📖 Telecall AI Fleet — Wiki

The telec.ai portal is the central hub for everything Telecall — product information, AI fleet operations, project tracking, and team accountability. Here's a quick map of what you'll find and who each section is for.

Plane is the team's project management system — think of it as a shared to-do list that both humans and AI agents can read and update. Every piece of work that matters gets tracked here: from a quick fix to a major project milestone.

The URL is plane.kelmanajuz.com. AI agents update tasks automatically as they work. Humans use it to track progress, set priorities, and review completed work.

🗂️ Understanding Task Types

There are four types of tasks, each with different levels of review. The review level reflects how much impact the task has on the business:

📌 Task Status Flow

Every task moves through these stages:

🏃 Projects in Plane

ALMA and HOLLY are the two AI agents you'll interact with most as a human team member. They are your interface to the entire Telecall AI fleet — you talk to them, they coordinate everything else.

✅ What You Can Ask ALMA or HOLLY

• Create a new task for the team: "ALMA, we need to update the Vibra coverage map by Friday."
• Get a status update: "HOLLY, what's the current status of the Colombia launch preparation?"
• Assign work to a specific agent: "ALMA, ask ORACLE to research competitor pricing in Bogotá."
• Review completed work: "HOLLY, summarize what the fleet accomplished this week."
• Schedule or plan: "ALMA, set up a weekly review meeting for the Colombia team."
• Get explanations: "HOLLY, explain what TURING is working on right now."
• Escalate issues: "ALMA, this task has been stuck for 3 days — escalate it."
• Request reports: "HOLLY, give me a report on fleet health this week."

🔐 Permission Levels — Explained Simply

Not everything needs the same level of authorization. Here's how it works:

💬 How to Reach ALMA or HOLLY

• Mattermost #fleet-ops: Tag @ALMA or @HOLLY in any message
• RocketChat: Direct message their accounts
• Telegram: Use the Maestro bot — it relays requests to the right agent

Our AI agents are powerful, but they operate within clear boundaries. These guardrails exist to protect the business, maintain data integrity, and ensure human oversight stays in control of important decisions.

✅ Things You CAN Ask Agents to Do

🚫 Things Agents Will NOT Do (Hard Limits)

• No fabricated data: Agents will never invent numbers — subscriber counts, revenue figures, or market statistics must come from real sources. If data isn't available, the dashboard shows "No Data Available."
• No unsupervised spending: Any commitment that costs money requires human approval first. Agents cannot authorize purchases or contracts on their own.
• No external communications without approval: Agents won't email customers, partners, or press contacts without a human reviewing the message first.
• No production changes without verification: Any change to a live service gets tested and verified before being considered complete. If something breaks, it gets rolled back immediately.
• No bypassing review: When a task requires human approval, agents will wait — they will never proceed without the sign-off they're supposed to get.
• No self-modification: Agents cannot change their own configuration, permissions, or safety settings. Only Allan can do this.
• No data fabrication for dashboards: Charts and metrics only show real, verified data. Empty charts are honest — they mean data isn't available yet.

🔄 The Human Override Principle

You can always override or pause an agent's work. If something doesn't look right, say "stop" or "pause" and the agent will halt immediately. Human oversight is always the top priority — AI efficiency comes second to human control.

The Colombia Launch is Telecall's first commercial deployment of Vibra 5G Fixed Wireless Access. This is the flagship project for 2026 — bringing AI-native internet connectivity to Colombian homes and businesses.

👥 Project Leadership

🗂️ Project Workstreams

📅 Key Milestones

🤖 AI Agents Assigned to Colombia

• ORACLE: Market research — competitor analysis, pricing benchmarks, regulatory landscape
• LINCOLN: Legal and compliance — contract review, regulatory monitoring
• DALIO: Financial modelling — unit economics, cash flow, investor decks
• GIRARD: Sales pipeline — lead tracking, partner onboarding, conversion metrics
• TURING: Technical integration — AI Ops for Colombia network, API connections
• NEXUS: Project coordination — sprint planning, task tracking, milestone reporting

The Telecall AI fleet consists of 17 specialized AI agents. Each agent has a defined domain and area of expertise. Think of them like a team of specialists — you wouldn't ask your accountant to fix your network, and you wouldn't ask your network engineer to write a legal contract. The agents work the same way.

You don't need to contact agents directly. Use ALMA or HOLLY, and they'll route your request to the right specialist.

🔗 How to Reach an Agent

You never need to contact agents directly. The workflow is simple:

1. Tell ALMA or HOLLY what you need. Be as specific or as general as you like — "I need a report on Colombia competitor pricing" is enough.

2. ALMA identifies the right specialist and routes your request to them. For the example above, she'd assign ORACLE.

3. The agent executes the task, updates the Plane tracker, and reports back when done.

4. ALMA notifies you when the work is complete and delivers the output.

Several important tasks run automatically on a schedule — no human needs to trigger them. This section explains what runs, when, and what you should expect to see as a result.

📅 Daily Automated Jobs

📅 Weekly Automated Jobs

📅 Monthly Automated Jobs

⚠️ What Happens When Something Goes Wrong

If a critical service fails or an automated job produces an alert, the escalation chain is automatic:

1. GRID is notified immediately and begins investigation.

2. If GRID doesn't respond within 15 minutes, TURING takes over.

3. If unresolved after 24 hours (P0 issues), Allan is alerted directly via Telegram.

4. The leadership team is informed via Mattermost #fleet-ops when any service disruption is confirmed.

RaC stands for Responsibility and Accountability. It's our weekly and monthly cadence for making sure the team — both human and AI — is delivering on commitments. Think of it as a structured check-in ritual that keeps everyone honest and prevents things from falling through the cracks.

📅 Weekly RaC — Every Friday

The weekly RaC is a structured review of the past 7 days. It's automatically compiled by NEXUS and delivered to the leadership team. Here's what it covers:

📅 Monthly RaC — Last Friday of Every Month

The monthly RaC is deeper. It's a retrospective that looks at the full month and asks harder questions:

• What did we commit to this month — and did we deliver? Line-by-line review of monthly objectives.
• Where did we miss? Honest analysis of any targets that were missed — no excuses, just root cause and fix.
• What rule was broken? If any standing rule was violated (e.g. a service was changed without backup, a task sat unverified), it's documented here.
• What did we learn? New patterns, new risks, new opportunities identified during the month.
• What changes for next month? Updated priorities, new standing rules, reassigned responsibilities.

📊 The RaC Dashboard

The live RaC dashboard is available at telec.ai/rac-dashboard. It shows real-time task status, sprint progress, and key metrics. Updated automatically by the fleet every time a task changes state.

👤 Human Accountability in RaC

The RaC ritual isn't just for AI agents. Human team members have commitments tracked too: