Automate. Augment. Assist. Human-Only. Every call yours.

“Who decided AI does this work?” You did. And you can prove it.

NOW OS™ — the Nexus of Work Operating System — maps your organization's work down to individual actions and scores each one against your own rules — your policies, your risk appetite, your values. For every action it recommends one of four modes: Automate, Augment, Assist, or Human-Only. Then your people decide — up to the board — and every decision lands in a tamper-evident record with a name, a date, and a rationale. Not because we say so. Because your people decided, and anyone can check.

See a decision record See what should stay human — and why

In beta today with a major U.S. public university system, running one governed process end-to-end: 669,064 actions mapped and scored · 74,581 recommended Human-Only — scored by the same rules as the rest · 447 policies configured as scoring rules · every decision signed by a named human.

We have no stake in your answer.

The thesis

Work-first, not tool-first.

Most work was never designed. It accreted. The conventional play starts from the agent or tool, grants it autonomy, bolts on guardrails afterward, and calls the result automation. NOW OS starts from the work: it maps the organization down to individual actions, then scores the human–AI split for each one — and your people decide it. Same conventional AI primitives — a radically different methodology.

The copilot-and-guardrails playNOW OS
Allocation assumed — automate it, or don'tAllocation scored — four modes per action, not a binary, and your people choose
Governance bolted on — guardrails after autonomyGovernance structural — NOW OS recommends, a named human decides, nothing promotes itself
Grounding RAG-stapled to a chatbotGrounding graph-native — reasoning over a connected model of the organization
One copilot doing everythingA coordinated estate of role-specialized agents

Process-mining tools map your work. Agent platforms run agents. NOW OS maps, scores, governs, and executes — one loop, one ledger.

Independent research points the same way: redesigning the division of labor is the single largest driver of value from AI — and only a small minority of organizations have actually done it. NOW OS is built for exactly that redesign.

The four modes

Not "automate or don't." Four modes for every action — recommended by NOW OS, decided by you.

NOW OS scores every action against your own rules — your policies, your risk appetite, your values — separating "can it run autonomously?" from "at what consequence?" Two equally automatable actions can land in different modes purely because of stakes. Your people make every call.

AutomateAI runs it end-to-end; humans audit out of band. It runs only after a named person approves it.
AugmentAI does the work; a named owner reviews what it produces.
AssistYour people do the work; AI helps them do it faster.
Human-OnlyHeld to the same scoring, evidence, and rules as the other three — yours. It's one of four because the complete answer includes it, not because we favor it. Your named people make the call, and the record keeps their names.

Use cases

One method. Every kind of work.

A deal, a shift schedule, a wire payment, a marketing campaign, a staffing plan. Different work, different endings — the same frame every time: each step named, each performer visible, and the moments where a named human must decide, shown as exactly that. Select any card, or just watch.

Automate — AI runs it Augment — AI executes, a named human reviews Assist — a human works, AI helps Human-Only — a person decides. Full stop.

Proof, not promise

In beta today: one governed process, end-to-end.

Alpha-Reveille — the first NOW OS release — operates a complete security-exception process for a major U.S. public university system: intake to decision to audit, on a work model mapped down to the individual action.

669,064individual actions in one process's Task Atlas
88.9%of actions recommended for autonomous execution — under rules the institution configured
74,581recommended Human-Only — scored by the same rules as the rest
447policies read and grounded — 1,856 pages
18 / ~130agent families / role-specialized agents
~12Bfacts in the full digital-twin knowledge graph

Figures are live platform values, bound to the running system — not marketing estimates. See the worked case →

See it

This is the real product, not a mockup.

Every screenshot below is the running interface, shown with the reference instance's branding — customer instances ship under your own name and identity, or NOW OS's. The same product ships as an always-on evaluation instance — every surface clickable, both roles, zero integration risk.

The approver's dashboard

Open exceptions, decisions waiting, cycle time, autonomy rate, and labor saved — with every request analyzed and arriving with a recommendation, not a blank form.

Reveille — Overview
The Reveille overview dashboard

A decision, fully assembled

NOW OS runs intake, classification, and risk assessment — dozens of individual actions — then stops at the one human gate that matters. The approver sees the recommendation, the residual risk, the cost math, and the step-by-step record of exactly what AI did and where a person decided.

See the full worked case →

Reveille — A worked request
A worked security-exception request with recommendation and step record

The product family

One agnostic core. Many products.

The platform knows nothing about any one customer or industry. A vertical is an overlay; a customer is a mapping — never the foundation. That's what makes the platform reusable across industries.

The four-layer product family: platform, vertical, digital twin, module
1
NOW OSThe platform

The organization-agnostic platform core: 10-level work decomposition, configurable scoring tuned to your own rules, the governance envelope, the agent estate, the governed data lane, the verified ledger.

2
Verticale.g. Higher-Ed

Scopes the operational domains and regulatory frameworks for an industry — FERPA and Title IX for higher education; HIPAA and Joint Commission for healthcare. A new industry is a new overlay, not a new platform.

3
Digital TwinThe whole customer

One customer's entire work, decomposed to atomic actions across every operational domain, connected to their people, policies, and systems in a reasoning knowledge graph.

4
Modulee.g. Reveille

One process, productized: narrow, polished, go-to-market. Reveille — the Exception-Process module — is the first, in beta today. It borrows the twin's knowledge read-only, with full provenance.

Governance

NOW OS recommends. Your people decide — on the record.

Governance in NOW OS is structural, not optional — it cannot be switched off, and it cannot be satisfied by a disclaimer.

Nothing auto-promotes

Everything the agents discover or recommend is filed as a candidate. A named human on the customer's Strategic Advisory Board approves, denies, or defers — every time.

A verified ledger

Every governance decision and execution event lands in a tamper-evident, hash-chained audit log — and the chain is re-verified on every read, not stamped once and trusted forever.

Enforced floors

Restricted work — FERPA, HIPAA, PCI, export-control — carries enforced control floors, so it can never be silently over-automated. Non-compliant work is refused at ingest, fail-closed.

Services

Delivered as a partnership, not a license key.

Four service lines wrap the platform — from decomposing your work, to standing up your twin, to running it with you and training your people.

01

Consulting

Atomic-task decomposition, AAA analysis, workforce-transformation strategy.

02

Implementation

Standing up a customer Digital Twin: data integration, agent wiring, real-system connections.

03

Managed Service

Running the platform as an ongoing operation, with governance support.

04

Training & Enablement

NOW OS University: curriculum, exams, and stackable certifications for your people.

How delivery works Request a demo