Neivan Intelligence
Neivan gives a small business AI agents that do real work — chasing invoices, answering support email, tidying the books — and stop to ask you before anything that matters. Start with one pack, or run the whole platform.
Endless automations.
Built in minutes.
If your team does it on a computer, Neivan can help automate it — no engineers, no long project. Built for owners and operators of businesses with roughly 10 to 200 people.
Automate anything you can describe.
Describe a task in plain words and Neivan builds the agents and workflows to do it — then change any step, rule or schedule. Everything on this page, for your whole team. Free beta for selected companies from early November 2026; you pay only for the AI your agents use.
Ask to join the betaLet Neivan find what to automate.
With your team’s consent, Observation watches the work you allow, measures the jobs that repeat and drafts the automations for them. Try it as a fixed-price Assessment, from $1,500.
How Observation worksJust one job? Start smaller.
Monthly agents that each do one thing well — draft support replies, tidy the books, send a weekly brief. Connect your accounts, answer a few questions, approve the cards.
See the agentsChat tools answer questions.
Someone still does the work.
The busywork piles up
Chasing invoices, sorting email, categorizing transactions, answering the same support question again. Small teams lose hours a week to it.
AI that can't touch anything
A chat window can draft a reply, but you still copy it, paste it, look up the invoice and hit send. The work hasn't moved.
Or AI you can't trust
Letting software send email or change your books on its own is a real risk. You need to see what it will do before it does it.
Neivan agents connect to your real accounts, do the job on a schedule, and stop for your approval before anything that matters.
One
agent, start to finish
Built for small teams
Describe it. Try it in shadow.
Approve what matters.
You describe the job in plain words. Neivan builds the agent, runs it beside you in shadow mode until it matches how you work, and puts anything risky on an approval card first. Every run is logged and priced, so you can see what it saved.
What you get
today.
Your numbers on one screen. Problems flagged for you.
The home screen shows the figures you care about, read live from your connected apps — overdue invoices, cash, Power BI reports and refreshes.
- A weekly digest email every Monday morning
- Alerts posted to the Slack channels an admin picks (off until switched on)
- Hand a finding to an agent — it comes back as an approval card, not an action
- Suggested widgets for things nobody is tracking yet
Describe the job. Get an agent.
Start from one of 35 templates or describe what you need in plain words. A new agent can run in shadow mode — doing the work beside you without acting — until you trust it.
- Pick what each agent may touch: which apps, which folders, read or write
- Test runs hold every outgoing message back so you can check it
- Each risky action stops on an approval card for a person to decide
- Every run is logged with what it did and what it cost
Not everything needs AI.
Steps that only move data or send a known message run as plain steps, with no AI and no AI cost. Use AI only for the parts that need judgment.
- The Automations tab lists every workflow that runs without AI
- Mix plain steps and agent steps in one workflow
- Run on a schedule, a webhook, an event or by hand
Ask what your expert knows.
A Brain holds what a colleague knows — their documents and recorded expertise — so others can ask it questions when that person is busy. A Brain answers questions. It never takes actions.
- Each person keeps their own Brain in Settings
- Who may ask is controlled by permission
Show it once. Keep the steps.
Record yourself doing a task — screen, video, voice or written steps. Teach turns it into steps an agent can follow, tries them as a dry run, and can save them as a draft SOP for your handbook.
- A shared SOP library for the whole team
- Dry run before an agent uses what it was taught
A limit on what AI can spend.
Every AI call is metered in dollars. Set a monthly limit for the company, a department or a single agent. You get an alert in the app before the limit; at the limit, new runs are refused until someone raises it.
- See cost per run and per agent
- Choose a cheaper model for simple work and a stronger one where it matters
25 apps today.
Said plainly.
Our catalogue lists 214 apps, but most aren't ready yet. These are the ones that work in production now.
Ready-made actions, each risky one approved first.
The agent writes each request; every call waits for approval unless you set it to run unattended.
You can connect these today; agent actions on them are not proven yet.
That is 25 of the 214 apps in our catalogue. The rest are listed, with their real status, on the integrations list.
The controls IT will ask about —
and where each stands.
What is live, what is built but switched off, and what we don't do yet. More detail on the security page.
Demo: the business, people and numbers are made-up examples.
Single sign-on
People sign in with email and password plus a two-step code. SAML single sign-on through your identity provider is available on request.
SCIM provisioning
Live: your identity provider can create, update and deactivate users. A deactivated user is suspended and removed from workspaces. Group sync isn't built yet.
Roles and per-person permissions
Four roles — owner, admin, manager, employee — and about fifty permissions, each overridable for one person. Access can also be limited by department or team.
Approvals
Agent actions can require two or more named approvers before they run. Requests people submit, like expenses or time off, follow multi-step chains with named approvers and deadlines.
Audit log export
Every meaningful action is logged with who, when and the request id. Admins export it as CSV; each agent also has an evidence pack export.
Bring your own AI keys
Built: use your own AI provider accounts, with keys sealed in an encrypted vault. Switched on per customer — ask us.
Tenant isolation
Each company's data is kept apart in the database itself, with row-level security on every table that holds customer data.
US hosting
The app runs on DigitalOcean in New York; the database on Supabase in AWS us-east-1. No EU hosting or dedicated deployment today.
Templates to start.
A platform without limits.
The platform has 35 agent templates to start from. Use one as it is, change it, or describe a new agent in plain words. You choose its instructions, the apps it may use, the model it runs on, and what needs your approval.
Connect the apps it needs and try it in shadow mode first.
Every prompt, tool, data source, policy, and approval chain is yours to edit.
No template fits? Describe the agent you need in plain words.
Fifteen of the 35 templates.
These are starting points — not the menu. Every card below can be cloned, edited, composed with others, or used as the seed for something new entirely.
Morning Brief
One short daily brief: overnight email, today's calendar, files, and agent activity.
HR Assistant
PTO lookups, cited policy answers, and onboarding checklists; sensitive issues escalate.
IT Helpdesk
Level-1 support — password resets, install requests, diagnostics — with security escalation.
Legal Intake
Intakes contract reviews and NDAs, runs conflict checks, and routes to the right attorney.
AP Clerk
Matches invoices to purchase orders, flags variances, and routes for approval.
Sales Lead Qualifier
Enriches and scores inbound leads against your ICP, logs to CRM, drafts the opener.
Customer Support Agent
Drafts ticket answers cited from your knowledge base; exceptions escalate with context.
Content Marketer
Turns a brief into on-brand posts and emails — always a draft, never auto-published.
Recruiting Coordinator
Screens applicants against the role rubric, schedules panels, and keeps candidates warm.
Data Analyst
Answers business questions from your connected data with charts and a short narrative.
Executive Assistant
Inbox triage by priority, replies in your voice, calendar defense, and meeting prep.
Procurement Agent
Intakes purchase requests, compares vendor quotes, and drafts POs for approval.
Month-End Close
Works the close checklist daily and marks items done only when the evidence exists.
Reconciliation Prep
Assembles the reconciliation and lists exceptions with sources — a person concludes.
Build any agent you can imagine.
Templates are starting points. Build your own agent with your instructions, the apps you choose, your approval rules and your choice of model.
Customer stories
We don't publish testimonials we haven't earned. As our first customers go live, their stories — with real names and real numbers — will appear here.
Mobile app
There is no Neivan mobile app yet. It is on our roadmap; we'll say so here when it ships.