Enterprise Box AI
Institutional Intelligence

Every ticket resolved, every call handled, every decision made adds to your collective intelligence. A synthesis layer that reasons across all of it in real time — and never forgets.

Knowledge Graph|Graph-Guided RAG|Compounding Intelligence

Deployed on your infrastructure. Gets smarter every day. Never retires.

Scroll

Deployed across enterprise organizations

DTIQPacketFabricElement8QwiltWelinkDTIQPacketFabricElement8QwiltWelink

Beyond Search

What becomes possible with
institutional intelligence

A knowledge base answers “what do we have about X?” The intelligence layer answers “why does this keep happening, what worked last time, and what should we do next?”

Compounding Returns

Every resolved ticket, every call, every decision enriches the model. Year 3 is dramatically smarter than year 1 — and it never forgets, never retires, never takes a day off.

10x
Democratized Expertise

A new hire with the intelligence layer performs like a 10-year veteran. Instant access to the synthesized knowledge of everyone who came before.

Cross-Domain Synthesis

Connects dots across tickets, calls, contracts, and changes that no single person could. “The last 4 P1s on this circuit happened within 48 hours of a change request.”

Institutional Intelligence

Every data source feeds one brain

Phone calls, emails, tickets, KB articles, uploads, and portal chats — all flowing into a single intelligence core that compounds daily.

KB Articles
File Uploads
Transcripts
Portal Chats
Resolutions
Support Tickets
Email Threads
Phone Calls
Your
Intelligence
LIVETicket #4521 resolved → ingested to knowledge base

Graph-Guided Retrieval

Watch a query flow through three layers of intelligence

Instead of searching everything, the knowledge graph narrows the search space by 2,000x before vector similarity ever runs.

intelligence-layer
>|
0x
Search Reduction
<0s
Synthesis Response
0M+
Tokens Per Company
0
Intelligence Layers
0+
MCP Tools
0
AI Models

The Precision Advantage

Flat search vs. graph-guided retrieval

Every other RAG system searches everything. Ours narrows the space by 2,000x before vector similarity ever runs.

BEFORE — Flat RAG
1Embed query → vector
2Search ALL 100M tokens
3Return top-K “similar” chunks
Mediocre

Returns chunks about “compliance” from everywhere. Missing the causal chain. No context for why.

vs
AFTER — Graph-Guided
1LLM extracts entities from query
2Graph traversal → linked cluster
3Vector search scoped to ~4,200 tokens
Precise

Returns the exact policy, its origin incident, the person who created it, and the full causal chain.

2,000x
Search space reduction
<3s
Full synthesis response
100%
Source attribution

Knowledge Capture

Every interaction makes your organization smarter

Passive extraction from existing data streams combined with active knowledge elicitation from subject matter experts.

Passive Capture

Auto-extracts entities from every ticket, call, and document on ingest. Zero workflow change for agents.

🎤
SME Interviews

AI agent periodically interviews experts. Structured questions about processes, decisions, and edge cases.

📝
Decision Logging

“You just escalated this to Tier 3 — why?” Captures the reasoning that would otherwise be lost.

Knowledge Ingestion — Live
📞Call transcript ingested → 3 new knowledge chunks, 2 entities linked
🎫Ticket #4521 resolved → linked to "Billing Process" node
📝Decision logged: "Escalated to Tier 3 — client has custom SLA"
⚠️Gap detected: "VPN Setup Process" has 0 linked chunks
🎤SME Interview complete: Sarah M. → 12 procedural nodes captured

Knowledge Types

Three kinds of institutional knowledge

Not all knowledge is created equal. The system captures three distinct types, each requiring different extraction strategies and carrying different value.

Easiest to Capture

Declarative Knowledge

What is true. Policies, SOPs, configurations, rules, org charts, system specs. The “reference material” of your company.

Password rotation policy
Escalation matrix
System architecture docs
Compliance requirements
Capture method:Document ingestion, KB articles, file uploads
Medium Difficulty

Procedural Knowledge

How things actually work — which often differs from what's documented. The “real” process vs. the “official” process.

How to actually debug the billing system
The real onboarding sequence
Workarounds for known issues
Undocumented API quirks
Capture method:Workflow observation, SME interviews, ticket analysis
Hardest · Most Valuable

Causal Knowledge

Why things are this way. Historical context, failed experiments, hard-won lessons. This is the tribal knowledge gold.

Why we switched from vendor A to B
The incident that changed the deploy process
Why client X gets special handling
The failed approach we tried in 2022
Capture method:Decision logging, SME interviews, incident post-mortems

Institutional Intelligence

Every interaction makes your company smarter

A closed-loop intelligence system where every customer touchpoint compounds your advantage.

Phone Call
Transcript
RAG Ingest
Knowledge
AI Response
Resolution
compounds

Knowledge Compounds

Phone calls become searchable transcripts. Resolved tickets become training data. KB articles become instant answers. Your institutional knowledge grows every day — automatically.

Closed-Loop Intelligence

AI classifies intent → routes to the right team → surfaces relevant knowledge → learns from the resolution. Every loop makes the system smarter.

Your Data Moat

After six months, your system knows things no generic SaaS ever will. Your product names, your escalation patterns, your customers' preferences. Competitors can't buy this.

What's in Your Box

Six services. One platform.

A complete AI ecosystem deployed on your infrastructure. Purpose-built, deeply integrated, entirely yours.

Intelligent CRM

AI-powered CRM with smart ticket routing, SLA enforcement, email threading, ITIL change management, and NOC dashboard. Every interaction feeds the intelligence loop.

303 Endpoints55 Models47 Controllers
$ entbox deploy crm --company acme
Provisioning database... done
Running 55 migrations... done
✓ Live at crm.acme.entbox.ai

CRM Frontend

Next.js 14 dashboard with Cmd+K agentic search, AI assistant panel, NOC monitor, change calendar, and customer self-service portal.

81 Routes94 Components49+ Pages

AI Phone Agent

Voice AI handling inbound calls in real-time. Every call transcript auto-ingests into your RAG knowledge base.

<1s Latency40+ Metrics

RAG Knowledge Engine

pgvector retrieval with tier-boosted search. Auto-ingests resolved tickets, call transcripts, KB articles, and file uploads.

3-Tier Boostpgvector

MCP Tool Server

70+ agentic tools across 10 categories. AI orchestration for autonomous actions.

70+ Tools10 Categories

Real-time Analytics

Live dashboards for SLA tracking, agent performance, call analytics, ticket throughput.

PrometheusCustom Dashboards

DevOps & IaC

Terraform with 8 modules, 6 CI/CD pipelines, auto-rollback, GCP Cloud Run orchestration.

8 Terraform Modules6 Pipelines

AI Everywhere

Intelligence in every layer

Six distinct AI capabilities woven into every layer of the platform.

Agentic

Cmd+K Agentic Search

Natural language search across your CRM. AI returns structured answers with sources.

AI Assistant Panel

Context-aware assistant that sees your ticket, suggests responses, summarizes history.

Multi-tool

AI Workspace

Full-screen AI with file upload, code execution, and multi-tool orchestration.

6 Channels

NL Engine (RAG)

6-channel search: KB, tickets, docs, call transcripts, file uploads, web. Tier-boosted for relevance.

Portal AI Chat

Customer portal with AI chat, ticket submission, circuit viewing, change schedules, and KB browsing.

AI Insights Dashboard

Revenue opportunities, sentiment analysis, training needs, recurring issue detection, SLA risk prediction.

The Process

From zero to AI-powered

A proven deployment playbook. Weeks, not months.

01

Diagnose

We map your operations, identify high-impact AI opportunities, and design your custom architecture.

02

Deploy

Your complete stack goes live — CRM, phone agent, RAG engine, analytics — configured for your business.

03

Own

Full source code. Full data control. Terraform IaC, CI/CD pipelines, and the keys to the kingdom.

Your Stack

Your Complete AI Solution

Everything for AI success, packaged in a solution you completely own and control.

AI Models

Claude, GPT-4, Deepgram — trained on your data

Data Pipeline

pgvector, embeddings, tier-boosted RAG retrieval

Business Logic

47 controllers, SLA enforcement, smart routing

Your Infrastructure

GCP Cloud Run, Cloud SQL, Terraform IaC

Complete Ownership

Source code, databases, secrets, CI/CD pipelines — everything. Deploy anywhere, customize freely.

Head to Head

Box AI vs. the incumbents

Feature-by-feature comparison against the enterprise tools Box AI replaces.

Capability
ServiceNow
Zendesk
Salesforce
Box AI
AI Foundation
Bolt-on Einstein
Older ML models
Einstein (limited)
✓ Claude-native in core
Vector Knowledge (RAG)
Manual API
✗ No native RAG
Einstein Search
✓ pgvector + tier boost
Voice AI
Phone only
Recording only
Limited
✓ Twilio + Deepgram + Claude
Intent Classification
Rules-based
ML-based
ML-based
✓ Keyword + LLM dual-path
Data Isolation
Shared multi-tenant
Shared multi-tenant
Shared multi-tenant
✓ Per-company (own DB)
Deployment
Vendor-hosted
Vendor-hosted
Vendor-hosted
✓ Your cloud, your code
Time to Deploy
Weeks (pro services)
Days
Weeks
✓ Days (CI/CD)
Customization
Config + scripting
Config + marketplace
Apex + config
✓ Modify source directly
Vendor Lock-in
✗ Full lock-in
✗ Full lock-in
✗ Full lock-in
✓ Zero — you own the code
Cost / Agent / Year
$1,800–2,880
$1,440–2,400
$1,200–2,400
~$120
Traditional SaaS
$1,800–2,880
per agent / year
Box AI
~$120
per agent / year
15–25x cheaper

Enterprise Security

Built for compliance

Enterprise-grade security at every layer.

Per-Company Auth

Firebase Auth per company. Separate domains, separate user pools, complete isolation.

RBAC & API Keys

Role-based access with timing-safe API key validation. GCP Secret Manager for service auth.

PII Masking

Automatic PII detection and masking in logs. Phone, email, SSN, credit cards sanitized.

Container Hardening

Non-root containers, read-only filesystems, VPC networking, Trivy scanning in CI.

Why Box AI

Three principles. Zero compromise.

Enterprise support teams pay $100K+/year across 5–7 separate SaaS vendors, own none of the IP, and bolt on “AI features” as expensive add-ons.

Replace, Don't Rent

Stop paying per-seat fees to 5+ vendors. Box AI replaces your CRM, help desk, knowledge base, phone system, customer portal, and analytics — with one platform you own outright. Deploy on your cloud. Run it forever.

Zendesk $50–99/seatServiceNow $150+/seatSalesforce $100+/seat

AI-Native, Not Bolt-On

Claude intelligence is woven into every layer — not a $50/mo add-on. RAG-powered knowledge retrieval, intent classification, natural language engine, agentic search, and real-time voice AI are all built into the core. 6 AI models integrated. Zero extra licensing.

Claude Sonnet & HaikuOpenAI EmbeddingsDeepgram ASR/TTS

Per-Company Isolation

Each deployment is fully isolated — own database, own auth project, own infrastructure. This is not shared multi-tenant SaaS. Zero cross-company data leakage by architecture. Full compliance and data residency guarantees.

Own Cloud SQL DBOwn Firebase AuthOwn Cloud Run

Get Started

Ready to own your intelligence?

Join the companies building institutional intelligence they own forever. The semantic layer, intelligent automation, and SaaS replacement — in one platform.