Practical Work Delivery with Multiple AIs

Why this book exists

1. How this book was made
2. Edition and rights

Reader guide and glossary

1. Practical Work Delivery with Multiple AIs
2. Companion files
3. Glossary
4. Chapter list

Chapter 01 — From an idea to a usable result

1. Start with the use, then choose the output
2. A small research assignment
3. Picture the finished result
4. Give the check a different job
5. Give each tool a clear job
6. Write your own brief and test the handoff

Chapter 02 — Plan, Do, Check, Act with two or more AIs

1. Start with one small result and a limit
2. Give producing and checking different jobs
3. Follow an actual correction, with its limits visible
4. Match the check to the output
5. Keep a small cycle record

Chapter 03 — Prepare your computer and AI CLIs

1. Start with a small, known setup
2. Check Python before the trial
3. Complete one workstation checklist
4. Choose licences and a trial allowance
5. Make a workspace and run one bounded task
6. Keep the diagram editable
7. Check your result

Chapter 04 — Give each AI an identity and the right access

1. Separate six things that look like one identity
2. Give each role the access it needs
3. Set up the shared-tool accounts and connections
4. Read the reference map honestly
5. Make your own map and check it
6. Keep shared GitHub use attributable
7. Use access evidence without creating false blockers

Chapter 05 — Use Jira to manage work units

1. Make the work small enough to finish and check
2. Copy this work-unit card
3. Understand the states before moving them
4. Try a four-item practice queue
5. Connect as the right actor and make one small action
6. From this queue to bounded automatic pickup

Chapter 06 — Use Confluence for durable records and shared memory

1. Leave five connected records
2. Make current and superseded unmistakable
3. Write a handoff the next session can use
4. Revise a diagram without losing the editable source
5. Keep missing screenshots in a separate inventory
6. Check what the next AI can recover

Chapter 07 — Use GitHub for durable files and outputs

1. Understand a version without learning a programming workflow
2. Use the versioned sample
3. Check a fingerprint with one command
4. Write a manifest that removes guessing
5. Choose storage that fits the files
6. Recover, edit and export the native diagram
7. Separate the service actor from an author label
8. Check the exact delivery

Chapter 08 — Connect the tools and prove the setup works

1. Start from the accounts you already prepared
2. Give the trial a finite boundary
3. Make and check one small output
4. Revise a picture from its editable source
5. Observe a denied operation and a real stop
6. Copy this connection record
7. Know what makes this automatic

Chapter 09 — Write work units an AI can complete and check

1. Describe a result, then its proof
2. A small card that is actually usable
3. Add detail when the decision needs it
4. Fill the operating fields before pickup
5. Make missing information a defined outcome
6. Split by useful outputs and dependencies
7. Check that the card can travel

Chapter 10 — Build the queue and automate ticket pickup

1. Put a boundary around the queue
2. Claim once, then show the claim
3. One owner and one delivery — BOOK-VIS-010
4. Give publication its own gate
5. Reference configuration you can reproduce
6. Read what the evidence actually proves
7. Try the protocol on your work

Chapter 11 — Plan and do the work

1. Load the job before doing the job
2. Make a plan with a finish
3. Do the work and leave checkpoints
4. The candidate exists
5. A handoff another AI can use

Chapter 12 — Check the result and expose blind spots

1. Give the checker the whole job
2. Match the evidence to the result
3. Research: challenge the claim, not just its wording
4. Presentations: inspect the thing the audience sees
5. Software: use observable results and counterexamples
6. Write a finding someone can act on

Chapter 13 — Act on findings and repeat with purpose

1. Start with the finding and the evidence
2. Choose a response that fits
3. Correct a real error, then check the change
4. Disagree with a reviewer when the source warrants it
5. Ask a precise question when facts are missing
6. Use another cycle for a stated purpose
7. Leave a disposition record someone can follow

Chapter 14 — Set budgets, stopping rules and escalation

1. Make the allowance belong to the work
2. Write an enforceable policy
3. Keep the record when work stops — BOOK-VIS-008
4. A retry is not a new plan
5. Recognise a stop that matters
6. One set of resume-authority classes
7. Make stopping practical
8. Separate measured usage from guaranteed limits
9. Check that the policy really stops work

Chapter 15 — Recover from failures and resume without repeating work

1. Keep enough evidence to restart
2. Use the recovery card
3. Decide what may happen next
4. Repair the record before repeating the delivery
5. Resume from the original record — BOOK-VIS-009
6. Try six finite exercises
7. Record the outcome a colleague needs

Chapter 16 — Deliver outputs and close the work unit

1. Check the result against the whole request
2. Assemble one usable package
3. Reopen what the recipient will receive
4. Make closure follow the recorded authority
5. A completed delivery card
6. Preserve partial work without calling it complete
7. Close the loop with a useful lesson

Chapter 17 — Complete a research assignment

1. Define the question and the stopping point
2. Let an eligible work unit be picked up
3. Collect evidence and look for a contrary answer
4. Follow the recorded correction
5. The complete research report
6. Publish the result and the work history

Chapter 18 — Create and verify a PowerPoint presentation

1. Make a presentation someone can use
2. Plan the message and the checks
3. Produce an editable story
4. Check the file people will receive
5. Follow the actual correction trail
6. Deliver and leave a useful record

Chapter 19 — Build a software deliverable

1. Specify useful behaviour
2. Retrieve and run the exact package
3. Check behaviour independently
4. Use review to challenge the implementation
5. Hand over a usable result

Chapter 20 — Use lessons to improve the next cycle

1. Make the next task easier to get right
2. Turn an observation into a checked change
3. Measure the work without claiming more than you know
4. Compare equivalent work before changing the process
5. Maintain the setup that makes delivery possible
6. Complete the cycle

Companion files

1. Obtain and restore the files
2. Choose an exercise
3. Check an output before starting an AI
4. Run automation within a new, explicit allowance
5. Revise editable pictures and keep the record

Edition evidence and limits

About the author