How it works
From the first inquiry to the repeat customer: who does what.
A business is a chain of small jobs. Each system here takes one link, and a person takes the links that need judgment.
Free audit, no accounts, we change nothing.
Or ask a person: aimarketing@gonzo.co.in +91 7390990227 WhatsApp
Seven stages, and who holds each
A chain breaks at its weakest link: an inquiry that waits overnight, a quote typed by hand, a customer who never hears from you again. The table follows one customer from first search to repeat order and names the system on each link. It also shows where a person still holds it.
| Stage | What is happening | What covers it | Status |
|---|---|---|---|
| 1. Awareness | A stranger searches and finds you, or does not | The search and AI-answer agents audit your own site | Sandbox. No ranking is promised |
| 2. Inquiry | Someone writes or calls | Lead Response Engine for forms, WhatsApp and email. AI Receptionist for the phone and chat | Sandbox |
| 3. Qualify | Is this person a fit? | The same two ask the questions you set. The AI Outreach & Lead Qualification Engine scores a list you already have | Sandbox |
| 4. Quote and negotiate | Price, objections, a quote | The finance desk answers price questions, handles objections and counters, inside limits you set. Quote-to-Invoice Automation issues the quote | Sandbox. Above your limits it stops and asks you |
| 5. Onboard | The contract is signed | Client Onboarding Pipeline | Sandbox |
| 6. Serve | Day-to-day questions and work | E-commerce Support Assistant for orders and returns. Connected Operations Layer for one operator across your tools | Sandbox. The Connected layer is the least finished |
| 7. Retain | Keep them, collect, keep it all running | Dormant List Revival for quiet customers. Reminders and payment matching in billing. Managed Automation Retainer for the systems themselves | Sandbox |
At every stage a person can take over. The system hands off for a complaint, legal language, a large or important account, a request for a human, and any money question outside your limits.
Seven checks, always in the same order
No message or action gets out without passing all seven. This is the whole design: because the checks never vary, routine work can run without someone tapping approve on every message.
- Kill switch and breaker. Before acting, each agent reads the switch, and if it cannot read it, it stops. If failures repeat, a breaker trips, and only a person can reset it.
- Finance sort. Anything about money, coming in or going out, is sent to the finance desk. If a routine agent drafts a price, the draft is blocked and rerouted.
- Consent and opt-out. No consent record, no message. A STOP is honored at once.
- Your limits. The finance desk works only inside the limits you set. Above them, the question comes to you.
- AI disclosure. The first message says it is from an AI, on every channel, and no setting turns that off.
- Outbound scan. The finished draft is scanned a second time before it leaves.
- Audit log. Each action is appended to a log in which any later change shows.
The ladder: how much it does alone
| Rung | What it allows |
|---|---|
| Paused | Nothing runs. Each agent reads the engaged switch before it acts, and stops |
| Dry run | Agents do the full job but send nothing. Each message waits in an outbox, unsent, for you to read. Every system starts here |
| Routine work | Routine messages go out without approval: replies, follow-ups, reminders. Consent, disclosure, opt-outs and the daily cap still apply, and anything about money goes to the finance desk |
| Finance inside your limits | The finance desk quotes, counters and invoices without asking, but only inside your limits. Above them, the question comes to you |
You decide how high a system climbs. A system ships switched off, and turning it on is one recorded step by a named person.
Rules decide, the model words
Templates and rules choose what is said and what is sent. Where a language model is used, its job is wording only. Before the model sees a text, email addresses, phone numbers and amounts are replaced by pattern with placeholders, and names are replaced where they have been supplied. That lowers the risk. It is not proof that no personal data remains. If the model's version changes a figure or adds a claim, the version is thrown away. Which AI, and how personal data is handled is on the trust page.
What has run so far
- The nine systems and the finance desk have each been built and tested in a sandbox, with simulated customers, simulated messages and simulated money.
- The replay on the home page carries a label. It is either a sandbox run or, until its data has been generated from one, sample data written for layout. Read the label before the numbers.
- Nothing has run against a live mailbox, WhatsApp number, payment account or calendar. Each system's page lists what is still needed.
- Gonzo has no paying clients yet, so no page here carries a client result.
- A first live start would sit at the dry-run rung, and nothing would leave until you have read the outbox.
What goes wrong, and what happens next
A failed action is retried, then set aside in a queue you can see; it is not quietly dropped. A question outside the data you supplied goes to a person instead of being guessed. A STOP, an erasure request or a complaint ends the conversation after one neutral line at most, and you are told. The trust page covers data, disclosure and the limits, the doubts page lists what a careful buyer asks, and prices are fixed per system.
Start with the free audit.
Send one web address. In about five business days you get a short written report. If a system fits, a fixed-price scope follows.
Free audit, no accounts, we change nothing.
Or ask a person: aimarketing@gonzo.co.in +91 7390990227 WhatsApp