Chatbot Solution Providers: Types, Costs & How to Choose (2026)
Searching for a chatbot vendor returns two very different kinds of result, and most buyers do not notice until they are on a sales call.
One kind sells you software you configure yourself. The other sells you a project: discovery, build, deployment, and a retainer. They are priced an order of magnitude apart, they fail in different ways, and the right answer depends on something most comparison articles never ask about.
TL;DR#
- Chatbot solution providers fall into four types: self-serve platforms, managed platforms, development agencies, and systems integrators.
- The deciding question is not features. It is whether your content is ready and who is going to maintain it.
- A platform costs $40 to $900 a month. An agency build starts around $15,000 and rarely ends there.
- The failure mode of agencies is a bot nobody updates after handover. The failure mode of platforms is a bot nobody configured properly.
- Ask every provider how the bot decides which passage to answer from, and whether it cites the source. Most cannot answer the first question.
The Four Types of Chatbot Solution Provider#
1. Self-serve platforms#
You sign up, point the tool at your content, and deploy a widget. No contract, no implementation call. Pricing runs from free tiers to a few hundred dollars a month.
Best for: teams whose knowledge already exists in documentation, a help center, or a website. Weak for: deep integrations into internal systems, or anyone without someone to own the content.
Examples include Denser, Chatbase, Botsonic, and SiteGPT.
2. Managed platforms#
The same software, plus a human who configures it, tunes answers, and reports on performance. Usually sold as an enterprise tier rather than a separate product.
Best for: organizations that want the software model but lack the internal owner. Weak for: small budgets, since the managed layer typically doubles the price.
3. Development agencies#
They build a custom bot, often on top of an existing platform or an LLM API. Deliverables are a working system and, if you negotiate it, the code.
Best for: unusual workflows, deep integration with proprietary systems, or industries with requirements no platform serves. Weak for: ordinary question answering, where you pay a project fee for something a platform does out of the box.
Expect $15,000 to $150,000 depending on scope, plus a maintenance retainer.
4. Systems integrators#
Large consultancies deploying enterprise stacks such as Salesforce, Zendesk, or Microsoft alongside the rest of a transformation program. Six figures and up, measured in quarters.
Best for: enterprises already committed to one of those ecosystems. Weak for: anyone who wants something running this month.
What Each Type Actually Costs#
| Provider type | Setup | Ongoing | Time to live |
|---|---|---|---|
| Self-serve platform | None | $40 to $900/mo | Hours |
| Managed platform | $2,000 to $10,000 | $500 to $3,000/mo | 2 to 4 weeks |
| Development agency | $15,000 to $150,000 | Retainer, often 15 to 20% | 2 to 6 months |
| Systems integrator | $100,000+ | Enterprise contract | 3 to 12 months |
The trap in this table is the third row. Teams often buy an agency build for a problem the first row solves, because the sales process is more reassuring. A custom build is justified by integration complexity, not by wanting the project to feel serious.
The Question That Separates Good Providers From Bad Ones#
Ask how the bot decides which passage to answer from.
Weak providers describe the model. "We use GPT-5" or "we use Claude" tells you nothing, because every provider has access to the same models. The model is not the differentiator and has not been for years.
Strong providers describe retrieval: how your content is chunked, how the right passage is found, and how the system behaves when nothing relevant exists. That is what determines whether an answer is correct. A capable model that retrieves the wrong paragraph produces a confident, wrong, well-written answer, which is worse than no answer at all.
Follow it with a second question: does every answer cite the passage it came from? If it does, you can verify a reply in one click and hand the bot to customers with some confidence. If it does not, every answer is a claim you are trusting.
A Short Evaluation Checklist#
- How much content can one bot hold? Ask in your units, not theirs. Some vendors cap at 20 MB, others at tens of thousands of pages.
- Does accuracy hold as content grows? Ask what happens at ten times your current volume. Retrieval that works on 50 pages often degrades badly at 5,000.
- What does it do when it does not know? A grounded refusal is correct behavior. A confident guess is a liability.
- How do you find out what it failed to answer? Without this, gaps surface as customer complaints.
- What is the real monthly bill? Credits that burn faster on better models, per-resolution fees, branding removal, and API access are the four costs that surprise people.
- Who owns the content after handover? For agency builds especially, establish who updates it in month seven.
Platform or Agency? A Simple Test#
Answer these three:
- Is your knowledge already written down somewhere, even messily?
- Does the bot mostly need to answer questions rather than perform transactions in other systems?
- Do you have anyone who could own it for an hour a week?
Three yes answers means a self-serve platform will serve you, and an agency would be selling you a project you do not need. A no on the third means look at managed. A no on the second, because the bot must write to your ERP or run multi-step transactions, is the real case for custom development.
How Denser Fits#
Denser is a self-serve platform in the first category. You point it at your website, documentation, or files, and it builds a chatbot that answers from that content and cites the exact passage behind every reply. Retrieval runs on Denser Retriever, which is open source, so the mechanism above is one you can read rather than take on trust.
It also reports which questions it could not answer, grouped by topic, which is the checklist item most providers skip.
Plans run from free to $899 a month, published in full on the pricing page, and every paid plan starts with a 7-day trial. If you are comparing named vendors rather than provider types, the comparison pages put Denser head to head with eleven of them.
For a product-by-product ranking rather than a guide to vendor types, see the best AI chatbot solutions. For large deployments specifically, see enterprise chatbot solutions.
FAQs About Chatbot Solution Providers#
What is a chatbot solution provider?#
A company that supplies the software, the build, or both for a conversational AI assistant. The term covers four quite different businesses: self-serve software platforms, managed services on top of those platforms, development agencies that build custom bots, and systems integrators deploying enterprise suites.
How much do chatbot companies charge?#
Self-serve platforms run roughly $40 to $900 a month with no setup fee. Managed platforms add $2,000 to $10,000 in setup plus a higher monthly. Development agencies charge $15,000 to $150,000 for a build and usually a maintenance retainer. Systems integrators start in six figures.
Do I need a chatbot development company?#
Usually not. Custom development is justified by integration complexity, such as a bot that must transact inside proprietary systems, rather than by question answering. If your knowledge already exists as documents or web pages, a platform will deploy the same capability in an afternoon.
What should I ask a chatbot vendor?#
Ask how retrieval selects the passage an answer comes from, whether every answer cites its source, how much content one bot holds, what happens when the bot does not know, and how you find out which questions it failed. The first two expose more about answer quality than any feature list.
Are AI chatbot services different from chatbot software?#
Generally yes. "Services" usually signals a provider selling implementation work, whether an agency or a managed tier, while "software" signals a platform you configure yourself. The distinction matters most for cost and for who owns maintenance once the bot is live.