Agent workforce
Four functions, covered
Sales and outbound, customer support, marketing and content, back-office operations. The agents run inside the application newc0 builds you, and everything binding, irreversible or above your threshold stops at you.
Last updated
Sales and outbound
prospecting · follow-up
Models feeding it: sales-forecasting, lead-scoring
What the agents run
- Finds and enriches leads
- Scores and ranks them
- Writes and sends opted-in outbound
- Books and reschedules calls
Never without you
Any discount above your threshold. Any pricing statement. Any binding commitment.
Customer support
tickets · escalation
No trained model confirmed in this domain. Language agents only.
What the agents run
- Answers tickets from your own documentation
- Triages and escalates what it cannot answer
- Keeps the backlog at zero overnight
Never without you
Refunds, credits and anything outside stated policy. Anything a customer could hold you to.
Marketing and content
drafts · publishing
No trained model confirmed in this domain. Language agents only.
What the agents run
- Drafts articles, emails and landing copy
- Keeps a publishing schedule
- Reports on what was read and what converted
Never without you
Anything published in your name. Positioning. Claims about the product.
Back-office operations
invoices · records
No trained model confirmed in this domain. Language agents only.
What the agents run
- Reconciles invoices against the bank feed
- Chases overdue payments
- Merges duplicate records and keeps data clean
Never without you
Any payment out. Any spend above your threshold. Anything irreversible.
How does the approval model work?
Agents execute continuously and queue anything that crosses a line. The lines are fixed for some things — pricing, positioning, anything legally binding, anything irreversible — and set by you for spend, where you choose the threshold. Queued items wait. They do not time out into approval.
In practice you meet the queue once a morning, in a report that shows what was done overnight, what needs you, and what the models predict. Realistically that is about fifteen minutes — a typical morning, not a guarantee.
Why does newc0 refuse to call these employees?
Because they are not, and the precision is load-bearing. An agent runs a function. It does not exercise judgment, hold accountability, or notice that something has gone strange in a way nobody wrote down. Describing software as staff sets an expectation that the software will quietly fail to meet.
This is also why newc0 is for founders who have not hired yet. The functions get covered. The org chart never gets created.
Questions about the agents
What can AI agents genuinely do in sales today?
- They can find and enrich leads, rank them, write and send outbound to people who opted in, book calls and chase replies. They cannot judge whether a deal is worth doing, hold a negotiation, or decide a price. The first list is most of the volume; the second list is most of the value.
Do the agents have names?
- No. They are functions, not people, and newc0 does not dress them up as hires, teammates or employees. A named agent invites you to trust it the way you trust a person, which is exactly the wrong instinct to build in someone running a company on software.
What should you never let an agent decide?
- Anything you cannot take back, anything that binds you legally, what you charge, and what the company claims to be. Those four cover most real damage. newc0 queues all of them for you by default, along with any spend above a threshold you set yourself.
What happens when an agent does not know the answer?
- It escalates rather than guesses. In support that means the ticket lands in your queue with the context attached. An agent that improvises an answer to a customer question is not saving you work, it is creating a problem you will find out about later.