Custom AI agent or off-the-shelf chatbot?

What’s the difference between an AI agent and a chatbot?

A chatbot answers questions. An AI agent works towards a goal, such as a qualified lead or a booked visit, and takes actions in your systems: it updates your CRM, books calls, routes leads and hands over to your team. Both can run on similar AI models. What differs is what each one connects to and may change.

Many website chat widgets are chatbots. They match a visitor’s question to an answer from your FAQs or help pages, and newer ones use an AI model to word the reply. Once the question is answered, the job is done. Some also collect a name and phone number and email them to you.

An agent has a job to finish. A lead qualification agent asks your qualifying questions, writes each answer into its own CRM field, decides by your rules whether the lead is hot, books a call in a salesperson’s calendar and alerts them. When a lead asks something your approved information doesn’t cover, the agent passes the conversation to your team with the full context instead of guessing.

The line between the two keeps moving. Many chatbot platforms now add AI replies and a few actions, such as creating a contact in a CRM. So when you compare options, set the label aside and ask three things: what the tool can read, what it can change, and who sets the rules.

How do they compare side by side?

An off-the-shelf chatbot is quicker and cheaper to start, and your team can usually edit it without a developer. A custom agent takes longer to build, but it follows your process and writes to your own systems. The table compares the two on the points that usually decide it.

Neither column wins on every row. The chatbot wins on speed, entry cost and self-service edits. The custom agent wins on fit, integrations and control. Which rows matter most depends on how your leads arrive and what happens to them next.

When is an off-the-shelf chatbot the better choice?

Choose an off-the-shelf chatbot when the job is answering common questions from documents you already have, capturing simple leads at low volume, or going live within days. If the bot doesn’t need to write to your CRM and your team wants to edit the flows itself, a subscription costs less and starts sooner.

If most chats ask about opening hours, delivery times, returns or fees, the answers already sit in your FAQs and help pages. A chatbot that reads those documents and replies in plain language covers the job, and there is nothing else to connect it to.

Simple lead capture works the same way. If a few enquiries arrive each week and a salesperson calls every one, a bot that takes a name, a number and one or two details, then emails them to you, is enough. At that volume, seat or usage fees stay small.

Some teams want to change the wording on a Friday afternoon without asking anyone. Chatbot platforms are built for that, with a visual editor, templates and a preview. You can change a custom agent too, but the change usually goes through whoever built it.

Timing counts as well. If you need something live for a launch next week, a chatbot can be running in days, while a custom agent needs weeks of build and testing.

Before you sign, price the plan at the volume you expect in a year, and read how the price changes as you grow.

When does a custom agent pay back?

A custom agent pays back when qualifying a lead depends on your CRM, your rules and your data, when it has to act across several systems, or when volume makes per-conversation platform fees expensive. It also suits businesses that want to own the set-up and keep their conversations in their own accounts.

Qualification that your sales team trusts rarely fits a generic form. A real estate developer may route leads by project and location, check whether the lead is already in the CRM with another salesperson, and offer a site visit only if the budget fits the units still available. That takes an agent that reads and writes your data while it asks the questions.

The actions cross systems too. A booked visit should land in the salesperson’s calendar, the lead’s CRM record and the follow-up sequence at the same moment. We connect agents to the major CRMs, WhatsApp Business, email, calendars and spreadsheets, and build custom integrations wherever an API exists.

Volume changes the sums. Fees charged per seat or per resolved chat are easy to accept while chats are few, and harder once they grow every month. Price a custom agent the same way, because its AI model and messaging usage grows with volume too.

Then there is control. Conversations sit in your own accounts, so the history stays with you. We use business-grade AI providers that don’t train on your data, and access is limited to the people who need it.

WhatsApp-first sales make the case stronger. An agent on WhatsApp has to work inside Meta’s rules, from the official Business Platform to approved templates and the 24-hour reply window, and still hand hot leads to a person straight away. Our guide on how to qualify WhatsApp leads with an AI agent covers the set-up, the questions and the handover.

How should you decide?

Start with the cheapest option that meets your must-haves. For FAQs and simple lead capture, that is usually a SaaS chatbot. Go custom where your process is specific and the volume is real: where qualification needs your CRM and rules, and where platform fees grow faster than the value each lead brings.

Write your must-haves down before you look at any tool, then put each option through the same checks before you sign:

• Data ownership. Where are the conversations stored, and who can see them?
• Export. Can you take the full conversation history and lead data with you, in a format you can use?
• CRM writes. Can it write each answer into its own CRM field and update existing records, or only create new contacts?
• Handover. Can your team take over any conversation with the full history, and does the bot pass on questions it can’t answer?
• Pricing at your volume. What will you pay at the volume you expect in a year, including any per-seat, per-conversation or per-resolution fees?

If a SaaS chatbot passes every check, use it. If it fails on CRM writes, handover or pricing at your volume, cost a custom agent before you commit.

You can also run both: a chatbot for FAQs on your website, and a custom agent for the leads your sales team cares about.

If you can’t tell which side you’re on, our two-week automation audit reviews your lead flow and response times, lists the manual tasks that take up your team’s week, checks your tools and data, and ends in a prioritised plan. To see what each engagement includes, start with our AI automation services.