The Enterprise AI OS

One AI workforce for the entire enterprise.

BeforeQuery is the operating system for enterprise AI. Deploy AI employees across IT, Support, Sales, HR, and Finance — powered by shared knowledge, shared skills, shared tools, and one control plane. One identity, one audit trail, one governance model.

Architecture

Five layers, one platform.

Every capability BeforeQuery ships is built on the same primitives — so a new department is a configuration, not a new platform.

  1. 05
    AI Employees, Applications & Workflows

    The workforce customers actually deploy — departmental agents, AI-native apps, and multi-step procedures.

  2. 04
    Agent Control Plane

    The registry, identity, versions, deployment, permissions, secrets, memory, and evaluation of every agent — one console.

  3. 03
    Skills, Knowledge & Tools

    Reusable capabilities, one shared context layer, and one integration catalogue that every employee draws from.

  4. 02
    Enterprise Systems

    The systems your team already runs — CRM, ITSM, HR, ERP, docs, identity, cloud, code. Connected, not replaced.

  5. 01
    Security & Governance

    The layer everything else sits on — RBAC, SSO, SCIM, agent identity, audit logs, approval policies, and compliance.

AI Workforce

An AI employee for every department.

Named, permissioned, evaluated. Each employee runs on the shared platform — one agent runtime, one identity model, one audit trail.

IT

IT Employee

  • Access requests and password resets
  • Software provisioning and licence checks
  • Onboarding and offboarding cascades
  • Device requests and IT troubleshooting
Support

Support Employee

  • Ticket classification and routing
  • Customer questions from your knowledge
  • Refund requests within policy
  • Escalation to the right specialist
Sales

Sales Employee

  • Lead qualification and enrichment
  • Account research and briefing packs
  • CRM hygiene and follow-ups
  • Proposal drafts and meeting prep
HR

HR Employee

  • Policy lookup and answers
  • Onboarding and offboarding workflows
  • Leave, time-off, and benefits requests
  • Employee documentation retrieval
Finance

Finance Employee

  • Invoice processing and reconciliation
  • Expense review under policy
  • Payment requests and routing
  • Finance questions from the handbook
Operations

Operations Employee

  • Vendor and contract questions
  • Incident response coordination
  • Approval workflows across systems
  • Operational reporting on demand
Anatomy

What's inside every AI employee.

Eleven configurable primitives. Change any one and the employee's behaviour changes — no rebuild, no redeploy.

Identity
Identity

Its own name, email, and account. Every action attributable back to it.

Identity
Instructions

The system prompt that shapes voice, scope, and refusal behaviour.

Identity
Models

The reasoning model behind it. Bring your own or use the platform default.

Knowledge
Knowledge

The corpora it reads from. Grounded, cited, and refresh-aware.

Knowledge
Memory

What it remembers across sessions — user, thread, or workspace scope.

Capability
Skills

Reusable capabilities. One skill can serve many employees.

Capability
Tools

The APIs, MCP servers, and enterprise systems it can call.

Capability
Workflows

Multi-step procedures with branching, approvals, and retries.

Boundary
Permissions

Scoped access — never more than the human it's acting for.

Boundary
Policies

Risk tiers that decide when a human must approve.

Boundary
Evaluation

Continuous accuracy, cost, latency, and hallucination scoring.

Shared primitives

Build shared. Not siloed.

The most common failure mode of enterprise AI: each team builds its own agent with its own knowledge, its own tools, its own permissions. BeforeQuery shares four primitives across every employee — so adding a new department is configuration, not a rebuild.

01

Skills

Reusable capabilities — search a wiki, open a Jira ticket, provision an Okta user. Build once, and every AI employee that needs it can call it.

02

Knowledge

One shared context layer across documents, code, tickets, and databases. Grounded retrieval, source paragraphs, and permission-aware access.

03

Tools

One integration catalogue — SaaS, identity, cloud, databases, MCP servers, custom APIs. Connect once, use everywhere.

04

Workflows

Multi-step procedures with branching, approvals, human-in-the-loop, retries, schedules, and agent-to-agent calls.

Human-in-the-loop

Policies decide when a human must approve.

Every action carries a risk tier. Reads run on autopilot; reversible writes route to the correct approver; destructive writes add step-up MFA on the reviewer before execution.

Low risk
Automatic

Read-only lookups, cited answers, non-privileged writes.

Medium risk
Manager approval

Reversible writes — CRM updates, ticket edits, calendar changes.

High risk
Security approval

Destructive or privileged writes — access grants, refunds, deletions.

The one rule

An AI employee should never have more access than the human it acts for.

HumanAI EmployeeSkillToolAction
Enterprise-grade governance

Every action attributable. Every risky action gated.

Agent Identity
email, groups, audit trail
RBAC
role-based per surface
SSO
Okta · Entra · Google
SCIM 2.0
auto-provision
Tool Permissions
scoped per employee
Data Permissions
row-, doc-, and field-level
Approval Policies
low / medium / high tiers
Guardrails
input, output, and tool-call
Audit Log
every action, immutable
Secrets
AES-256-GCM at rest
Data Residency
EU / US / private
Compliance
SOC 2 · GDPR · HIPAA-ready
Agent Control Plane

Every AI employee, one console.

Registry, identity, versions, deployment, permissions, secrets, memory, policies, logs, costs, evaluation — the operating layer for the whole workforce.

What it manages
Registry
Identity
Configuration
Versions
Deployment
Permissions
Secrets
Policies
Memory
Logs
Costs
Evaluation
Observability
last 24h
94%
Accuracy
91%
Task Success
98%
Tool Success
2%
Hallucination
$0.08
Avg Cost
4.2s
Avg Latency
trend↑ 6% wow
Four categories · one platform

What BeforeQuery covers.

AI Workforce

Named AI employees for IT, Support, Sales, HR, Finance, and Operations. Deploy the department; skip the build.

Agent Platform

Build and deploy custom agents. Visual builder plus SDK/API/CLI/MCP for the developer path.

Enterprise Context

Knowledge, data, memory, and permissions unified across the systems your team already runs.

AI Operations

Governance, observability, evaluation, and security — the operating layer under every employee.

Workflows

Multi-step work, agent-to-agent, human-in-the-loop.

An HR event triggers an onboarding cascade. The Sales employee delegates to Engineering for a technical answer. Finance auto-approves under threshold, routes over. The platform makes it one graph, not many.

  • Branching, conditions, and retries
  • Approvals, delegation, step-up MFA
  • Schedules, webhooks, agent-to-agent calls
  • Every step logged, replayable, reversible

Deploy your AI workforce. On your terms.

Book a demo and see the platform on your own systems — one employee live in 45 days, the workforce shortly after.