← All writing

Why I started building Khan

The project-management problems inside Big0 that became the starting point for Khan, and the decision to keep people in control.

Adapted from the Big0 case study. Khan began at Big0 and is now my focus at Khan AI. Big0’s original case study.

Khan began with a problem inside Big0: a small team was running several client projects, while the founder was also handling sales and engineering. Tasks, budgets, and client updates all needed attention. Recording that work was easier than noticing when something had started to slip.

The original Big0 case study describes the product we built around that problem. This is an account of those design choices, rather than a list of current product features.

Start with a question the team needs answered

A board can show which tasks are overdue. The harder question is what that means for the project: is a dependency blocked, is the sprint overloaded, or does someone need to speak to the client?

Khan’s starting point was to look across project information rather than treat each task as an isolated record. The original analysis covered seven areas:

  • Overdue tasks and work that had stopped moving.
  • Risks with no mitigation plan.
  • Gaps in communication with stakeholders.
  • Sprint progress and budget health.
  • Empty backlogs that could leave the team without a clear next step.

These checks gave the analysis a concrete scope. “Something is wrong” is less useful than identifying the work item, the relevant context, and a decision that needs attention.

Separate observing from acting

The workflow followed four steps: observe, analyse, propose, and act. A finding could lead to a proposed reassignment, a change in sprint scope, or an escalation. The proposal included the reasoning behind it.

Execution was a separate step requiring human approval. The system’s job was to make a decision easier to evaluate, with enough context to accept or reject it. Reviewing those proposals also created a record of the decisions made.

This boundary mattered because project information is incomplete. A delayed task might already have an agreed workaround. A budget variance might reflect a deliberate change in scope. The person responsible for the project needs to be able to supply that context.

Use project-management concepts underneath the interface

The original design drew on the PMI project lifecycle, risk registers, stakeholder engagement, and Earned Value Management. Those concepts supplied structure for the analysis: what information to collect, which signals to inspect, and what kind of response to propose.

The surrounding platform brought together project work, client information, and finance. That reduced the need to reason about a budget without its tasks, or a client relationship without its recent communications.

From an internal problem to my next company

The useful part of Khan’s origin story is its specificity. It began with coordination problems we were encountering in our own client work, rather than a general ambition to add AI to a task list.

Following Big0’s acquisition in August 2026, Khan AI became my current focus. The underlying question remains the one that started the project: how can software help a project manager notice the right problem early enough to make a useful decision?

More from the blog →