Where a typical chatbot guesses, ours stops and asks you.
No names, no call-outs — just the two behaviors side by side. The difference matters most on the two things that cost you money: a price, and a date.
| Capability | Typical AI chatbot | Saleringo |
|---|---|---|
| Source of answers | Scraped pages, whatever it found | Facts a human approved |
| When info conflicts | Picks one, silently | Stops, flags, asks you |
| When info expires | Keeps answering | Stops answering |
| When unsure | Guesses confidently | Hands off with context |
| If a customer disputes what they were told | A chat log, and an argument | The approved answer it used, who signed it off, and when |
| Learning from chats | Auto-learns | Queued for your approval |
Somebody on your team
ticks every fact.
Import drafts it. Your desk approves it. Until that tick lands, the answer does not exist as far as a customer is concerned — and when it expires, it stops existing again.
Photography is illustrative — the kind of business Saleringo is built for, not named customers and not testimonials.
How an answer gets
permission to exist.
Your website and price list come in. Somebody on your team ticks the ones the AI may use, and sets a date to look at them again. Everything it later quotes, books or files into a customer record traces back to one of those ticks — nothing answers by accident.
Read a full call, start to finishSix stages, and a person stands at the one that matters. An answer that nobody signed off never reaches a customer.
When a conversation goes to a human
A meaningful minority of conversations should reach your team, by design. We are not going to promise you a fraction — the dashboards on this site show sample data; the only honest number is the one your own logs produce after a month. What we can tell you exactly is when it happens. These are the five triggers.
No approved fact matches the question closely enough. Instead of guessing, the AI hands over.
→ Not certain enough — handing to your teamTwo approved facts disagree — say, two prices for one package. Answering pauses on that topic until you resolve it.
→ Paused — two prices found for one packageThe only relevant fact is past its expiry date. Last season's prices and old hours never reach a customer.
→ Fact expired — a person will confirmMedical, legal, safety, or complaint territory goes to a person — however sure the AI is.
→ This needs a person — handing to staff"Can you guarantee the date?" Booking confirmations, dates, and contract terms always go to a person.
→ Date confirmation needs a person- Summary already written
- Full transcript, translated
- The facts the AI used
- A suggested reply to edit
Before the handoff
Hands full, question open.
The five triggers exist so the questions that need you actually reach you — instead of being answered badly while you are mid-order.
A summary, not a transcript dump.
Your desk opens one card: what was asked, what the AI already answered, which facts it used, and a reply waiting to be edited.
The photograph is illustrative — the kind of work these handoff rules were written for, not a named customer and not a testimonial.
What every answer leaves behind
A chat log tells you what was said. A verified answer also tells you why it was allowed to say it.
What this does not do
A product that states its limits out loud is a product you can trust on everything else.
It does not make the AI perfect.
It makes the AI accountable to you. Facts can still be wrong if what you approved was wrong — but you'll always know which fact was used.
It does not replace your team.
A hand-off is the design, not a shortfall — and the five triggers above decide when one happens, not a quota we set.
It does not sign contracts.
Saleringo will not confirm a booking, promise a date, or agree to terms on your behalf. Those always go to a person.
The deny list is a file you edit, not a promise we make. Every trade gets a different one — a dental boundary and a venue boundary are not the same shape.
Now go and see
what it actually does.
Answering accurately is table stakes. Booking against your calendar, quoting from your own price list, filing the customer and showing you on Monday what it earned — that is the product.