Beta

Send feedback
( 01 ) How it works

Four steps. You're in control of every one.

This is the one page where we open the hood. Everywhere else on this site we talk about what changes for you — here's the mechanism underneath it.

The four-step Provicity sequence: rich inputs, collaborative AI research, human review and decision, continuous learning loop
Step 01

Rich inputs

It starts with what you already have — your CV, your roles, your contacts, your process. No blank slate.

Step 02

Collaborative research

It cross-references companies, roles and the people behind them in the background, continuously.

Step 03

Human review & decision

Every action waits for your explicit approval. No unsupervised drift, no autopilot — supervision at every step, by design, not by exception.

Step 04

Continuous learning loop

What you approve, skip or edit teaches it what works — so it gets sharper every week.

( 02 ) The mechanism, in detail
01

It watches

Once you connect email, calendar and your records, it always knows where things stand — on both sides of every conversation: who said what, what was promised, what's due, and when someone last replied. Nothing has to be re-explained to it.

02

It decides

It's continuously asking one question: what would actually move this forward right now? Sometimes that's chasing a reply. Sometimes it's recognising nothing needs to happen yet, and scheduling the next check for the right moment — rather than pinging you constantly in between.

03

It acts or asks

Routine steps — updating a record, logging a reply, scheduling a follow-up — it just does. Anything that puts your name on something, or commits you to a decision, comes to you first: drafted, reasoned, one tap from done.

04

It learns

Every approval, edit and skip is a signal. Over time it learns which messages you send exactly as drafted, which you always rewrite, and which opportunities you pass on — and adjusts what it brings you next accordingly.

( 03 ) Step 03, closer up

Supervision at every step — that's not a slogan, it's the architecture.

Autonomous systems drift when nobody's watching. Ours doesn't get the chance to: every execution — sending, updating, committing — is a proposal until you approve it. Unchecked automation is where things go wrong; a human in the loop is where they don't.

Approve every actionSupervision at every stepThe danger of dark automation
( 04 ) Built to be trusted

Your data, properly guarded

Bank-grade encryption, strict isolation between organisations, and permissions down to the individual field. What someone shouldn't see is never even sent to their screen.

Intelligence, spent wisely

It draws on leading AI models from multiple providers and automatically routes each task to the most cost-effective one that can do it well, with built-in budget guards — never a runaway bill.

Configured in sentences, not code

What it watches, when it acts, what needs your approval — set up in plain language, or from a ready-made Business in a Box. Change your mind any time, no engineer required.

Business-in-a-Box templatesZero-code personalization
( 05 ) Start already built

Don't start from a blank page.

Every use case ships as a ready-made Business in a Box — the data structures, the dashboards, the connections, all wired up before you arrive. Point it at your inbox and calendar and it's live the same day. Everything about it stays adjustable afterwards, in plain language, with no engineering team required.