The network for creativity
Join 1.25M professional creatives like you
Connect with clients, get discovered, and run your business 100% commission-free
Creatives on Contra have earned over $150M and we are just getting started
Back to feedPost
What happens between an AI recommendation and an approved action? Most AI demos stop at generating a signal, ignoring the operational gap where production risks actually live.
I built Investment DecisionOps AI to engineer that workflow from the ground up—focusing on auditability, strict permission boundaries, and deterministic replay rather than autonomous execution.
Key architectural choices behind the system:
• Read-Only Agent Boundaries: CrewAI agents research documents and portfolio context via a stdio MCP tool server restricted strictly to read and compute tools. Agents can analyze and propose, never approve or trade.
• Tamper-Evident Evidence Logs: Decision contexts are frozen with per-item SHA-256 hashes so historical replays never drift when underlying data changes.
• Payload-Bound Approval Gates: Orders require previewing, step-up re-authentication tied to the exact payload hash, and broker reconciliation.
• Lean Modular Monolith: Built with React, FastAPI, Docker, Ollama (qwen3:4b), and PostgreSQL 18 + pgvector handling relational state, vectors, full-text search, and FOR UPDATE SKIP LOCKED background jobs without Redis or Kubernetes bloat.
Designing the brakes takes just as much thought as designing the agents. How do you enforce hard boundaries between AI outputs and state-mutating actions?
Code & architecture: https://github.com/Oren1984/investment-decisionops-ai
The network for creativity
Join 1.25M professional creatives like you
Connect with clients, get discovered, and run your business 100% commission-free
Creatives on Contra have earned over $150M and we are just getting started
Related posts
Proud to officially add the AWS Certified Machine Learning Engineer - Associate certification to my toolkit!
Building robust machine learning pipelines and scaling AI workloads securely on the cloud has always been a key focus of mine. Earning this certification has been a great deep-dive into mastering AWS ML infrastructure, MLOps best practices, and production-grade architectures.
Ready to bring these advanced capabilities to upcoming data science and engineering projects! Let's build something impactful. 💡💪
🔗 Verify Certificate:
Validation Number: 74c5808986a64b84addcaa9db725b8bd
Verification Link: https://aws.amazon.com/verification
#AWS #MachineLearning #AI #CloudComputing #DataAnalysis #MLOps
Congratulations on earning the AWS Machine Learning Engineer certification, Hazem khaled! Robust pipelines and cloud automation make all the difference for production-grade AI systems.
Everyone is shipping AI agents right now. Almost no one is testing them — and that is the whole problem.
A demo agent and a production agent are different animals. The demo works because you walked it down the happy path. Production breaks on the paths you never imagined.
A real example from my own work: a support assistant correctly handed a conversation to a human when someone wrote "I want to talk to a person" in Spanish — but not in English, because one phrase was missing from a rule. Every automated test passed. It only surfaced when I ran real conversations through it.
That is the gap the "agents will replace everything" wave keeps skipping. The hard part was never making an agent act. It is making it:
• say "I don't know" instead of guessing,
• hand control back to a human at the right moment,
• and prove, with tests, that it still behaves when the input gets strange.
Autonomy is cheap. Trustworthy autonomy is the real engineering.
Before you let an AI agent near real users or real data — what is the first thing you check?
Most inbound bottlenecks aren't caused by low lead volume—they happen during triage.
When inquiries roll into a Google Sheet or CRM, operations teams often waste 10–15 minutes per submission reading through messy paragraphs, deciphering budget mismatch, and manually deciding who gets an immediate call vs. a standard rate sheet.
Instead of adding a bulky CRM add-on, I documented a lightweight 3-step automation pipeline built inside Make:
Intake & Extraction: An active listener triggers immediately upon a new submission, extracting core variables: client scope, target deliverables, timeline, and stated budget.
Context-Aware Evaluation (Gemini 2.5 Flash): The payload passes through an AI evaluation node with strict guardrails:
Evaluates budget against minimum agency criteria.
Assigns a triage score: Hot (immediate priority), Warm (nurture), or Cold (disqualified/template reply).
Generates a strict 15-word operational summary—no long paragraphs.
Determines an automated next step (e.g., routing to an account lead or flagging for asynchronous follow-up).
Data Return & Clean Pipeline: Appends the structured classification directly back to the master operational tracker, keeping the sheet filterable and clean.
The outcome: The sales and operations team can filter the entire board by priority status in two clicks, cutting first-touch response time to minutes without manually reading every single submission.
I’ve documented the full scenario map, execution logic, and data schemas below.
Curious how other agency operators and automation leads here handle their lead enrichment layer—are you doing AI scoring at the form level, or later in the CRM pipeline?:
For the budget gate, I'd keep an unstated budget separate from one explicitly below your minimum. Does the Make route ask for clarification when the amount or currency is missing, so a potentially suitable lead doesn't fall into Cold by default?