Collect

Ask forty people for the same things. Get back one answer.

Mabel Collect sends each person a link, works through your list with them in a conversation, keeps every answer in a typed field with the documents attached — and then reconciles all of it into one document you can act on, with the evidence behind every line.

Free to set up. Write your first template, send one link to yourself, and see what comes back before you invite anybody.

Three steps, then it runs itself

Nothing for the person answering to install, and no account for them to create.

Write what you need

A markdown list, one line per field. Start from a pack — client onboarding, supplier compliance, pulse, AI readiness — and edit it, or start from nothing.

Send the links

One unguessable link each, to four people or four hundred. No account, no password, no app. They can stop halfway and come back to exactly where they were.

Read one answer

Every response reconciled into tables and cited claims. Not four hundred transcripts to read — one document, and the evidence behind each line of it.

The same machine, five jobs

Every one of these ships as a starting template. Take it as it is, or edit it into the questions you actually ask.

Accounting and law firms

Client onboarding, before you may act

Send one link to a new corporate client. The agent works through the company, what it does, who represents it and who ultimately owns it — and where the answer is "a few shareholders", it asks for names and percentages instead of writing that down.

What comes backA complete client file with the documents attached, an approve / reject / more-information decision on the record, and a five-year retention floor that nothing deletes early.

Procurement

Supplier compliance, once a year, across forty suppliers

Certifications, insurance, subcontracting and incidents — asked of everyone at once, in their own language, with certificates uploaded against the item they belong to rather than emailed as a pile.

What comes backHow many hold what, every certificate and policy that lapses inside the next twelve months, and a list of exactly who still owes you which items.

People teams

An employee pulse that produces sentences, not just a score

Seven statements on a scale and three questions in their own words. Nobody is chased by a spreadsheet; the agent asks, they answer when they have five minutes, and they can stop and come back.

What comes backA score per statement and per team, the words that keep recurring in what people would change, and every comment reproduced in full.

Consultants and COOs

An AI readiness probe, one respondent per team

Where the repetitive time actually goes, how ready the systems and the people are, and — the question most readiness surveys leave out — what in this work must stay a human decision.

What comes backA readiness index by dimension, an estimate of the hours a year that repetitive work costs, a coverage map of who answered, and what each team said must not be automated.

Anything with the same shape

Emissions data, renewals, declarations, grant reporting

Write the list yourself as a markdown document. Each line is a field with a type — a date, a number, a yes/no, a document — and the agent collects exactly those and refuses to invent one you did not ask for.

What comes backOne table across every respondent, and a template you can keep editing as you learn what you actually needed to ask.

Careful with other people’s information

You are asking someone for their passport, their owners or their honest opinion of their manager. That has to be handled properly, or not asked for at all.

A conversation, not a form

It asks the follow-up a form cannot: “you said a lot — how many a week?”, “that shareholder is a company, so I need its owners too”. Where an answer is genuinely not applicable, it says so and moves on rather than inventing a value to turn a box green.

They can see the whole list first

The requirements are visible before anyone starts, so your client can go and find their documents. And the progress counter is honest, because completeness here is objective — a required field is filled or it is not.

Typed values, not a transcript

Every answer lands in a named field with a type, and carries a link back to the sentence in the conversation that produced it. You are not reading chat logs a week later trying to find where somebody said 2019.

Documents come back too

Uploaded against the item they belong to, checked by their actual contents rather than their file extension, and kept in a vault — never a public link, and served only to the person reviewing the file.

Every number carries its evidence

A line in the output resolves to the records behind it, and each record to the words somebody typed. A citation that does not resolve is removed, and a claim left with none is dropped whole — so nothing unsupported can borrow the authority of everything around it.

The calculator is yours

The script that turns forty responses into one number is an ordinary skill in your workspace. Open it, change the thresholds, change what counts as a pass. Aggregating your own data should not need us to ship a release.

Two things worth knowing before you ask

Answers are attributed

Every response is tied to the person who gave it, and an employee survey says so on its first screen rather than in a footnote. An anonymous mode — aggregate-only, with a minimum team size below which nothing reports at all — is being built, and it is a guarantee rather than a setting, so it will arrive when it is real.

It is early, and we are close by

You can set this up yourself in a few minutes, and most people do. But Collect is young, the packs are opinionated, and the first campaign is the one worth getting right — so write to us and we will build the template with you and watch it land.