Trust and governance

Your company deploys it. Your company sets the rules.

Your company deploys Nolli on company devices and decides what it may learn from. Every finding shows its evidence, and agents act only within the approvals you give them.

Request a demo Email the founders

Company policy

What Nolli may learn from

Set by your admin

Work apps

SalesforceOn
Gmail and CalendarOn
Google Sheets and SlidesOn
SlackOn
AI agent logsOn

Always kept out

Passwords and password managers
Login, checkout and billing pages
Card numbers, keys and ID numbers, removed before any model

Where data is kept: Your company decides
AI models and keys: Approved models, your keys
How long it keeps what it learns: Set by your admin
Agent actions: A person approves anything risky

The privacy model

Five rules, for every team and every agent.

They hold for every source Nolli learns from: work on company devices, the email and calendar you connect, and the logs of the AI agents people use.

Your company sets the policy.
An admin decides where Nolli keeps its data, which AI models and keys it uses, which systems and documents it may read, and how long it keeps what it learns. Employees never have to make these calls.
It runs where your company deploys it.
Your IT team installs Nolli on company devices, and your admin decides which work apps, systems and documents it may learn from. Your admin can see, correct, exclude, export or delete what it learned.
Sensitive data comes out first.
Passwords, password managers, and login, checkout and billing pages are kept out. Keys, card numbers and ID numbers are removed before any AI model sees the work.
Findings are about the work, not the person.
Leaders see workflows, patterns and results across teams. Each person sees their own work map.
Every finding shows its evidence.
Each step, number and recommendation links back to where it came from: a chat, a system record, a document or observed work. Corrections are kept, never silently overwritten.

Governance

Control over agents, AI tools and evidence.

Agents

Agents work inside your guardrails.

Nothing an agent drafts or suggests is sent on its own. A person approves anything risky, where your team already works, like Slack or email. Every run leaves a record of what the agent read, what it did and who approved it, and an agent gets more responsibility only as it proves itself on your own work.

A person approves anything risky
An audit trail for every agent run
Shadow mode before anything goes live

Guide: AI agent observability and governance

Governance Your company

Live

ApprovalsAgents ask a person before anything risky, in Slack or email.On for every agent
Audit trailEvery agent action is recorded: what it read, what it did, who approved it.Kept for every run
AI tool policyWhich AI tools and data each team may use.14 tools approved

AI tools in use

Chat assistantApprovedCoding assistantApprovedMeeting notesApprovedSlide generatorIn reviewBrowser extensionUnapproved

AI tools

Every AI tool on the record.

See which AI tools and models each team uses, approved or not, and set which ones are allowed and with which data. Shadow AI is found, not guessed at, and so are the seats nobody uses.

Every AI tool in use, by team
A policy for which tools and data are allowed
Unapproved tools flagged as they appear

Guide: What is shadow AI?

AI spend Tools and seats

Updated today

AI tools in use1714 approved

Seats unused31%across all AI tools

Seat savings found$6.9ka month

Chat assistant420 seats · 61% activeReclaim 160 unused seats$4.8k a month
Meeting notes2 tools · Same jobKeep one of the two$1.6k a month
Slide generator40 seats · 22% activeIn review: few people use it$0.5k a month
Coding assistant180 seats · 92% activeHealthy—

Plus $10.8k a month (22% of model spend) from cheaper models that pass your checks

Evidence

Every finding shows where it came from.

Each step on the map, each number in a benchmark and each line of a PRD or SOP links back to its source: a chat, a system record, a document or observed work. When someone corrects Nolli, the correction is kept and the old version stays on record.

Sources on every finding
Corrections kept, never silently overwritten
People and agents scored on the same checks

Guide: How to map a process for AI

Weekly revenue review prep Plan

PRD · Weekly revenue review prep

Problem

Six people spend about 14 hours each week assembling the Monday revenue review from five apps.Observed 4 Mondays6 people

Requirements

Gather account updates from email threads, #revenue-updates and Salesforce stage changes.14 threads a week
Match closed-won deals to the forecast workbook. Flag any conflict for a person to resolve.3 conflicts last month
Draft slides 2–4 from the workbook’s summary tab.Same layout 4 weeks running

Out of scope

The final recommendation stays with the Revenue Ops lead.Always written by hand

Built from four weeks of observed work. Every line links to the evidence behind it.

Where your data lives

Your company decides.

Where data is kept and which AI models it reaches are company decisions, not employee ones. Your admin sets them once, as policy, and Nolli follows that policy for every person and every agent.

Where Nolli keeps its data
Which AI models and keys it uses
What it may read, and how long it keeps it

Questions from your security team? Email the founders

Questions leaders ask

Good to know

Something we haven’t answered? Email the founders.

Is Nolli employee monitoring?

No. Nolli studies how the work gets done, not how hard someone works. Leaders see workflows and patterns across teams, and your admin decides what Nolli may learn from.

What does Nolli learn from?

The work done on company devices, in the apps your admin allows: which app and screen is in use, the steps of a task and the text that matters to it. Plus the email and calendar your company connects, and the logs of the AI agents people use, such as Claude Code and Codex.

Where does Nolli keep our data?

Your company decides. Your admin sets where Nolli keeps its data as part of your company’s policy, and Nolli follows it for every person and every agent. Employees never have to make that call.

Which AI models does Nolli use, and whose keys?

The models and keys your company approves. Within them, Nolli picks the cheapest model that passes your own checks for each task, and shows you the cost per task.

How is sensitive data handled?

Passwords, password managers, and login, checkout and billing pages are kept out. Keys, card numbers and ID numbers are removed before any AI model sees the work.

Can we see and delete what Nolli learned?

Yes. Your admin can see what Nolli learned, correct it, exclude work that shouldn’t be included, export it or delete it.

Can agents act on their own?

Only within the approvals you set. Nothing an agent drafts or suggests is sent on its own: a person approves anything risky, and every run is recorded in an audit trail.

Does Nolli train AI models on our data?

No. Nolli doesn’t train, fine-tune or adapt general AI models on your company’s data.

How can our security team review Nolli?

Email the founders. We’ll walk your security team through how Nolli works and how your policy is applied.