Verification, limitations & trust
What has actually been built, what remains a simulated workflow, and what must be integrated and tested in a customer's environment.
No published customer testimonials, prior paid deployment claims, unverified logos, or invented success metrics are used.
Internal AWS intake: bounded QA evidence
DEEPLINK's own inquiry flow uses API Gateway, AWS Lambda, DynamoDB, scheduled processing and SES. Internal testing confirmed a saved inquiry, one accepted SES send, a SENT record and email receipt after a configuration/permissions fix. A previous failed request remained available for review.
Limitations: this is not an external customer deployment. Alert-delivery fault drills, SLA, mail inbox placement for every message, production penetration testing and business-continuity validation have not been represented as completed.
View project context ↗Technical evidence matrix
| Module | Demonstrated | Not demonstrated |
|---|---|---|
| Reporting | Browser CSV parsing, validation, KPIs, exports | Production connectors, refresh scheduler, hosted analytics |
| Lead intelligence | Rule scoring, reasons, missing fields, duplicate block, local approval | Claude API, CRM authentication, real messages |
| RevOps | Intake, lifecycle stages, idempotent review tasks, local audit | Role-based authentication, third-party CRM, cron scheduling |
| Private knowledge | Local keyword retrieval, document attribution, export policy | Ollama/LLM inference, encrypted vault, Notion/n8n sync |
| Webhook lab (existing) | Browser-safe retries, failure inspection, idempotency illustration | Deployed production webhook endpoint |
Data handling
All initial records are fabricated examples. Interactive demos run JavaScript locally in the browser with no API calls or persistent storage.
Files selected in the reporting demo are parsed locally; visitors should not upload confidential information because local processing alone is not a complete information-security guarantee.
No live client CRM, accounts, payment system, Notion workspace, or email inbox is accessible to these demonstrations.
AI-assisted client delivery: AI tools may support analysis, implementation and testing only within the agreed project boundaries. Client confidential data is not forwarded to external AI providers without appropriate consent and a defined handling method. The public contact form must not be used for secrets.
Acceptance approach
01. Document expected behavior, data sources, failure modes and access constraints.
02. Establish a limited first milestone with a written acceptance checklist.
03. Test both normal and adverse cases, recording failure handling and limitations.
04. Hand over editable code/configuration, test evidence and operator notes.
Reproduce the engineering validation
The packaged source includes tests/test_mvp_core.js and tests/test_browser.py. The former tests deterministic business rules; the latter visits each demo, performs primary interactions, and captures desktop/mobile screenshots. Test status and environment are recorded in the packaged docs/TEST_REPORT.md.
Need a production integration?
Show us an anonymized workflow and the acceptance criteria. We will scope the real authentication, integration, logging and deployment requirements.