A chatbot is software that talks with customers or staff in natural language, on a website, in a messaging app or by voice, to answer questions and complete simple tasks. In 2026 the real decision for most businesses is no longer whether to use one. It is whether to configure a chatbot platform or commission a custom AI build. The answer depends on how much of your own data, systems and security rules the chatbot has to work with.
The evidence for caution is clear. A Gartner survey of 3,566 customers published in July 2026 found that customers are about three times more likely to use third-party generative AI tools than a company's own chatbot to resolve service issues. Use of company-provided chatbots has stayed statistically flat since 2022. This guide explains why, what a better chatbot looks like, and how to choose between a platform and a custom build.
What chatbot technology looks like in 2026
Three kinds of tools are now sold as "chatbots". They behave very differently:
| Type | How it works | Best for | Main risk |
|---|---|---|---|
| Rule-based chatbot | Follows scripted flows and button menus | Simple, predictable questions such as opening hours or order status | Dead ends when a question falls outside the script |
| AI chatbot (LLM-based) | Uses a large language model to understand free text and write answers | Support and sales questions with many phrasings | Confident wrong answers if it is not tied to your approved content |
| AI agent | Understands the request and also takes actions in your systems, such as booking, refunding or updating a record | Resolving whole tasks, not just answering | Needs strict permissions, logging and human review |
The same Gartner release shows where the opportunity is. Among customers who use generative AI, 58% say they have used it to complete a task on their behalf (74% in B2B settings). Gartner's advice to service leaders is to move away from standalone chatbots and toward "AI-enabled service journeys" that resolve issues across digital and voice channels. In practice that means chatbots that connect to your systems and finish the job, not ones that only answer questions.
Benefits of chatbots for business
When a chatbot is well scoped and connected to accurate content, the benefits are practical:
Faster service at any hour
A chatbot answers straight away on your website, Instagram and other messaging channels, including outside office hours, and can handle many conversations at the same time. That shortens waiting for routine questions and lets people reach you on the channel they already use.
More conversions and better lead capture
Visitors who get a quick answer to a pricing, availability or fit question are more likely to stay. A chatbot can also ask a few qualifying questions and pass a warm lead to your sales team with the conversation attached.
Less routine work for frontline staff
Automating repeat questions frees your team for complex cases, complaints and sales conversations that need a person.
Useful business intelligence
Chat logs show what customers ask, where they get stuck and which answers are missing from your website. That feeds content, product and service decisions.
Set expectations on return
The same Gartner research found that service and support leaders put a median 12% of their 2025 budget into AI, the highest share of ten business functions, yet only 24% of them could show positive financial returns across their AI use cases. Pick one use case with a baseline you can measure, rather than a broad launch.
Chatbot platform or custom development?
Once you know the use case, you reach a crossroads: configure a chatbot creation platform or commission a custom build from a development partner. Neither is better in every case.
| Factor | Chatbot creation platform | Custom development |
|---|---|---|
| Speed to launch | Fast: templates and visual builders | Slower: scoping, build and testing |
| Upfront cost | Lower, usually a subscription | Higher, project-based |
| Fit to your process | Limited to what the platform supports | Built around your workflows |
| Integrations | Standard connectors only | Any system with an API, including older ones |
| Your data and security | Governed by the vendor's setup | You set the controls, hosting and logging |
| Ongoing work | Your team maintains content and fixes | Can be covered by the partner |
Using a chatbot creation platform
A chatbot platform lets people without coding skills build a bot using a visual editor, pre-built templates, basic analytics and a set of standard integrations. For a simple, well-defined use case it is often the more cost-effective solution to start with, and many platforms now include AI answers on top of scripted flows. Pricing usually depends on the number of conversations, users or bots.
The trade-offs are limits and ownership. If you need an integration the platform doesn't offer, such as a particular store, CRM or internal system, you either change platform or go without. You also need someone on your team to keep content accurate, review conversations and fix problems promptly, because a wrong answer reaches customers immediately.
Custom chatbot development
A custom build suits businesses that need the chatbot to work with their own systems and rules. It can be grounded in your approved documents so that answers come from your content rather than the model's guesses (see our explainer on retrieval-augmented generation), connected to your CRM, ticketing or booking tools, and designed with the security, logging and approval steps your industry requires.
With a partner such as VisioneerIT, you also hand over setup, testing and future updates, so you don't need technical staff to run it. The cost is higher and launch takes longer, which makes a custom build the better choice when the use case is high value, regulated or tied closely to your systems.
How to choose
- Choose a platform if the use case is simple, the data is not sensitive, and you have a person to own the bot.
- Choose custom if the bot must act in your systems, handle personal or regulated data, or follow strict approval rules.
- Start small either way: one use case, one channel, a measurable baseline.
Risks to manage before you launch
- Wrong answers are your responsibility. In Moffatt v. Air Canada (February 2024), a British Columbia tribunal held the airline liable after its chatbot gave a customer wrong information about bereavement fares, and awarded about $650 CAD in damages. The tribunal rejected the argument that the chatbot was a separate entity and noted that the correct policy appeared elsewhere on the site. Ground the bot in approved content and test it on real questions.
- Tell people they are talking to AI. Article 50(1) of the EU AI Act requires chatbot systems to inform users that they are interacting with AI, from 2 August 2026, unless it is obvious to a reasonably well-informed person. If you serve EU customers, check how this applies to you with your legal adviser, and read our guide to AI compliance best practices.
- Always offer a human. Gartner reports that customers expect the option to reach a human agent when a company uses AI in customer service. Build a clear handoff that passes the conversation along.
- Protect the data. Decide what the bot may see and store, restrict its permissions, and log what it does. The NIST AI Risk Management Framework and a proactive cybersecurity strategy are good starting points.
How to launch a chatbot: 5 steps
| Step | What to do | Output |
|---|---|---|
| 1. Pick one use case | Choose a high-volume question or task, such as order status, bookings or lead qualification | A problem statement with a baseline (volume, handling time, cost) |
| 2. Prepare the content | Gather approved answers, policies and FAQs; remove out-of-date pages | A reviewed knowledge source |
| 3. Choose the build | Use the comparison above to pick a platform or custom development | A decision with named owner and budget |
| 4. Test with real questions | Run past customer questions, edge cases and attempts to break the bot; set up the human handoff and AI disclosure | A test report and go-live checklist |
| 5. Launch and measure | Track resolution, handoff rate, customer satisfaction and cost per resolved issue; review chat logs weekly | A monthly improvement list |
For a wider view of rolling out AI across the business, see our complete guide to AI implementation.
Frequently asked questions
What is chatbot technology?
Chatbot technology is software that holds a conversation with people in natural language, through text or voice, to answer questions and complete simple tasks. Modern chatbots use large language models to understand free-form questions.
What is the difference between a chatbot and an AI agent?
A chatbot mainly answers questions. An AI agent also takes actions in your systems, such as updating a booking or opening a ticket, so it needs tighter permissions, logging and human review.
Should I use a chatbot platform or a custom chatbot?
Use a platform for simple, low-risk use cases that your team can maintain. Choose custom development when the chatbot must integrate with your systems, handle sensitive data or follow strict rules.
Do I have to tell users they are talking to a chatbot?
Under Article 50(1) of the EU AI Act, chatbot providers must make sure users are informed they are interacting with AI, from 2 August 2026, unless it is obvious. Telling users is also good practice everywhere, and you should confirm your obligations with your legal adviser.
How do I know if my chatbot is working?
Measure the share of conversations resolved without a person, the handoff rate, customer satisfaction and the cost per resolved issue against the baseline you set before launch. Review the chat logs regularly for wrong or missing answers.
Key takeaways
- Gartner's July 2026 survey found customers are about three times more likely to use third-party generative AI than a company's own chatbot, so a generic bot is not enough.
- Chatbots work best when they are tied to your approved content and can complete tasks in your systems.
- Choose a platform for simple, low-risk use cases and custom development for integrated, sensitive or regulated ones.
- You are responsible for what your chatbot says, so test it, log it and always offer a human.
- Start with one use case and a measurable baseline.
