Cadre: Digital Employees You Can Audit
Cadre runs AI as role-based digital employees, and every job is source-checked, scored, approved and costed before it reaches a client. Here is how.
By NGMOB, published . 4 min read.
Key points
- Cadre, from NGMOB, runs AI as role-based digital employees: a product manager, an industry researcher, a content operator, a designer and a support agent.
- Before work is handed over, Cadre checks that citations link to real sources and scores the output automatically.
- Nothing reaches a client until it passes an approval gate and a delivery page where the client signs off or asks for changes.
- Every job records its cost per model, with an automatic cut-off, in a tamper-evident hash chain.
- Cadre runs on Claude, GPT and Qwen and offers an OpenAI-compatible API; companies apply for a free trial.
What is a digital employee?
A digital employee is an AI system given a job title rather than a chat box. In Cadre, NGMOB's enterprise platform at highcadre.com, each digital employee has a role, a workflow and a set of rules, and its work arrives with the evidence a manager needs before anything goes to a client: where its facts came from, how good it is, what it cost and who approved it.
That last part is the point. Most teams can already get AI to produce something. What they lack is a way to trust the output, review it before it goes out, and know what it cost. Cadre treats those three as the foundation rather than add-ons.
Five roles, briefed like colleagues
Cadre starts with five ready-made roles. You brief each one the way you would brief a colleague: goal, audience, format and constraints. If a role does not fit your business closely enough, clone it and tune the copy.
| Role | Typical brief | What comes back |
|---|---|---|
| Product manager | A feature request | A PRD, a flow chart and a prototype |
| Industry researcher | A market or competitor question | A sourced report or weekly briefing from the web |
| Content operator | A topic | A draft with images, ready to deliver |
| Designer | A campaign or post | Posters, illustrations and visuals |
| Support agent | A customer question | An answer from your knowledge base, or a ticket |
What every job records
Every job in Cadre leaves a record. The example on highcadre.com is an 800-character competitor weekly written in Chinese:
| Field | Example value |
|---|---|
| Brief received | Competitor weekly, Chinese, 800 characters |
| Citations verified | 12 of 12 have a source |
| Evaluation score | 92 out of 100, scored automatically |
| Cost | US$0.184, cost cut-off not triggered |
| Approval | Waiting for a named reviewer |
| Hash chain | A tamper-evident fingerprint of the record |
Behind those fields sit the platform's checks:
- Citation checks. A claim must link to a source before the work is handed over.
- Machine scoring. Output is scored automatically, so quality is measured rather than assumed.
- Model comparison. The same brief can run on several models, and Cadre keeps the cheapest result that passes.
- Approval and delivery. Work waits at an approval gate, then goes to a delivery page where the client confirms it or asks for changes. A change request opens a revision ticket automatically.
- Cost control. Token and dollar costs are visible per job and per model, and a job stops when it hits its limit.
- Audit trail. Records are chained with hashes, so later edits show. Tenants, permissions and per-role quotas keep teams apart.
Three model families, one workflow
Cadre runs on Anthropic Claude, OpenAI GPT and Alibaba Qwen. A role can be pinned to one of them, or jobs can be routed by type. Cadre's own guidance pairs Claude with long-form writing and research, GPT with general reasoning, prototypes and visuals, and Qwen with Chinese-language fluency, in-country requirements and high-volume work at a low unit cost. Switching models does not change the workflow, and each job's cost is split by model.
Getting started in four steps
- Pick a role from the ready-made digital employees, or clone one and tune it.
- Write the brief: goal, audience, format and constraints.
- Review the output at the approval gate or on the delivery page; feedback becomes a revision ticket.
- Keep what works as a reusable skill, or set it to run on a schedule.
Who it is for
Today Cadre fits small and mid-sized teams whose work is heavy on content, marketing, product or research: brand and cross-border e-commerce content teams, SaaS growth teams, and consulting or research studios. Private deployment for regulated large enterprises is the next step Cadre is building towards. Companies apply for a free trial at highcadre.com/apply, and colleagues join through the employee sign-in. The interface is in Chinese, and developers can call Cadre through an OpenAI-compatible API.
Where it fits at NGMOB
Cadre is the business-side counterpart of agents-md, our open-source rules for AI coding agents, which we wrote about in Eleven Rules for Letting AI Agents Commit Code. Both start from the same position: AI can move fast as long as the system around it stays accountable.
Questions
What is Cadre?
Cadre is NGMOB’s enterprise digital-employee platform at highcadre.com. It runs AI as role-based digital employees (product manager, researcher, content operator, designer and support agent) whose work is source-checked, machine-scored, approved by a person and costed per job, with a tamper-evident audit trail.
How is a digital employee different from a chatbot?
A chatbot answers in a conversation and leaves the checking to you. A digital employee in Cadre has a role and a workflow, proves its work with citation checks and scoring before hand-over, waits for approval, and records what each job cost and who signed it off.
Which AI models does Cadre use?
Anthropic Claude, OpenAI GPT and Alibaba Qwen. A role can be pinned to one model, or the same brief can run on several and Cadre keeps the cheapest result that passes its scoring.
How much does Cadre cost?
Companies apply at highcadre.com/apply for a free trial. Inside the platform, each job’s token and dollar cost is shown per model, and jobs stop automatically when they hit their limit.
Can developers use Cadre through an API?
Yes. Cadre offers an OpenAI-compatible /v1/chat/completions API. Point an existing OpenAI SDK at the platform with a Cadre key, and usage counts toward the company’s quota.
Filed under Cadre, AI agents, Enterprise AI, Governance.