04 - Demo Script

Demo Objective

Show a believable end-to-end enterprise workflow that combines AI agents, policy governance, and human oversight to deliver measurable value.

Audience

Technical recruiter
Hiring manager (Solutions Engineering / Sales Engineering)
Cross-functional panel (product + GTM)

Demo Length

20 to 25 minutes

Storyline

"A mid-market SaaS company is reviewing a growing volume of vendor MSAs and procurement agreements. We will show how a LangGraph Studio workflow accelerates standard contracts, escalates risky language to a reviewer, and leaves behind a clear audit trail."

Live Flow

Intake (3 minutes)
- Run two local fixture contracts by filesystem path in LangGraph Studio:
  - Contract A: standard, low-risk.
  - Contract B: non-standard liability language, high-risk.
- Show run metadata and graph kickoff.
Agent Extraction (4 minutes)
- Show structured clause extraction and evidence spans.
- Show confidence scores and normalized output schema.
Policy + Routing (4 minutes)
- Contract A routes to auto-path due to policy pass.
- Contract B routes to human review due to high-risk clause.
Human Review (5 minutes)
- Reviewer uses the LangGraph interrupt to edit risk tags or approve/reject Contract B.
- If helpful, show that the interrupted review can be reloaded through GET /api/runs/{thread_id} after a backend restart.
- Reviewer approves final action.
Audit + Reporting (4 minutes)
- Node-level event history with provider/model metadata, policy reasons, and reviewer action.
- SQLite / JSONL / Markdown run artifacts.
KPI Snapshot (3 minutes)
- Baseline vs pilot metrics dashboard.
- Call out agent-hours of reviewer capacity unlocked as the headline business outcome.
- Discuss rollout readiness criteria and what would be added next.

Example Inputs

Use the repo fixtures with absolute paths in Studio:

{
  "contract_path": "/absolute/path/to/agentic-contract-review/fixtures/low_risk_vendor_msa.txt",
  "provider": "openai",
  "policy_pack": "/absolute/path/to/agentic-contract-review/policies/default_policy.yaml",
  "run_label": "demo-low-risk"
}

{
  "contract_path": "/absolute/path/to/agentic-contract-review/fixtures/high_risk_vendor_msa.txt",
  "provider": "openai",
  "policy_pack": "/absolute/path/to/agentic-contract-review/policies/default_policy.yaml",
  "run_label": "demo-high-risk"
}

When the high-risk contract pauses in human_review, resume with a payload like:

{
  "decision": "edit",
  "edited_extractions": [
    {
      "clause_id": "clause-002",
      "summary": "Reviewer clarified the indemnity obligation."
    }
  ],
  "edited_risks": [
    {
      "clause_id": "clause-002",
      "risk_level": "high",
      "confidence": 0.98,
      "reviewer_reason": "Broad indemnity remained unacceptable."
    }
  ],
  "reviewer_notes": "Escalated risk remained valid.",
  "reviewer_id": "demo-reviewer"
}

Screens or Artifacts to Prepare

LangGraph Studio graph view.
Processing timeline view per contract.
Clause extraction JSON (cleanly formatted).
Human review interrupt payload and resume action.
Reviewer inbox or pending-review API response.
Audit trail table.
KPI scorecard.
One generated Markdown report from runtime/reports/.

Q&A Prompts You Should Be Ready For

"How do you prevent hallucinated clauses from triggering actions?"
"What changed when confidence was wrong but high?"
"How quickly can this adapt to a new customer policy?"
"What does rollback look like if a model update regresses quality?"
"Why LangGraph Studio first instead of a customer-facing app?"
"How do you keep the demo running if an API key is missing or a provider call fails?"

Strong Closing

"This project is intentionally built as a deployable customer workflow: measurable outcomes, clear governance controls, a working LangGraph Studio execution path, and a practical roadmap from demo-quality orchestration to customer-ready deployment."

FilesExpand file tree

04-demo-script.md

Latest commit

History