Khan — AI Project Manager for Small Agencies

The Business Problem

Big0 runs 3–5 client projects at any given time across different time zones. The founder is the project manager, the sales lead, and half the engineering team. When you're wearing that many hats, things slip. A sprint falls behind and nobody notices until the deadline. A client hasn't received an update in two weeks. A budget overrun gets caught at invoicing instead of mid-project.

We tried the usual tools. Trello and Notion were too simple — fine for task lists, useless for actual project management. Jira and Monday.com were built for 50-person teams, not a small agency. None of them did the one thing we actually needed: watch our projects and tell us what was going wrong before it became a fire.

So we built it.

The Solution

Khan is a project management platform with an AI analysis engine that runs a continuous loop across all active projects. It follows a four-step cycle: see, analyze, propose, act.

The AI sees everything. It monitors tasks, risks, timelines, budgets, and stakeholder engagement across every project. Overdue tasks, stale work items, engagement gaps, sprint velocity drops, budget burn rate anomalies — all tracked continuously.

Analysis is specific, not generic. Khan doesn't flag "project at risk" and leave you to figure it out. It runs seven distinct checks: overdue tasks, stale in-progress items, unmitigated risks, stakeholder engagement gaps, sprint health, budget health, and empty backlogs. Each check produces concrete findings with context.

Proposals include reasoning. When Khan finds a task that's been overdue for 10 days, it doesn't just highlight it red. It proposes a specific action — reassign to a team member with available capacity, adjust the sprint scope, or escalate to the client — with an explanation of why.

Humans stay in control. Always. Every proposed action requires explicit approval before execution. This isn't a setting you can toggle off. It's a design principle. Khan handles the monitoring and analysis. You make the decisions.

The platform covers four integrated modules. Project management handles tasks, dependencies, sprints, risk registers with probability/impact scoring, and budget tracking with Earned Value Management. Client management includes CRM, pipeline, proposals, and a client portal for project visibility. Finance covers invoicing, time tracking (billable vs. non-billable), and revenue forecasting from pipeline data. Khan AI ties it all together with cross-project analysis and a knowledge hub that learns team patterns.

The Methodology

Khan follows the Project Management Institute framework across the full project lifecycle: Initiation, Planning, Execution, Monitoring & Controlling, and Closing. This distinction matters. Most PM tools for small teams are glorified to-do lists with a Kanban view. They track what you tell them to track. They don't understand project management as a discipline.

Khan's AI analysis is grounded in PMI concepts — Earned Value Management for budget health, velocity tracking for sprint forecasting, risk registers with probability and impact matrices, stakeholder engagement assessments. The AI doesn't just pattern-match on overdue dates. It understands why a CPI below 0.9 is a problem and what the standard responses are.

The Results

Khan is now a standalone product in free beta at khan.big0.dev, used on real agency projects. It was built from our own operational pain, validated on our own client work, and released because we believed other small agencies had the same problem.

The platform catches issues that used to surface too late. Sprint velocity drops get flagged at 40% deviation, not at the missed deadline. Budget anomalies surface during the project, not at invoicing. Clients get regular updates because the system flags engagement gaps before they become relationship problems.

The human-in-the-loop model means teams trust the tool. There's no anxiety about an AI making changes autonomously. Proposals are reviewed, approved or rejected, and the reasoning is preserved — creating a decision log that improves future analysis.

What started as an internal tool to manage our own chaos became a product because the problem is universal. Every small agency with 5–20 people running multiple client projects has the same blind spots. Khan watches for them.