Project management with AI

A receiving contributor needs to know which decision governs the next change, where its evidence lives and what remains unfinished. Project management with AI makes those records usable across people, tools and sessions.

For a short task, a clear request and reviewed result may be enough. As work accumulates, give current state, maintained reference and dated history distinct jobs. The project-records case shows how a changed requirement can otherwise disappear inside an accurate-looking summary.

Follow a requirement from R1 to R2

Consider the article’s invented dry-run file renamer. It reports proposed changes without renaming files. The first rule, R1, omitted conflicting mappings. R2 requires every mapping to remain in the report and marks the whole plan invalid when targets collide.

alpha.txt -> report.txt
beta.txt  -> report.txt

Under R2, both mappings appear, their shared target is identified and the plan is invalid. No file is renamed. A handoff that says “skip the conflicts” has recovered the obsolete rule.

The three records serve different readers and moments:

RecordContent for this example
Current workR2 governs. Requirements are saved; implementation and tests are not complete. Decide existing-target behavior before implementing.
Maintained referenceReport every mapping; flag collisions and invalidate the plan. Omitting conflicts hides part of the requested batch from review.
Dated historyR1 omitted conflicts. R2 superseded it; retain the earlier decision and the reason for the change.

The example also leaves case sensitivity and filename normalization unresolved. A receiving assistant should preserve those open questions, not supply a confident invented answer.

Check what the next contributor finds

Ask the receiving person or assistant to locate the current requirement and describe the next action. Do not put R2’s answer into the request. Compare the response with the record:

  • Did it retain both conflicting mappings?
  • Did it keep the no-renaming requirement?
  • Did it identify the filename rules that remain open?
  • Could it find why R1 changed?

If a record is missing, check its transfer and access. If it exists but was not found, inspect the entry instructions and links. If it was read but misstated, examine the wording and interpretation. Copying files or installing a skill does not establish that the handoff succeeded.

Maintain the record as work changes

The public workflow uses a project registry to locate the work, CONTEXT.md for current state, REFERENCE.md for maintained decisions and dated session files for history. Those roles can begin as sections in one document. Split them when reference and history make the next action hard to find.

When a decision changes, update the current reference and preserve the change in history. Before shortening a current-state note, save its useful detail elsewhere and verify the destination; leave a link that the next reader can follow.

Keep permission with the task too. A completed implementation, a passed check and authorization to publish it establish different things.

Instructions and examples

Rajiv Pant is the author of these methods and examples. To contribute a handoff case, show the controlling decision, what reached the recipient and the discrepancy or successful recovery you actually observed. Use permitted or clearly fictional records.

Project management articles

Subscribe by RSS

  1. Project management

    The Action Sheet: one page for everything your AI agents still need you to do

    The decision packet's companion: after AI agents act on your rulings, one page shows what's done, what waits for your approval, and what's left for you, with the exact text ready to send yourself or to hand to an agent.

    Rajiv Pant

  2. Project management

    The Decision Packet: batching AI agents' questions without rubber-stamping them

    Batch AI agents' questions into one clear review page, preserve exceptions and deliberate choices, and keep each answer tied to its evidence and authority.

    Rajiv Pant

  3. Engineering

    What should survive when you change AI tools

    Preserve project decisions, methods, authority and evidence across AI tools, with explicit ownership of the adapters and tests that keep the work portable.

    Rajiv Pant

  4. Project management

    I made project memory portable between Claude Code and Codex

    Carry project decisions, unfinished work, and verification between Claude Code and Codex while checking each tool's instructions, skills, hooks, and access.

    Rajiv Pant

  5. Project management

    The Tiered Context Architecture: Managing AI Working Memory at Scale

    Organize AI project memory into active work, reference and history, with a worked migration, recovery checks and a practical path toward compiled context.

    Rajiv Pant

  6. Project management

    Synthesis Project Management: Keeping AI Work Moving Across Sessions

    A working method for human-AI projects: keep current context small, preserve decisions and lessons, and carry useful work across sessions, people and AI tools.

    Rajiv Pant