💼 Cost & business

Uptime and reliability: questions to ask a chatbot vendor before you commit

Updated 2026-09-19 · 5 min read · By the Convo360 team

In short

Before committing to a chatbot vendor, ask what happens when your own AI or telephony provider fails, how often your data is backed up, whether failed broadcast sends retry automatically, whether you can see system status, and whether you keep ownership of your WhatsApp number and API keys. A vendor that answers all five clearly is far less risky than one that only talks about their own uptime percentage.

Key takeaways
  • Most real 'downtime' a business experiences is a failing AI or telephony provider key, not the platform itself — ask how that's detected and alerted.
  • Backup frequency and scope matter more than an uptime percentage, since a platform can be 'up' while your data is unrecoverable after a mistake.
  • A broadcast or campaign send that fails partway should retry automatically, not silently drop messages.
  • You should be able to see what's happening on your account without opening a support ticket first.
  • Owning your own WhatsApp number, WABA and API keys is the single biggest protection against a vendor outage becoming your emergency.

Why 'uptime percentage' alone doesn't tell you much

Vendors often lead with a headline uptime number, but for a WhatsApp or voice bot, the platform being technically 'up' is not the same as your bot actually working. The more common failure in practice is your own connected provider: your AI key running out of credit, a telephony account's balance hitting zero, or a WhatsApp access token expiring, silently breaking the bot while the platform itself reports no incident at all. The right questions focus on how a vendor handles these everyday failure points, not just their own server uptime.

This is worth pressing on specifically during a sales conversation, because 'uptime' is an easy number to quote and a hard one for a customer to verify independently. Asking for the failure scenarios above, and how each is actually surfaced to you, tells you far more about day-to-day reliability than a percentage on a website.

What happens when your AI or telephony provider fails

Since a BYOK platform depends on your own connected keys and accounts, ask specifically: does the platform detect when your OpenAI, Claude, Gemini, or telephony provider starts returning errors, and does it tell you clearly, or does it fail silently until a customer complains? This is the single most common real-world 'my bot stopped working' cause, and it has nothing to do with the platform's own servers.

Ask thisRed flag answerGood answer
How would I know if my AI key expired?'You'd notice when conversations stop coming through.''An alert appears on your dashboard flagging the specific provider and error.'
What if my telephony balance runs out mid-campaign?'Calls would just start failing.''You get a clear notification, and can top up before it affects more calls.'

Backups: frequency, scope and how restore actually works

Ask three specific things, not just 'do you back up data': how often (nightly is a reasonable minimum), what's included (conversation history, contacts, flows, uploaded media, not just the database), and whether you've actually seen a restore work, versus it being an assumption nobody has tested. A vendor who can answer specifically ('nightly database backup, plus a separate nightly media/uploads backup') is giving you a real answer; 'we have backups' with no detail is not.

It's also worth asking how far back backups are retained. A nightly backup that's overwritten the next night protects against a server failure but not against a mistake discovered a week later. Retention of at least a few daysgives you a genuine safety net rather than a single point-in-time snapshot.

What happens to a failed broadcast send

A WhatsApp broadcast or campaign to thousands of contacts will occasionally hit individual failures: a number that's blocked the business, a template rejected mid-send, a temporary rate limit from Meta. Ask whether the platform retries these automatically or simply marks them failed and moves on, since the difference between the two can mean hundreds of customers missing an important message (an appointment reminder, a payment notice) without anyone noticing unless someone manually checks the send report.

Can you see what's happening without opening a ticket

Ask whether you get visibility into conversation logs, delivery status and provider errors directly on your own dashboard, or whether diagnosing a problem always means waiting on support to check logs you can't see yourself. Being able to look at a flow's execution log and see exactly which step a specific conversation failed on is a meaningfully faster way to fix an issue than describing symptoms over a support ticket and waiting for a reply.

This also matters for accountability during a busy period. A festival campaign or a sudden spike in enquiries is exactly when something is most likely to go wrong, and exactly when waiting on a support queue to diagnose it is most costly. Self-serve visibility into logs and status turns a potential half-day delay into a five-minute check.

Ownership: the real insurance policy against vendor downtime

The most important reliability question isn't about the vendor at all, it's about what you own if the vendor has a bad day. If your WhatsApp Business Account, phone number, AI provider keys and telephony account are registered to you (not the vendor), a platform-side outage is an inconvenience you can work around; if the vendor owns these on your behalf, an outage can mean you can't even message your own customers through another channel until it's resolved. Always ask directly: whose name is the WABA registered under, and can I export my number, contacts and flows if I ever need to leave?

How Convo360 answers these questions

Convo360 runs provider-health monitoring that detects when a customer's own AI or telephony provider errors, such as an expired key or an exhausted balance, and shows a clear alert on the dashboard rather than a silent failure. Nightly database and media/uploads backups run automatically. Failed sends in WhatsApp Studio broadcasts get retry handling rather than silently dropping. Analytics and per-step Flow Logs give visibility into exactly where a conversation succeeded or failed, without needing a support ticket to see it. Because the platform is BYOK, your WhatsApp number, WABA, AI provider keys and telephony account are yours, registered to your business and not the platform's, and flow, contact and template data can be exported, so a platform-level issue is never a reason you can't reach your own customers.

Step by step

  1. Ask about provider-failure detectionConfirm the platform alerts you when your own AI or telephony provider key or balance fails, rather than failing silently.
  2. Get specifics on backup frequency and scopeAsk for the exact schedule and what's included — database, contacts, flows and media — not just 'we have backups'.
  3. Check broadcast retry behaviourConfirm failed sends in a campaign retry automatically instead of being dropped without notice.
  4. Check what you can see without a support ticketAsk whether conversation logs and provider errors are visible on your own dashboard directly.
  5. Confirm ownership of your number and keysVerify your WABA, phone number, AI keys and telephony account are registered to you, and that export is available if you ever switch.

See Convo360's reliability features

Start free in minutes — no code, no credit card. Bring your own AI keys and pay no per-message markup.

Frequently asked questions

Is a high uptime percentage enough to judge a chatbot vendor's reliability?

No — most real disruptions come from a connected AI or telephony provider failing, not the platform's own servers being down. Ask specifically how the vendor detects and alerts on provider-side failures.

How often should a chatbot platform back up my data?

Nightly is a reasonable minimum for both the database and any uploaded media. Ask what's included in the backup, not just how often it runs.

What happens if a WhatsApp broadcast fails partway through sending?

A reliable platform retries individual failed sends automatically, using backoff for temporary errors like a Meta rate limit, and reports which contacts ultimately failed in the campaign log, rather than silently dropping messages without notice or leaving you to guess who was missed.

Why does it matter who owns my WhatsApp Business Account?

If your WABA, number and provider keys are registered to your business rather than the vendor, a platform outage or dispute doesn't cut off your ability to reach customers through another route.

Can I see why a specific conversation failed without contacting support?

On platforms with per-step flow logs and analytics visible on your own dashboard, yes — you can trace exactly which step a conversation failed on without waiting for a support reply.

Does BYOK affect reliability?

It shifts responsibility for provider uptime to your own AI and telephony accounts, which you control directly, while the platform's job is to clearly surface any failure on those accounts rather than hide it.

Related reading

Available across India — see WhatsApp & AI calling automation by state and city.