Two to ten, one Nova, roles that do not collide.

What breaks when two to ten people share Nova without roles?

A small team breaks when everyone can search, everyone can send, and nobody owns the domain. Two AEs mail the same owner. A builder deploys while a strategist is still on the diagnostic. A contractor replies from Echo with a tone the account lead would never use.

The product is the same loop as Solo. The frame is collision. Shared guardrails. Shared roster. Separate work. Work means a business uuid, a Pulse board, a sender, a portal, a DEPLOY target.

Nerxus One will not invent a full enterprise RBAC brochure on this page. It will say Supervised is the default, Relay needs a named approver, DEPLOY needs a named approver, and Scout should not be able to contact anyone even if a junior finds the send button in another tool. Nova is shared. The roster does not fork. Voice stays inbound. DEMO is a Forge sample. Health stays a band. v1 still has no white-label sub-account resale. Overnight runner is still not a production claim.

If your headcount is two and your books are many, you also need the agency pages. Size and team fear stack. They do not replace each other. Clerk org is the tenant. Two legal brands should assume two tenants. A business owner who wants to sit in Nova hears the same no as the agency size page.

How should a two-to-ten shop divide Scout, Radar, Forge, and Relay?

Give hunt and save to the person who can write an ICP. Give deep diagnostic to the person who can defend modeled dollars on a call. Give Forge to the person who understands DEMO versus DEPLOY. Give Relay to the person whose name is on the domain reputation.

Nova stays shared. It is a door, not a personality each seat customizes into a different company. Slash-delegate if a specialist is obvious. Do not run five contradictory prompts on the same row in the same hour and call it collaboration.

Pulse boards are how the rest of the team sees the hunt without exporting CSV. Signal-named. Next action visible. DEV and FEE split. A board called "General" is how two AEs collide.

Steward belongs to whoever owns the relationship after the win. Leftover findings should not silently become a builder's side project on the wrong client.

How does a shared Autonomy Matrix stay shared without becoming a free-for-all?

A principal sets the matrix. Per agent, per action. Juniors can live on Supervised even if Scout search is looser. Relay and Muse publish and DEPLOY stay tight until the team has a month of clean activity.

The four locks are not a team vote. Cold voice, cold SMS, review gating, and the spirit of the email gate do not flip because a standup wanted volume.

Mission Control shows approvals and activity. Use it. A small team that ignores the queue will still say they are Supervised while three people click yes on the same thread.

Overnight dispatch, when the runner is real, is a principal decision. Until the runner is the open half, do not tell the team playbooks already ran in production. Governance is live. That sentence is allowed.

What does Monday look like when two people hunt and one person fulfills?

Monday the hunters run ICPs that the fulfiller can actually build. A hunter who saves medical ADA findings for a studio that only ships brochure sites is creating Wednesday fights. ICP exclusions are a team document, not a private note.

Saves should say who owns the next action. Pulse can hold that. A Slack screenshot of a Scout list is how rows get worked twice.

Watchtower diffs should land on a board, not in five DMs. Timing is a finding. It is also an assignment.

Keys stay in the tenant. A hunter who pastes a Places key into a personal script is off the team, not off a guideline. Secrets never belong in the browser or in a chat.

What does Wednesday look like when strategist and builder share one finding?

Wednesday the strategist finishes Radar before the builder treats Forge as a blank canvas. The fulfillment action id is the brief. A builder who ignores it and "makes it nice" is back to spec work.

Atlas branding should be a profile the team agrees on: agency mark, or a client-facing cover. Two covers for the same memo is how you look unprepared on a call.

The AE waits for a DEMO link that is real. Relaying a promise that the builder will "have something by Friday" is the old motion. Sample-first means the link exists when the mail exists.

If the strategist and builder disagree about the finding, fix the diagnostic. Do not A/B two sites to the same owner from two seats.

What does Friday look like when one person closes and another runs Steward?

Friday the closer hands a business uuid, not a PDF and a hope. Steward already knows the leftovers if the diagnostic was done. The account lead should not re-audit in a new doc.

Portal language should be written once. Plain. No costs. No internal jokes. A small team will leak tone if five people type into the client surface.

Echo replies need a voice guide the same way Muse needs an approval. Two reviewers with two tones is how a shop sounds drunk.

Health bands keep the Friday standup honest. Healthy, Watch, or At-risk. Do not let a hunter paste a modeled DEV number into Slack as if it were cash collected. Forecast FEE. Atlas still refuses bare dollars on the memo that leaves. Portal language is written once. Echo needs a voice guide by vertical.

Which compliance lock matters most when juniors can reach the agent screen?

Relay and the collect page. A junior sending from an unverified domain, or "improving" collect by hiding public buttons they saw on a YouTube funnel, is an incident. The product blocks the second. The first is a verification gate plus a role.

Voice and SMS locks matter the first time a client asks the intern to "have the AI call the list." The intern needs a sentence that is true: locked.

DEPLOY is a compliance event for studios inside the team. DNS is not a group click.

Medical and professional-services limits are team training, not a tooltip. If you hunt those verticals, everyone who can draft a reply reads the industry page.

How do shared senders and shared calendars go wrong at this size?

One domain for every offer and every metro is convenient and fragile. When a complaint arrives, the whole team stops. Tighter senders cost setup. They save the book.

Marketer calendars must not merge with client-business booking grids. Two to ten people will mix those links in a hurry. Meetings product copy is isolation. Follow it.

Forms consent is the only SMS birth. A teammate importing a spreadsheet of numbers into a text tool is not a Nerxus workflow and not a loophole we will bless.

If two brands share a calendar because "we are small," you are already in the mix-up. Split the grid or split the tenant.

How should a small team use Pulse so two AEs do not work the same owner?

One row, one owner, one board. Signal-named campaigns. If two AEs sell different offers into the same metro, the ICP and the exclusions must say so, or the listing ID will still save the same shop twice and both will call.

Stop-on-reply is team-wide. An AE who keeps a personal sequence running in another mailer after the inbox classified "not now" is the collision.

Forecast FEE, not DEV. Two AEs stacking modeled recovery will lie to the principal. Atlas credibility gate exists because that lie is common.

When a row closes, it leaves the hunt board. A closer who keeps won deals in "negotiation" so the column looks healthy is training the hunters to distrust Pulse.

When should a small team stay split across specialist tools?

Stay split if each human is a specialist firm under one LLC and the only shared object is a bank account. Forcing them into one tenant will create the mix-up on purpose.

Stay split if GoHighLevel already is the OS after the close and you only need detect-and-sample in front. Do not migrate on a size page's advice.

Come together here when the hop is the tax: list in one tab, PDF in another, site in a third, mail in a fourth, reviews in a fifth, and nobody can find the uuid.

Lighter tools still win at their one job. We close the loop. Size does not change that trade.

What should a principal refuse the first month a second seat joins?

Refuse shared personal Gmail. Refuse a contractor on DEPLOY. Refuse autonomous Relay as a welcome gift. Refuse white-label promises to make the new hire's quota easier. Refuse a fake team price on a proposal.

Refuse hunting verticals the fulfiller cannot support. Refuse unknown sources displayed as green. Refuse Yelp live-reply claims while assisted.

Refuse putting client revenue on a health widget. Bands only.

Refuse a second Clerk org "just for testing" that still uses the production domain. Test mail on a domain you can burn, or do not send.

How does a small team become an agency size problem without noticing?

The tenth client and the second brand. If portals, senders, or keys need a harder wall than roles, you are on the agency size page. Isolation is the feature. Resale is still not in v1.

Headcount eight with one book is still a small team. Headcount three with two legal entities is already an isolation conversation. Count books, not only humans.

Nova can stay shared inside a tenant. Nova cannot be the excuse to retrieve the other brand's diagnostic because the names sounded alike.

Read Agencies for the fear story you will tell clients. Read Agency for the procurement sentence. Stay here for who clicks what this week.

How should a small team write ICPs so hunters and builders stay on the same offer?

The fulfiller writes the exclusions. The hunter writes the metro and the vertical. Together they name the problem they can DEMO this week. An ICP that only the hunter loves is how Wednesday starts with a fight.

One offer per profile until you have a month of clean handoffs. Stacking dentists, roofers, and clinics in one hunt because "local is local" is how medical limits get ignored by the person who only wanted booking DEMOs.

Share the ICP object, not a Slack paragraph. Tomorrow's hunt should be the same filter. Watchtower needs a stable saved set. A team that changes the paragraph every standup will never see a useful diff.

When two AEs sell two offers in one metro, two ICPs and two boards. The listing ID will still save the same shop once. Ownership then decides who speaks. A shared "we will both try" is the collision.

What does the inbox require when more than one person can reply?

One inbox, classified, stop-on-reply. Suggested reply still gated. A second personal mailbox running a leftover sequence is how a small team looks like two vendors.

SLA clocks sit on a named human. "The team will get back" is how a question sits until the owner goes cold. Pulse can show the clock. Someone still has to own it.

Tone belongs in a short guide the same way Muse belongs to an approver. Two AEs writing like two brands on one domain is a reputation finding you gave yourself.

If a contractor needs inbox access, they need a seat and a role, not a forwarded password. Sharing the Clerk login to stay "two to ten on paper" is how you skip the role talk and earn the mix-up.

How do you handle a contractor who only builds and never hunts?

Give them Forge and the uuid. Do not give them Relay. Do not give them DEPLOY until they have shipped DEMOs you would send. The fulfillment action id is their brief. A contractor who ignores it and rebuilds the brand for fun is spec work you are paying for.

Pay them on FEE you collected, or on a craft rate you set. Do not pay them on DEV. Modeled recovery is not a contractor bonus pool.

Their Places key is not your Places key. Secrets stay in the tenant. A contractor who copies listings into their own scraper is off the team.

If they want to resell the chrome as their own AI studio, that is the v1 Not doing. They can sell the work. They cannot sell a sub-account factory we do not ship.

How should a small team brand Atlas and the portal so two humans do not invent two companies?

One branding profile per legal brand, with a live cover preview. The marketer owns the artifact. Two AEs uploading two logos because they each like a color is how a memo looks unprepared on a call.

Portal language is written once. Plain. No costs. No detector names. A builder who pastes Forge jargon into the client surface is leaking the shop. An AE who pastes a modeled DEV number there is leaking a lie.

If you operate two brands in one tenant, you are already in the isolation conversation. Split tenants or accept that a wrong cover is an incident. Size is not the same as legal entity.

v1 will not give each client a chrome-resale theme. Your mark on the report is not a sub-account. Do not promise the client a fake product login to make the new hire's deck prettier.

How do you pause a teammate without pausing the whole book?

Take them off Relay and DEPLOY first. Scout can stay if they still write ICPs. Pulse ownership moves to a named human. A contractor who leaves with the inbox password is the small-team version of a burned domain.

Kill switch is for a sequence or a key event, not for a mood. If their drafts are the problem, Supervised already required a yes. Ask who clicked.

Watchtower keeps running on the saved set. Timing findings do not belong to a person who quit. They belong to the uuid.

If they take a CSV with them, treat it as a leak. The product's job was to make the daily export unnecessary. Humans still download things. Isolation does not follow the laptop.

How should a small team run Watchtower so diffs do not die in five DMs?

Diffs land on the board that owns the uuid. A rating drop, a permit, a hours change. Someone is named. Five Slack screenshots is how two AEs both mail, or how nobody does.

Freshness clocks match the vertical. Roofers after hail are not a law firm's practice-area page. The hunter who wants a daily metro scrape because the channel felt quiet is paying for theater.

Unknown sources stay unlabeled as live. A Watchtower tile that implies Yelp depth you do not have will be screenshotted into a QBR. The principal owns that lie even if a junior colored it.

How should a small team handle a client who wants to sit in Nova?

They do not. The SMB is the customer. The portal is the surface. A business owner in the agent screen will see other rows, detector names, and the next hunt. That is the mix-up. v1 has no white-label seat for them.

If they want to approve copy, you still approve Muse and Relay unless a later product puts a thin button on their side. Read-only is the honest sentence in the kickoff. Selling them a fake login is the agency fear arriving early.

A business owner who insists on a seat should hear the agency size page: isolation is the feature, resale is not. If they still want chrome, wait. Do not share the tenant to be polite.

If the request is really that they want to see progress, send the portal link. Plain language. One business. That is the seat they can have in v1. Nova stays on your side of the glass. Roles stay on yours too. Shared Nova is not a shared inbox with the shop. That split is the size rule.

How does a two-to-ten shop join the waitlist without inventing an enterprise?

Say headcount, how you split hunt and fulfill, and how many books. Use /demo if you are evaluating at volume. Batches open. No dollar table.

While you wait, write the role sheet: who approves Relay, who deploys, who owns Pulse boards. The ten-clients guide is useful even if you never buy.

When seats open, one tenant unless two entities demand two. Places keys as the contract allows. One ICP per offer. Supervised. Named approvers. Then hunt.

If you are still one person, you wanted Solo. If you are already leaking books, you wanted Agencies. Same product. Pick the sentence you needed.

How should a small team handle Approvals so two people do not clear the same send?

A small team should handle Approvals so two people do not clear the same send by naming who may click yes and treating the queue as one item with two doors. Chat and the Approvals table should agree. A queue anyone can clear is how collisions ship.

Approve re-checks the gate. Edit merges reviewed input and re-checks. Reject down-weights the pattern. Snooze stores a date. A toast that hides the card is not a resolve. Mission Control shows the pending count. Treat that number as a job, not as proof the overlay is working.

A principal sets the matrix. Juniors can live on Supervised even if Scout search is looser. Relay, Muse publish, and DEPLOY stay tight until the team has a month of clean activity. The four locks are not a team vote.

Overnight dispatch, when the runner is real, is a principal decision. Until the runner is the open half, do not tell the team playbooks already ran in production. Governance is live. That sentence is allowed. A small team that ignores the queue will still say they are Supervised while three people click yes on the same thread.

What does a shared Places key cost you besides quota?

A shared Places key costs you besides quota the isolation story, because the key is per tenant and sharing a tenant to share a key is how two books touch. Quota is yours. Do not put a client's key in the agency tenant if the contract says otherwise.

A hunter who pastes a Places key into a personal script is off the team, not off a guideline. Secrets never belong in the browser or in a chat. The browser sees masked status. Rotating a key does not destroy the business uuid. Handoffs persist first.

Two legal brands mean two tenants and two key rings. Saving a seat by sharing a Places key across brands is how a QBR slide leaks a competitor-client's metro. Platform quota is not a shared pot we burn to make a trial look generous.

If you need a hunter in another country, they hunt inside the tenant, not inside a pasted .env. Headcount eight with one book is still a small team. Headcount three with two legal entities is already an isolation conversation. Count books, not only humans.

How do you onboard a hunter who has never seen a DEMO?

You onboard a hunter who has never seen a DEMO by running one hunt through Radar and Forge before they get Relay. Give them an ICP the fulfiller wrote exclusions for. Give them a Pulse board. A hunter who saves medical ADA findings for a studio that only ships brochure sites is creating Wednesday fights.

The fulfillment action id is the brief. A builder who ignores it and makes it nice is back to spec work. The AE waits for a DEMO link that is real. Relaying a promise that the builder will have something by Friday is the old motion. Sample-first means the link exists when the mail exists.

If the strategist and builder disagree about the finding, fix the diagnostic. Do not A/B two sites to the same owner from two seats. Atlas branding should be a profile the team agrees on. Two covers for the same memo is how you look unprepared on a call.

Nova stays shared. It is a door, not a personality each seat customizes into a different company. Do not run five contradictory prompts on the same row in the same hour and call it collaboration. Slash-delegate if a specialist is obvious.

What does Friday standup look at if DEV is forbidden as cash?

Friday standup looks at FEE invoiced, leftover findings, Low-band rows, and who owns the next action, if DEV is forbidden as cash. A hunter who pastes a modeled recovery number into Slack as if it were collected is training the room to lie. Atlas credibility gate exists because that lie is common.

Health bands keep the standup honest. Healthy, Watch, or At-risk. You cannot drag a card to pretend they are fine. Last touch is staleness, not a claim the shop is dying. Combine it with payment and asset errors before you start a win-back draft.

When a row closes, it leaves the hunt board. A closer who keeps won deals in negotiation so the column looks healthy is training the hunters to distrust Pulse. The account lead should not re-audit in a new doc. Steward already knows the leftovers if the diagnostic was done.

Portal language should be written once. Plain. No costs. No internal jokes. A small team will leak tone if five people type into the client surface. Echo replies need a voice guide the same way Muse needs an approval. Two reviewers with two tones is how a shop sounds drunk.

How should a small team write a voice guide for Echo?

A small team should write a voice guide for Echo by vertical, not by mood, because a burger complaint and a medical wait-time review do not share a tone. Drafts stay drafts. A named human posts. If you cannot staff that human for clinics or firms, do not sell those books.

The medical page already said auto-reply is how you get famous in the wrong way. Professional-services pages said fee disputes are not auto-reply. A reputation shop that ignores those limits because the inbox is one product will learn it in a screenshot.

Low-star replies stay held. Public buttons stay public. Private recovery sits alongside. There is no hide-the-Google branch a junior saw on a YouTube funnel. The product blocks the branch. The intern needs a sentence that is true: locked.

Yelp and Tripadvisor stay monitor or assisted until credentials are exercised. Do not tell a client you are live-replying there. Unknown is a first-class state. A Watchtower tile that implies depth you do not have will be screenshotted into a QBR. The principal owns that lie even if a junior colored it.