questions, and an AI chatbot if they arrive as free text.
- Does it need to change something in another system? No: an AI chatbot that collects details and hands off to a person.
- Can you write down every step in advance? Yes: an automation workflow, with one AI step if needed and a chatbot as the front door.
- If it gets one wrong, is the damage small and easy to undo? Yes: an AI agent within limits. No: an AI agent that only drafts, with a person approving every action, or keep the task with a person.
Choose a chatbot when
- Most questions repeat, and the answers already live on your site or in a help document.
- The worst outcome of a wrong answer is a confused customer, not a wrong refund or a leaked record.
- You want something live in an afternoon on a free plan. Tidio’s free plan includes a one-off allowance of 50 AI conversations for testing.
- Your volume is low enough that a flat monthly plan costs less than building anything.
Choose an agent when
- Solving the problem means doing something in another system, such as booking, refunding, updating a record or sending a follow-up.
- The steps differ every time, so you cannot write them all down. If you can, use a workflow instead.
- You can put an approval step in front of anything that spends money, messages a customer, deletes data or shares data.
- You can audit a sample of its work every week and act on what you find.
Three illustrative cases
These are worked examples, not case studies, and the numbers are the ones used earlier in this guide.
- A plumber with 80 enquiries a month. A rules-based flow, or a free AI chatbot trial plus a lead form, covers it. An agent would add setup and oversight cost without changing the outcome.
- An online shop with 1,000 chats a month, many asking where an order is. An AI chatbot that can look up order status, a read-only action, handles the bulk. Refunds stay with a person until a sampled audit shows the tool deserves more.
- A salon or clinic that loses bookings after hours. A voice agent with calendar access fits, because booking is an action. Limit it to narrow, reversible actions such as booking into open slots, and have it hand anything else to a person.
Three Starter Stacks for Three Budgets
Each stack names real plans, adds up the monthly cost and says what should push you to the next one. Prices are US dollars on annual billing, and none of the totals include your own time.
Stack 1: Under $30 a month, “Answer and capture”
- Tool: Tidio Free ($0) or Starter ($24.17 a month). Free includes 50 billable conversations a month, Flows for 100 visitors and a one-off 50 AI conversations. Starter raises billable conversations to 100.
- Setup: five to ten scripted flows for opening hours, prices, booking link, quote request and contact capture, plus a handoff to your inbox.
- Total: $0 to $24.17 a month.
- Move up when: you keep answering the same free-text questions by hand, or visitors hit dead ends in the script.
Stack 2: About $70 to $160 a month, “AI chatbot plus one supervised workflow”
- Chat layer: Tidio Starter ($24.17) or Growth (from $49.17) plus the Lyro AI Agent add-on (from $32.50 a month for 50 AI conversations). That is $56.67 to $81.67.
- Workflow layer: one automation with a single AI step, such as classifying new leads and drafting a reply that a person approves. At 1,000 runs a month, model and server cost is about $12 to $80 in the cost model, depending on model tier. Add the automation tool’s own plan if you do not self-host.
- Total: about $70 to $160 a month ($56.67 + $12 at the low end, $81.67 + $80 at the high end).
- Move up when: a sampled audit shows the AI chatbot solving most chats and the remaining work is actions in other systems, not answers.
Stack 3: About $250 to $550 a month, “Resolve and act”
- Option A, chat at scale: Intercom Essential with one seat ($29) plus Fin at 500 outcomes ($495). That is $524 a month at 50% resolution of 1,000 chats.
- Option B, voice: a voice agent handling 1,000 three-minute calls, which is $240 to $420 depending on platform and voice, before phone numbers and extras.
- Option C, a bigger plan: Tidio Plus starts at $300 a month plus usage.
- Guardrails that come with this tier: narrow permissions, approval before refunds or customer messages, a step log, a monthly spend cap and a weekly sample audit. The governance guide has the checklist.
If your budget sits between stacks, the small business AI budget guide shows how a five-person team’s AI spend adds up across tools, not just chat.
Five Common Mistakes When Choosing Between a Chatbot and an AI Agent
- Buying an agent to answer FAQs. If the job is answering from your content, an AI chatbot does it with less risk and less oversight.
- Trusting the demo. Test with a sample of your own past conversations, including the awkward ones.
- Comparing prices in different units. Convert per-minute, per-outcome, and flat-plan prices into cost per solved problem.
- Granting write access on day one. Start read-only, add one action at a time and log every step.
- Skipping the human handoff. Whatever you buy, a customer should always be able to reach a person.
FAQ: AI Agent vs Chatbot for Small Business
What is the main difference between an AI agent and a chatbot?
A chatbot responds inside a conversation: someone asks, it answers. An AI agent is given a goal, decides which steps to take and uses tools to act in other systems, such as booking, updating a record or sending a message. The practical difference is not intelligence but permission to act.
Is ChatGPT a chatbot or an AI agent?
Plain chat is chatbot behaviour. When the same product is connected to tools and allowed to take multi-step actions for you, it is acting as an agent. The label depends on how it is configured, not on the brand name.
Which is cheaper for a small business, a chatbot or an AI agent?
A chatbot is cheaper to start: a rules-based flow begins at $24.17 a month. Per conversation, a do-it-yourself agent can cost pennies in model fees ($0.012 to $0.08 in the Beeznez Tech cost model), while managed products charge about $0.99 per outcome. The comparison turns on build time, review time and cost per solved problem, not on the sticker price.
Do I need an AI agent for customer support?
Not necessarily. If most chats are repeat questions answered from your own content, an AI chatbot with a human handoff covers it. You need an agent when resolving the chat means doing something in another system.
Can a chatbot become an AI agent?
Yes, by adding tools and permission: connect it to a calendar or order system and let the model decide when to use them. Start with read-only access and add one action at a time.
Are AI agents safe for small businesses?
They can be, with limits. The risk comes from broad permissions and untrusted input, not from the idea of an agent. Give each one the least access it needs, require approval for spending, customer messages and deletions, and keep a log. The AI agent governance guide covers the details.
What is agent washing?
It is Gartner’s term for vendors rebranding chatbots, assistants and robotic process automation as agents without substantial agentic capabilities. In June 2025 Gartner estimated that only about 130 of the thousands of agentic AI vendors were real.
AI Agent vs Chatbot for Small Business: Final Thoughts
Chatbot versus agent is not a ranking, it is a permissions decision. A chatbot answers, and its mistakes are visible. An agent acts, and its mistakes can hide. So start where mistakes are cheap and easy to see, which is usually an answering chatbot. Measure cost per solved problem on a real sample of your own conversations, then add one action at a time, behind an approval, only when the numbers justify it. The vendors will keep calling everything an agent. Your job is to ask what it is allowed to change, how it is billed and what happens when it is wrong.
Sources
- Anthropic, Building effective agents: the workflow versus agent distinction, the advice to use the simplest solution that works, and the note on higher costs and compounding errors. One respected definition, not an industry standard.
- Gartner, Over 40% of Agentic AI Projects Will Be Canceled by End of 2027: agent washing, the estimate of about 130 real vendors and the cancellation forecast. Published 25 June 2025.
- Tidio pricing: plan prices, the Lyro AI Agent add-on, Flows, the definition of a Lyro conversation and Tidio’s 75% average resolution claim (a vendor claim). Read 21 October 2026.
- Intercom pricing: seat prices, $0.99 per Fin outcome, Fin without seats, and the customer quote citing 50% resolution (vendor material). Read 21 October 2026.
- HelpDesk, Intercom pricing and Pluno, Intercom Fin AI pricing: third-party reports of assumed resolutions and the 50-outcome standalone minimum. Both publish competing products, so confirm the terms with Intercom.
- Macha, Tidio pricing explained: the monthly-billing price and the roughly $0.70 per Lyro conversation figure, a third-party number not confirmed on Tidio’s page.
- How Much Does It Cost to Run an AI Agent Every Month?: the $12, $80 and $366 monthly totals, the 20% retry allowance and the $8 server.
- AI Voice Agent Cost Per Minute: the Vapi, Retell, ElevenLabs and Bland rates, verified 16 October 2026.
- Beeznez Tech Editorial Policy: how research, AI assistance and verification work on this site.
Updated 21 October 2026. Prices change often, so confirm any figure on the vendor’s live page. Tidio and Intercom figures were read from their pricing pages on 21 October 2026; voice and model-cost figures come from earlier Beeznez Tech guides on the dates shown there. The totals are arithmetic on published prices, not measurements, and this guide involved no hands-on testing of these products. AI assistance was used for research and drafting, as described in the Editorial Policy.
Related reading: 10 Real-World AI Agent Use Cases, How to Build a No-Code AI Agent Pipeline, AI Answering Service Cost and AI Automation for Small Business.