The Automated Sales Engineer
How we eliminated the admin bleed on sales engineering — and built the pipeline that proves it works.
Most AI companies talk about what their technology could do. We built ours into the front door. Every prospect who visits PropelAI experiences the same agentic pipeline we deploy for clients — generating a bespoke AI Opportunity Brief tailored to their business, automatically.
The Pipeline
Six Steps, Zero Humans in the Loop
From prospect submission to personalized deliverable — fully autonomous.
Intake
Prospect submits their company URL, primary challenge, and optional ROI fields via the /generate form. No sales call required.
Planning Agent
An AI agent analyzes the submission and creates a structured research plan — decomposing the prospect into industry context, competitive landscape, and specific automation opportunities.
Deep Research
Autonomous web research using Ollama models (20B parameter) running on-premise to build comprehensive company intelligence. No data leaves the building.
Synthesis
Research findings are synthesized into a tailored narrative with specific recommendations, ROI projections, and implementation pathways unique to the prospect.
Autonomous Coding
Claude Code CLI generates a complete, branded HTML microsite from scratch — not a template fill-in, but bespoke code written for each prospect.
Delivery
The prospect receives an email with a link to their personalized AI Opportunity Brief microsite, hosted on PropelAI infrastructure.
Architecture
Built Different, On Purpose
Three architectural decisions that separate this pipeline from chatbot wrappers.
Air-Gapped Security
All research and synthesis runs on NVIDIA DGX Spark hardware on-premise. Zero data leaves the building. CMMC-ready by default.
Agentic Planning
Not prompt-and-pray. A planning agent decomposes each prospect into a structured research task tree, ensuring comprehensive coverage.
Premium Execution
Claude Code CLI generates bespoke HTML from scratch for each prospect — no templates, no fill-in-the-blanks. Every brief is unique.
Translation Matrix
Same Platform, Different Value
How the same technical capabilities translate into vertical-specific outcomes.
| Technical Feature | GovCon Value | Service Ops Value |
|---|---|---|
| Air-gapped DGX Spark inference | CMMC 2.0 Level 2 compliance without cloud risk | Customer data stays on-site — no cloud liability for service records |
| Agentic planning with task decomposition | Structured capture management workflows | Automated dispatch sequencing and job routing |
| Autonomous code generation (Claude Code CLI) | Bespoke proposal sections, no template constraints | Custom reporting dashboards and workflow integrations per client |
| On-premise model hosting (Ollama) | FedRAMP-irrelevant — data never leaves your facility | No per-seat SaaS fees — runs on your hardware at fixed cost |
| Multi-model orchestration | Best model per task: research, analysis, generation | Right-sized models: fast triage for calls, deep analysis for estimates |
| Real-time pipeline monitoring | Full audit trail for DCAA/DCMA oversight | Live visibility into every automated workflow and agent decision |
Air-gapped DGX Spark inference
CMMC 2.0 Level 2 compliance without cloud risk
Customer data stays on-site — no cloud liability for service records
Agentic planning with task decomposition
Structured capture management workflows
Automated dispatch sequencing and job routing
Autonomous code generation (Claude Code CLI)
Bespoke proposal sections, no template constraints
Custom reporting dashboards and workflow integrations per client
On-premise model hosting (Ollama)
FedRAMP-irrelevant — data never leaves your facility
No per-seat SaaS fees — runs on your hardware at fixed cost
Multi-model orchestration
Best model per task: research, analysis, generation
Right-sized models: fast triage for calls, deep analysis for estimates
Real-time pipeline monitoring
Full audit trail for DCAA/DCMA oversight
Live visibility into every automated workflow and agent decision
The Difference
Generative vs. Agentic AI
Most companies are still stuck on the left side. Here's why the right side matters.
Generative AI
What most people are doing
- Prompt → Response
- Generic outputs
- Hallucination risk
- Requires constant supervision
- Template-based
- Good for first drafts
Agentic AI
What PropelAI builds
- Plan → Research → Synthesize → Execute
- Bespoke outputs
- Grounded in research
- Autonomous operation
- Code-generated
- Production-ready deliverables
See It In Action
Get your own AI Opportunity Brief — built by the same pipeline described above.