A customer types "I got the wrong color, I want a refund." A chatbot sends back a link to your returns policy and a form to fill out. An AI agent reads the order, checks the return window, processes the refund, and sends a confirmation — no human involved. Same question, two completely different outcomes, and most small business owners have no idea which one they just bought.
That confusion is expensive. Vendors have spent the last two years slapping the word "agent" on products that haven't changed much since 2022. Analyst estimates suggest that of the thousands of companies marketing "agentic AI," only a small fraction are the real thing. Everyone else is a chatbot with a rebrand.
What Actually Separates a Chatbot From an Agent
The distinction isn't marketing fluff, even though marketing has made it hard to see.
A chatbot — even a good, AI-powered one — matches a question to an answer. It retrieves information from a knowledge base or FAQ and responds. It's read-only.
An AI agent works differently: it reasons about a goal, breaks it into steps, calls outside tools or APIs, and takes action across systems without a person clicking "approve" at every stage.
Industry framing puts it simply: a chatbot deflects, an agent resolves. A chatbot's job ends when it hands the customer information. An agent's job ends when the actual problem is fixed — the refund is processed, the appointment is booked, the order is updated.
The technical difference comes down to a reasoning loop. A basic chatbot processes one message, gives one answer, done. An agent runs a loop — observe the request, reason about what to do, take an action (like calling an API), then evaluate whether the goal was met — and it repeats that loop until the task is actually finished. That's what lets an agent chain steps together instead of just answering once and stopping.
In practice, the line blurs. Add a single tool — say, an order lookup — to a chatbot, and it starts drifting toward agent territory. Most real products today live somewhere in between, which is exactly why the label on a vendor's homepage tells you so little (read the best AI chatbot platforms for small businesses).
The 2026 Agent-Washing Problem
"Agent washing" means taking an existing chatbot, RPA tool, or virtual assistant and rebranding it as "agentic AI" without adding any real autonomous capability. A call recording feature becomes a "transcription agent." A CRM integration becomes an "activity mapping agent." Nothing changed except the label.
The scale of it is the part worth sitting with. Gartner's estimates put the number of genuinely agentic vendors at roughly 130 out of thousands claiming the label. That's not a rounding error; that's a market where the overwhelming majority of "AI agents" for sale are chatbots with better copywriting.
Analysts also predict more than 40% of agentic AI projects will be cancelled by the end of 2027, mostly because of escalating costs and unclear ROI once businesses realize what they actually bought.
None of this means agentic AI is fake or not worth pursuing — it is expected to be inside a third of enterprise software by 2028. It means the burden is on the buyer to check, not assume, that a product does what the pitch deck says.
What Each One Is Actually Good At
Neither tool is universally "better." They solve different problems, and picking the wrong one either overspends on complexity or underdelivers on what your customers actually want done.
| Feature | Chatbot | AI Agent |
|---|---|---|
| Best for | FAQs, pricing questions, hours, simple lookups | Refunds, bookings, order changes, multi-step requests |
| How it works | Matches question to pre-written or retrieved answer | Reasons, plans, calls tools/APIs, acts across systems |
| Typical resolution rate | 30–40% for straightforward FAQ bots | 50%+ when built with connected systems |
| Setup complexity | Low — point it at your FAQ page, live in minutes | Higher — needs integrations into booking/CRM/order systems |
| Realistic cost | $0–$150/month | $150–$800+/month, often more with per-resolution pricing |
| Risk if miscast | Frustrates customers with a low-risk, low-value answer | Overkill and wasted spend on questions that never needed reasoning |
A chatbot is the right call when volume is low and the questions are genuinely informational — shipping times, store hours, return policy text. An agent earns its cost when a meaningful share of your inbound involves an actual action inside your systems, not just an answer (see how to add an AI chatbot to your website).
What Reasonable Resolution Rates Actually Look Like
This is where the hype gets loudest. Vendor pages love to cite headlines that agentic AI will autonomously resolve 80% of common customer service issues — but that's a long-term target, not a current reality, and it's already been walked back with caveats about unified, cross-channel resolution that almost no small business setup has.
What's actually happening on the ground looks more modest. Around 30% of service interactions were AI-handled in 2025, with projections climbing toward roughly 50% by 2027. Entry-level tools like Tidio's Lyro deliver 30–40% deflection for a typical small business FAQ setup. Custom-built bots trained specifically on a business's own content can reach 60–80% resolution, but that range assumes real setup work, not a five-minute plug-in.
Intercom's Fin is marketed as resolving "50%+ of inbound tickets" — plausible, but that's measured against businesses with an already-solid knowledge base, and it's billed per resolution ($0.99 each) on top of seat fees (read AI customer support vs traditional help desk). Run the math at 2,000 monthly resolutions and you're at nearly $2,000/month in AI fees alone, before seats. That's a legitimate tool — it's just not a $30/month tool once volume shows up.
Real Tool Pricing in 2026
| Tool | Type | Starting Price | Notes |
|---|---|---|---|
| Tidio + Lyro | Chatbot | ~$29/month base + ~$39/month AI | Good entry point; jumps sharply at higher volume |
| Chatbase | Chatbot (RAG-based) | $0–$40/month | Credit-based; costs vary by chosen model |
| Crisp | Chatbot | ~€95/month flat | Unlimited conversations, predictable billing |
| Intercom Fin | AI Agent | $29–$132/seat/month + $0.99/res | Genuinely capable, but costs scale fast with volume |
| Voiceflow | Agent builder | $60/month | Better for teams building custom multi-step flows |
| n8n | Workflow automation | Free self-hosted / from $20/mo | Pairs a chatbot front-end with real backend actions |
The honest takeaway from this table: nobody serious is paying $29/month for a true agent at real volume. The advertised entry price is almost always a teaser for a single-digit conversation count. If a vendor won't show you their pricing at 1,000+ monthly conversations, assume it's worse than the homepage implies (read AI receptionist cost breakdown).
This is also where a lot of DIY setups hit their ceiling — someone wires together a basic chatbot and a workflow tool, it works fine at 50 conversations a month, and then it either breaks or gets expensive fast the moment the business actually grows.
At Brandywebs, we build custom sites starting at $999 with automation planned in from the start rather than bolted on after the fact.
How to Actually Decide Which One You Need
- Pull your last 100 customer messages or emails and sort them into two piles: "answered by pointing at information" and "required actually doing something in a system" (refund, reschedule, cancellation, order edit).
- Count the second pile. If it's under 20%, a chatbot covers you. If it's 30%+ and growing, you're paying a chatbot to fail at agent-shaped problems.
- Check what systems those actions touch — your booking calendar, your e-commerce backend, your CRM. An agent is only as useful as the number of those systems it can actually connect to.
- Ask any vendor demo for a production log, not a demo-environment metric: the last ten real decisions the tool made, what triggered them, and what happened next.
- Start small and measure. Most platforms offer free or under-$50 tiers specifically so you can test resolution quality before committing budget.
- Re-evaluate every 90 days. Resolution rates, pricing tiers, and what counts as "agentic" are all moving quickly in 2026.
Decision Framework
| If your business... | You probably need... |
|---|---|
| Gets mostly repetitive, informational questions | A basic AI chatbot (Tidio, Chatbase, Crisp) |
| Has growing volume but still FAQ-shaped questions | A mid-tier chatbot with better retrieval (watch credit limits) |
| Handles refunds, bookings, or account changes regularly | A genuine agent with system integrations (Intercom Fin, custom build) |
| Isn't sure which pile most tickets fall into | Start with a chatbot, log unresolved issues, revisit in 60–90 days |
Bottom Line
Most "AI agent" products on the market in 2026 are chatbots wearing a new label — analyst estimates put genuinely agentic vendors at around 130 out of thousands making the claim. That doesn't mean the technology is fake; it means the label alone tells you almost nothing, and the only way to know what you're buying is to check what it actually does with a real customer request.
The right choice isn't about which tool sounds more advanced. It's about what share of your customer messages require an actual answer versus an actual action. Most small businesses are still mostly in "answer" territory, which means a $29–$150/month chatbot is plenty.