AI-Powered SMS Support, Lifecycle Messaging & RAG Knowledge Platform
A trial user and a five-year paying customer text the same number. The system already knows which one it’s talking to — and which voice, which knowledge and which history belong to them.
Client: Confidential (Health-Tech / Remote Biofeedback). Designed and built by Auto Core System.
The system in four numbers.
Client: Confidential — Health-Tech / Remote Biofeedback Platform. Figures describe the system’s configured, always-on behaviour in live production.
One system wearing three channels, not three systems bolted together.
This client — a health-tech company behind a remote biofeedback wellness platform — supports two very different audiences on the same product: people still inside their free trial deciding whether to convert, and long-term paying customers who need fast, accurate answers to keep getting value from their device.
We designed and built a connected automation platform that meets both audiences everywhere they show up — over text, on a set schedule, and inside their support inbox — without ever making them feel like they’ve reached a different company depending on which door they walked through.
Underneath, it works like a single mind wearing three faces. However someone reaches out, the same understanding of who they are, what they’ve already said, and what the product actually does travels with them. Nothing was bolted on after the fact; every channel was built from the start to share one memory and one source of truth.
Two audiences, one inbox, and no way to tell them apart.
- No automatic recognition. Trial users need onboarding help and a reason to convert; active customers need product support. Without recognition, every inbound message required a person to first work out who they were talking to before answering.
- No structured follow-up after signup. Trial users who didn’t reach out simply drifted — no consistent timed sequence, and no ongoing cadence keeping paying customers engaged once converted.
- Knowledge scattered instead of centralised. Documentation existed, but nothing connected it to the moment a customer actually asked a question. Answers depended on whoever was available knowing the right thing.
- A support inbox that was a filing cabinet, not a channel. Text conversations were logged after the fact, and customers who reached out through chat got a noticeably slower, more manual experience.
- No shared memory across touchpoints. A customer in touch for weeks had no continuity — each new conversation risked starting from zero.
A message arrives, from wherever it arrives.
Every row on the right runs unattended, across channels, without adding headcount.
| Without this system | With Auto Core System |
| A person has to read every inbound message before knowing whether they’re talking to a trial user or a paying customer | Who someone is gets recognised automatically, the instant a message arrives, on every channel |
| Onboarding follow-up depends on someone remembering to check in on trial users | A paced, multi-step journey runs unattended from day one through the end of the trial |
| Support answers depend on whoever is available knowing the right documentation | Every reply is built from the company’s real, current product knowledge |
| Support chat gets a slower, more manual response than text | Chat and tickets are answered with the same speed and knowledge as a text message, in real time |
| A conversation that moves between text and chat restarts from zero | Shared memory keeps context intact no matter which channel a customer picks up on |
| New documentation has to be manually re-taught to whoever answers questions | Updating a guide is enough — it’s understood and searchable within minutes |
What changed.
- Every customer gets the right response, automatically. No one has to work out whether an inbound message is from a trial user or a paying customer before responding.
- Trial-to-paid conversion runs on a schedule, not on memory. The structured onboarding journey reaches every trial user at the right moment, without depending on someone remembering to follow up.
- Support answers stay accurate as the product changes. Because every response is grounded in the company’s own living documentation, updating a guide is enough to update every answer — no retraining, no rebuilding.
- The support inbox becomes a real channel, not just a record. Customers who prefer chatting directly get the same instant, well-informed experience as customers who text in.
- Nothing gets lost between channels. Shared memory means a conversation that starts by text and continues in chat keeps its full context instead of starting over.
What it’s built on.
| Layer | Tools | Notes |
| Messaging | Twilio (SMS & WhatsApp) | Inbound and outbound customer messaging |
| Support & ticketing | Zendesk | System of record and a native reply channel |
| Orchestration | n8n (Cloud) | Coordinated automation across every channel |
| AI & language | OpenAI | Trial & Active voices, lifecycle message generation |
| Knowledge store | Supabase (vector search) | Shared knowledge base, tuned per audience |
| Conversation memory | PostgreSQL | Persistent history, kept per customer |
| Document source | Google Drive | New documentation picked up automatically |