Most articles with this title answer yes, because most are written by vendors. The truthful answer is that it depends on four things, and two of them have nothing to do with the software.
When it is genuinely worth it
You get website traffic you are not converting. Check sessions on your pricing and contact pages against enquiries received. If there is a large gap, there is something to recover.
A meaningful share of visits are out of hours. Segment analytics by hour. Evenings and weekends are the pool nobody is serving.
Your enquiries are repetitive before they get specific. If the first five minutes of every enquiry is the same five questions, that is exactly the work to hand over.
Your website already explains your business. The bot answers from your content. Good content in, good answers out.
When it is not
Very low traffic. If ten people visit a week, a chatbot will not fix your business. Spend the money on getting found.
Almost entirely repeat phone customers. If your work comes from people who already have your number, a website bot addresses a channel you are not using.
A thin website. If your site is three pages and a phone number, there is nothing to answer from. Write the content first; you will get value from that alone.
Nobody to maintain it. Prices change, services change. A bot quoting last year's rates is worse than no bot.
The failure mode nobody warns you about
The most common disappointment is not a bad model. It is a bot deployed on a site with almost nothing on it, answering vaguely, and the owner concluding "AI does not work for my business".
The AI worked fine. It had nothing to work with.
How to decide in an afternoon
Write down the five questions customers ask most. Check whether the answers are on your website. If they are not, that is your first job either way.
If they are, build a bot from your site for free, ask it those five questions, and judge it on the answers. That is the whole evaluation.
Related: what missed enquiries cost and how much should one cost.