Pricing thread
Finance recommends Option B pending approval.
Slack
Google Workspace
Salesforce
HubSpot
Databricks
Figma
Atlassian
Confluence
Jira
TableauSetup
AI native in one hour.
No migration. No rip-and-replace.
01
Claim your address
Pick ai.com/yourcompany. That address becomes the front door for every agent in the organization.
02
Connect approved systems
Choose which tools agents may read. Permissions follow the ones your team already has.
03
Invite the team
Everyone gets a personal agent on day one, scoped to exactly what that person can see.
People
Open a person. See the week they actually had.
Not a status column somebody remembered to update on Friday. Each person carries what they moved forward and what falls to them next, read off the work itself, true at the moment you look.
Six people are on the customer launch. Three moved it forward this morning; one of them is waiting on somebody else.
Maya OkaforProduct · launch owner
MovingHolding the launch sequence together and running the 10:30 customer call.
Moved forwardCleared the last two launch dependenciesLaunch plan · this morningOwns nextDecide launch pricingBefore the 10:30 call
Customer launch4 of 6 launch items closed
Team stateProductMovingFinanceReadyLegalApprovedCustomerHeld
Read off connected work. Never messages sent, hours logged, or keystrokes.
Personal agents
Every employee gets an agent of their own.
It reads what its person is allowed to read. It drafts, it gathers, it prepares — and then it stops, because sending is a human act.
AgentMorning. Three things moved overnight: Legal cleared the launch terms, Finance re-modelled the volume tier, and the customer moved the call to 10:30.
Prep me for the 10:30 customer call. Draft the follow-up and flag anything I need to decide.
Reading company sourcesLaunch plan · Pricing thread · Legal review
AgentThe call brief is ready and the follow-up is drafted. One thing is still open: Finance needs a pricing decision before I can finalize it.
Launch planPricing threadCustomer call
Needs you — nothing sends until you approve
The customer follow-up is drafted with Option B, addressed to the customer thread. From Workflow · launch follow-up · run 1.
Message @northstar-maya-ai
Rooms
People and agents in one timeline.
An agent joins the room with a handle, answers with its sources attached, and can be overruled in the next message. The decision stays human; the homework stops being yours.
Today
Sofia Lindqvist9:12
Customer confirmed 10:30. They want the rollout dates in writing before they sign anything.
@northstar-jon-aiJon’s agent9:33
Pulled the current terms into the room. Three items are open, all of them pricing-dependent.
Amara Bello9:38
Finance is ready on our side. I just need the pricing call made.
Maya Okafor9:41
Legal approved the terms. Can we confirm pricing before the customer call?
@northstar-maya-aiMaya’s agent9:42
Finance has one choice left — Option B at the volume tier, or hold at the current rate. Option B is the only one Legal has cleared.
Pricing threadFinance · 8:20
Recommends Option B pending approval.
Legal reviewLena Voss · Aug 19
Terms approved. Option B only.
Launch planMaya Okafor · Aug 18
Pricing is the final blocker.
@northstar-jon-aiJon’s agent9:43
Adding one thing: the volume tier changes the renewal date to 30 days after rollout.
Vantiv renewalContract · clause 4
Renewal follows the rollout date, not the signature date.
Jon Alvarez9:45
Then we take Option B and I’ll own the renewal date. Maya, that’s your call to make though.
Two agents answered. A human decided.
Message #customer-launch
Calls
Leave the call with the record already written.
The agent listens with permission, keeps the agenda honest, and captures what was decided and who owns it — so nobody spends the afternoon reconstructing the morning.
Knowledge
Ask the company. Get the evidence.
Ask the way you’d ask a colleague. What comes back is a written answer with numbered citations, and every citation opens the exact thing someone actually said.
Customer launch
The rollout is consolidated around a single customer plan. Legal has approved the terms; Finance owns the last pricing decision.
What is still open
Finance recommends Option B at the volume tier1, which is the only option Legal has cleared2. The volume tier moves the renewal to 30 days after rollout3, so the rollout date is now a contract term rather than a scheduling preference.
| Work | State | Owner |
|---|---|---|
| Legal terms | Approved | Lena Voss |
| Launch pricing | Needs decision | Finance |
| Customer brief | Ready | Maya’s agent |
Linked from · 3
Briefing
Vantiv renewal
Decisions
Search the knowledge base, or ask it a questionesc
Workflows
Describe the work once. Run it forever.
Type what should happen and when. You get inspectable steps you can edit, test on sample data, and hand to the whole company.
Build a workflow
Describe the repeatable work in plain language.DraftWhat should happen, and when?Every weekday at 8:00, summarize what changed on the customer launch and send it to my agent.
0/240
Running now4 active
Contract renewals30 days before end date · 2 upcoming
Weekly pipeline reviewMondays at 7:00 AM · last run clean
New-hire onboardingOn accepted offer · 4 steps
Invoice exceptionsWeekdays at 6:00 PM · 2 flagged
Workflow draftedReady to activate4 editable steps
1TriggerEvery weekday at 8:00 AM
2GatherLaunch room · Pricing thread · Calendar
3SummarizeWhat changed, what needs a decision
4DeliverSend to @northstar-maya-ai
Sample deliveredSample data · no live action taken
Workforce
Managing a workforce in the AI era.
When agents do the fetching, contribution stops looking like activity. Two readouts: who the company’s work is actually built on, and who deserves to hear about it.
What the work is built on
Built on the most 14Passes work along 9
An edge is a citation — someone’s work was built on. Never a message count.
Who the company is built on
| Rank | Person | Built on | Contribution |
|---|---|---|---|
| 1 | ABAmara BelloFinance · GTM | 41 | 94 |
| 2 | MOMaya OkaforProduct · Launch | 36 | 89 |
| 3 | LVLena VossLegal · Ops | 28 | 81 |
| 4 | JAJon AlvarezEngineering | 22 | 74 |
| 5 | DRDaniel ReyesLegal | 17 | 68 |
Role-weighted visible output. Its components sit beside it everywhere — gaming one shows in the other four.
Trust
Every model. Your company’s rules.
Model choice is yours
Company policy defines the approved provider set and how requests route between them. Swap providers without touching a workflow.
Permissions are enforced, not assumed
An agent inherits exactly the access its person already has. No shadow permissions, no privileged back door into a system.
Every action leaves a trail
Who asked, which sources were read, which policy applied, which model answered, what changed. Exportable, and queryable.
We don’t train on your data
Your prompts, documents, and agent outputs never train our models. Requests you route to third-party providers stay governed by that provider’s terms.
Dedicated environments
Company data runs in an isolated environment with segregated storage and per-organization encryption keys.
Approved providers
AnthropicOpenAIGoogleDeepSeekQwenKimiGrokGLM
Anthropic
OpenAI
Google
DeepSeek
Qwen
Kimi
Grok
GLMFAQ
Before you connect the company.
How quickly can we get started?
Claim the address, connect the systems you approve, invite the team. Most companies have a working core workspace inside an hour — and you can add connectors later without redoing anything.
What can each employee’s agent access?
Exactly what that employee can access, and nothing more. An agent inherits its person’s permissions from the systems you connect, and your company policy can narrow that further by connector, by team, or by model route.
Do you train models on our company data?
No. We don’t use your prompts, documents, or agent outputs to train our models. If you route a request to a third-party provider, that provider’s terms govern that request — and you decide which providers are available at all.
How is our data isolated from other companies?
Each company runs in a dedicated environment with segregated storage and its own encryption keys. Nothing is pooled across organizations.
What happens before an agent sends or changes something?
It stops and asks. Consequential actions — sending a message, changing a record, committing a decision — pause for human review, and the person sees the proposed action, the context, and the sources before approving.
Can we choose which models are available?
Yes, and we expect you to. You define the approved provider set and the routing rules; the operating layer enforces them on every request. Adding or removing a provider doesn’t break the workflows built on top.
Do we have to replace the tools we already use?
No. This connects to Slack, Google Workspace, Salesforce, Atlassian, and the rest of your stack rather than replacing them. The work stays where it lives; the agents come to it.
What does this cost?
Company pricing is coming soon. Claim your address now and we’ll walk you through it before anything is billed.
Give your company a brain.
Claim the address. Connect the work. Every person gets an agent that knows what they can see and what has to happen next.