It learns the rules nobody wrote down
Who approves what, which customers get an exception, how the Monday report is really built. org.run maps how your company actually runs and uses that map to decide what to do next.
Refunds over $500 go to Priya.
The connected brain
A live map of how your company runs
People, customers, vendors, systems and commitments, and the real owners behind each. Not how the wiki says it works. How it really does.
Your tools
The systems org.run reads from and acts in — email, chat, CRM, finance, code, support and more.
Salesforce
Slack
GitHubEmail
Monitoring
Call recordings
Operations
SupportKnowledge base
What org.run has learned about your business — pricing, SLAs, approval limits and brand voice — applied automatically.
Projects
The live initiatives org.run pushes work into — linked to the tools, people and decisions behind them.
Rules & permissions
The limits you set — scoped access, approval thresholds, spend caps and data boundaries. It never crosses them.
People
Your team. org.run knows who owns what and routes each task and approval to the right person.
How it works
From history to rules you can read
- 01
It maps your history
Past tickets, approvals, deals and threads show who does what, for whom, and how decisions usually go.
- 02
It writes the rules down
Each pattern becomes a plain-language rule with an owner, a scope and the decisions it was learned from.
- 03
It asks about the gaps
Where decisions went different ways, it asks one specific question. Do enterprise refunds always go through Priya, or only above $1K?
- 04
It keeps learning
Your approvals, edits and rejections change how confident it is in each rule, so it asks less over time.
Off your plate
Stop being the place the rules live
When the only copy of a policy is in someone's head, every question routes to them. org.run takes that load.
- Answering who approves this, again
- Explaining the refund policy to every new hire
- Remembering which customers get net-60
- Briefing someone before they cover your leave
- Rebuilding context after someone leaves
- Hunting through threads for what was agreed
- Settling two teams that follow different rules
- Guessing what a policy change will break
What is inside
The part that decides, and remembers
Every rule is written once in plain language, then compiled into checks the agents enforce. People read the sentence. The system runs the check.
Company model
A live graph of people, teams, customers, vendors, systems, projects, contracts and commitments. Harborline's renewal being 26 days out comes from here.
Policy store
Rules, SOPs, decision rights, thresholds and tone. Each rule carries its source, confidence, owner, scope and version history, and it flags two rules that disagree.
Gap interviews
It finds decisions that nothing explains and asks the right person one specific question, instead of holding open-ended interviews.
Intake and triage
Turns signals into tasks. Removes duplicates, sets an owner, priority and deadline, and decides whether org.run, a person or both handle it.
Orchestrator
Splits work into steps, assigns them to agent teams, tracks dependencies across functions, and handles retries and escalations.
Learning loop
Approvals, edits and rejections change rule confidence. Each kind of action has its own autonomy level, shown as a dial you can see and set.
Rule editing
Edit any rule in plain language, then replay last month under the new rule to see what would have changed before it goes live.
Commitments tracker
Tracks the promises made in meetings and email, with their deadlines, so the Q3 pricing you said you would send actually goes out.
Cover for leave
When someone is away, their approvals and tasks go to the delegate they named, and come back when they return.
Questions about the company model
It reads the history in the tools you connect: past approvals, tickets, deals and threads. That gives it a first draft of how you work. Where the history is unclear, it asks one specific question instead of a long interview.
Yes. Every rule is a plain-language sentence you can read and edit, with an owner and a history. Before a change goes live you can replay last month under it and see the effect.
It flags the conflict and asks the owners to settle it. Until they do, anything the conflict touches stops for a person.
No. If you have them, it uses them. If you don't, it drafts them from how the work has actually been done, and you approve what it wrote.
Each rule carries the decisions it was learned from and a confidence level. Low-confidence rules ask before acting, and one rejection or edit is enough to correct it.
Early access
Get the rules out of people's heads
Connect your history and see the rules org.run finds in the first week.
Scoped, auditable and reversible from day one. Your data stays yours.