How to Choose a Chatbot Platform: Questions to Ask Vendors

To choose a chatbot platform, ask every vendor the same questions: where do the answers come from, what can the bot do on your site (and what can’t it), how is data handled, and what will you really pay? Then test it on your own content before you sign anything. Got three demos booked? Keep in mind that a polished demo is the vendor’s best script. It tells you very little about how the product copes with the questions your customers actually send. Below is a vendor-neutral checklist you can take into each call, so you can line the answers up side by side afterwards.

What should you decide before the first demo?

Pick one use case and collect a list of real customer questions first. Otherwise every demo will look good. Go through your inbox, support tickets and phone notes, and keep the awkward ones: vague wording, typos, two questions crammed into one message. That’s your test material. And it’s the one thing a vendor can’t prepare for.

Then settle the channel. Website chat, voice and messaging apps are different products, so decide between chat and voice before a salesperson decides for you. Start narrow. Not sure where? Pick the first use case that eats the most repetitive time today. Last step: write down your must-haves and nice-to-haves. Why bother? Because a shiny feature on screen has a way of quietly reshuffling your priorities mid-call.

Chatbot vendor questions about answers and knowledge sources

Start with the big two: where does the bot get its answers, and what does it do when it doesn’t know? A bot that answers from your own pages behaves very differently from one that leans on a general language model. The second kind can sound completely sure of itself while getting your prices or policies wrong.

• Source of answers: your website pages and documents, a general model, or a mix?
• Unknown questions: does it admit it has no answer, or does it guess?
• Updates: how do you change content, and how fast does the bot pick it up?
• Conversation logs: can you read through chats to catch wrong or missing answers?
• Languages: which ones are supported, and does the bot switch to the visitor’s language on its own?

Integration, handoff and actions: what the bot can actually do on your site

Ask straight out how the bot gets installed, and whether it only answers questions or also pushes conversations and data into other systems. Some tools go in as a single script pasted into your pages. Others rely on native integrations with your CMS, help desk or CRM. Those take longer to set up, but they can actually move data around.

Then check three things. First, can a conversation be handed to a live agent, and where does it land: email, a help desk queue, a chat inbox? Second, can the bot take actions like creating tickets, booking appointments or updating records, or is it answer-only? Third, does it live only as a website widget, or also in messaging apps and voice?

An answer-only widget installed with one script is a perfectly reasonable pick if what you mostly need is fast, accurate answers from your own content. Nothing wrong with that. If handoff or CRM sync sits on your must-have list, ask to see exactly that in a live demo. Not as a roadmap promise. Botino belongs in the answer-only group, as its list of the platform’s documented capabilities shows, so judge it by that yardstick.

Data protection: the chatbot data processing agreement

If a vendor processes personal data from your chats on your behalf, it’s a processor, and the GDPR says you need a contract with it. Article 28 of the text of the GDPR requires processing by a processor to be governed by such a contract. So raise it early. Finding the gap after purchase is a lot less fun.

• Will you sign a data processing agreement, and can we see it before the trial?
• Where is chat data stored, and for how long?
• Which subprocessors are involved?
• Can chat logs be deleted or exported on request?

This is not legal advice. Get your own lawyer or data protection contact to review the contract before you sign.

Comparing chatbot pricing models without surprises

Work out what one month of your real traffic would cost on each plan. Ignore the headline price. Vendors bill in different units (per conversation, per message, per token, per seat, per bot), so the offer that looks cheapest can turn into the most expensive one once your volume goes through it.

Ask about limits as well: how many bots, how many documents the knowledge base will take, token or message caps. And the question people forget: what happens when you hit a cap? Does the bot stop? Slow down? Start running up an overage bill? Pin down what’s included and what costs extra, like voice, additional languages, setup work and support. For Botino, you can compare the monthly plans and their token limits on one page.

How to run a chatbot trial before buying

A trial on your own content, with your own question list, is the only reliable way to tell a usable product from a polished demo. Run it the same way on every platform on your shortlist:

1. Load your real pages and documents, not a sample set the vendor hands you.
2. Ask the same question list on each platform.
3. Score every answer as correct, partly correct, wrong or refused.
4. Check how the bot handles questions it can’t answer.
5. Install it on a staging page and watch how it behaves in context.

Bring in whoever answers customer questions today. They’ll catch the subtle errors a manager tends to miss. Also keep track of the effort: how long setup took, and who on your team ended up doing it.

Chatbot platform checklist: questions to bring to every call

Use one shared list for every vendor, so you can compare answers line by line. Boiled down from the sections above:

• Knowledge sources and how content updates reach the bot
• Behaviour with questions outside the knowledge base
• Installation method: script or native integrations
• Human handoff and where conversations land
• Actions in other systems, or answer-only
• Channels: widget, messaging apps, voice
• Data processing agreement, data location and retention
• Pricing unit, limits and paid extras
• Trial terms and access to your own content during the trial
• Exit terms and data export

Get vendors to answer in writing. That way, anything promised in a demo can be checked later against the contract and your trial results.

So, to choose a chatbot platform with confidence: trust the answers you can test over the ones on slides, and let your own question list make the call. If you’re planning bigger changes to how your team works, our posts on digital transformation cover the next steps.

FAQ