Ten client brands. One flat bill. Your server.

One comparison, one vendor, one published price. Wati is the clearest case because it publishes India pricing openly.

Year one, all in: ₹75,000 one-time onboarding + ₹2,50,000 prepaid = ₹3,25,000. Year two onward: ₹2,50,000. Both exclude 18% GST. Your hosting is billed to you directly by your provider — for most deployments that is a four-figure monthly cost, and we size the box with you before we deploy.

Automation is priced separately almost everywhere in this category. To illustrate with the clearest published figures available — these are AiSensy's own listed prices, quoted to show how the category prices automation, not as a criticism of AiSensy specifically:

• Drag-and-drop chatbot builder: ₹2,500 / month for five chatbots
• AI agent builder: ₹3,500 / month, including 3,500 AI messages

Those are per account. Across ten client brands, the chatbot add-on alone is ₹25,000 a month — the entire cost of running everything, everywhere, on your own server.

With us, the automation engine, the visual canvas, all nineteen node types, the REST API, the outbound webhooks and the n8n nodes are the product. There is nothing to unlock.

Meta's message charges are yours either way. For India: ₹1.09 per marketing template, ₹0.145 per utility or authentication template, service replies free inside the 24-hour customer service window, billed per delivered template. Every platform in this comparison passes these on. The difference is that you pay Meta directly against your own account, so there is no third party between you and the invoice.

Oriel Chat Engine runs on Oriel Inc.'s own WhatsApp Business Account, registered with Meta directly. We built it to replace AiSensy across our ventures, and we are migrating them onto it now — Adiyogi Naturals is the first.

We have not deployed this for an outside customer. You would be the first, and we would rather tell you that than let you assume otherwise.

So in place of a reference list we do not have, here is what we found while building it. Every one is a real defect in our own work, caught before it reached anyone.

A certificate that had been failing renewal for weeks, silently.

A Let's Encrypt renewal on one of our live products had been failing every few days. Nothing alerted anyone. We found it during unrelated work, by accident. The cause was a missing IPv6 listener in a config file — a one-line omission that made the renewal challenge unreachable. Had we not stumbled on it, the first signal would have been customers seeing browser security warnings.

Six defects sitting green in a passing test suite.

Not failing tests. Passing ones, structurally incapable of failing. An assertion written so that it could never be false. A test helper whose own string-trimming removed the exact defect it existed to catch. A published password that nothing compared against the stored one. All six were found by clicking through the product, not by writing more tests. A test suite verifies a system against itself; only real use verifies it against reality.

A deploy script that ran against the wrong directory three times.

A hardcoded default turned "I do not know which deployment this is" into a confident wrong answer. It reached production three times before anyone noticed. The fix was not a better default. It was removing the default, so the script refuses to run rather than guess.

An icon eight ΔE units from a colour we must not evoke.

Our favicon sat 8.6 ΔE from WhatsApp's teal — close enough to read as affiliation on a site about the WhatsApp Business Platform, which is precisely what Meta's guidelines forbid and what this entire site is structured to avoid. Every automated brand check we run passed, because none of them measures colour distance. Being recoloured now.

The principle underneath all of it.

The failure mode of this kind of system is silence. A campaign that stops sending. A queue worker that dies. A backup that runs and cannot restore. A webhook endpoint that fails, retries forever, and tells nobody. None of these announce themselves. You find out from a customer, or you do not find out at all.

So the alarm is built with the feature, not after it. Backups are verified by restoring them, not by checking that a file exists. A failing webhook endpoint trips a circuit breaker and someone is told. A query that names no tenant refuses to execute rather than quietly returning every tenant's rows.

That is what you are buying at this stage. Not a reference list — someone who looks for the failures that do not announce themselves, and tells you about the ones he finds in his own work.

The demo needs no form and no email address. If it does what you need, a call is worth thirty minutes. If it does not, you have lost nothing.