AI at the core. Not bolted on
Most help desks staple a chatbot to a legacy inbox. TruSend is one AI engine — it answers your customers, drafts for your team, drafts your workspace setup and improves under test. Nerv and Portal are two faces of the same core
The difference is where the AI lives
Bolt a chatbot onto a legacy desk and it stays a bystander — its own prompt, its own logs, no safe way to ship a change. Build the desk on the engine instead, and the AI is in the work, not beside it
A chatbot stapled to a help desk
- A separate AI service, bolted on beside the desk
- Its own prompt, rules and logs — off to one side
- No versions, no canary — you can't ship a change safely
- The AI is a bystander to the work your team actually does
One AI engine, every surface a face of it
- Customer answers and every in-desk assist run one pipeline
- One prompt, one safety pass, one budget, one audit trail
- Shipped like software — versioned, canaried, sandboxed
- One log per request, so admins can review what it did
How the AI agent works inside
Every reply walks the same path — a real retrieval stack, not a prompt. Watch one request move through it, and see how the AI decided
Email addresses, phone numbers, card numbers and IP addresses detected in content sent to AI models are removed before it is sent. Contact details that your workspace publishes for customers in its knowledge base are not removed. Grounding and policy checks run before the reply is sent.
Below its confidence threshold it hands the conversation to a human in Portal, with the sources it was working from.
Every AI feature is that engine, keyed for the task
The customer bot and every assist in the desk run the same pipeline — one prompt, one safety pass, one budget, one trace. Only the task changes
Improve the AI without breaking what works
Test a smarter version in a live sandbox chat, roll it out to a 10% canary before everyone sees it, and publish when you’re sure. Knowledge gaps surface themselves. No prompt engineer required
Nerv couldn’t answer 22 requests about “SSO for enterprise.” Only 2 generic articles exist.
One predictable bill
Paid plans include a monthly allowance of resolves — conversations the AI closes end to end — and extra seats are a flat monthly fee. Hand-offs to your team are not billed as resolves. A live meter shows used, remaining and projected resolves
Your customers always know it’s the AI
The disclosure is on by default and locked — it stands down only once a real person takes over, never a silent switch. And every reply is checked on the way in and on the way out; email addresses, phone numbers, card numbers and IP addresses detected in content sent to AI models are removed before it is sent. Contact details that your workspace publishes for customers in its knowledge base are not removed
Data protection built in, not bolted on
On every plan: tenant isolation, encryption in transit and at rest, and PII redaction. Subscription payments are processed by Paddle, so your card details are not handled by TruSend
See it run on your own site
Paste your URL — the AI reads your live site and answers a real question grounded in what it finds, no setup. Then a working agent in about five minutes
Already running another help desk? Run TruSend alongside it and compare on your own traffic — how switching works →
See the support setup we’d start you with, drawn from your own site. Paste your URL