Built in a quoting office, not a lab
Tenderloop started as an internal tool for our own salespeople. It gave them their afternoons back — so we rebuilt it for everyone else.
Every morning, the sales desk at Alto opened an inbox full of requests for quote. Each one meant the same work: read the email, find every product in the price list, check the units and the quantities, copy the numbers into a Word template, export a PDF, write the reply. A quote that took twenty minutes to build took the customer ten seconds to read — and the next request was already waiting.
So we automated it for ourselves. The first version was a chain of off-the-shelf automation blocks wired together: it watched the mailbox, looked prices up in the list, and pushed a document out of the other end. It was held together with tape, and it still changed the working day of every person who used it. Quotes went out in minutes instead of hours, and nobody re-typed a price again.
The tape was the limit. Automation tools are very good at moving data between systems and very bad at understanding a customer — the same product named three different ways, a quantity in metres when the catalogue is priced per unit, a genuine request buried in a paragraph of small talk. We kept hitting the edge of what a flowchart can do, and past that edge a person still had to take over.
By the time the AI models were finally good enough to read those emails properly, we knew exactly what to build. We moved the whole pipeline off the automation tools and wrote a real product around it, with one rule we refused to bend: the model reads and matches the products, but every price comes from your own list and is recomputed on our servers. The AI never does the arithmetic that reaches your customer.
Tenderloop is that product. Nothing in it is theoretical — it is the tool our own team uses every day, generalised so that any company answering requests for quote by email can hand over the typing and keep the judgement.
From an internal fix to a product
- 1
An internal tool
Built for one sales team, to stop them re-typing prices out of their own price list and into a Word document. It was never meant to leave the building.
- 2
The limits of simple automation
Chained automation blocks took us surprisingly far, then stopped dead. Real customer emails need reading, not routing — and every exception landed back on someone's desk.
- 3
A product for everyone
Once AI could genuinely understand a request, we rebuilt the pipeline from scratch as a multi-tenant platform — and opened it up to industries far beyond our own.
Three things we refused to compromise on
Real quotes, not demos
Every decision was tested against actual customer requests, in a real catalogue, with the messy units and half-named products that come with it.
The price is never the AI's
The model matches products; your price list sets the price, and every line total is recomputed server-side. Anything off-catalogue is flagged for review, never invented.
No new tools to learn
Your mailbox, your Excel price list, your Word template. The work disappears; the way your team works stays exactly as it is.
Built and backed by Alto
Tenderloop is built and operated by Alto Perfis Pultrudidos Lda, a Portuguese manufacturer of pultruded profiles — the company whose sales desk this product was born on. They are still our most demanding customer: every release has to survive their inbox first.
Want to talk to the people behind it? Get in touch.
See it on your own price list
Start your free month, upload the list you already use, and watch the first quote answer itself.