JAGGYScope & prioritisation
Translate the goal into a bounded brief. Decide what matters now, what can wait and what falls outside the work.
JACK / PROJECT MANAGEMENT WITH AI
Human judgement.
AI assistance.
Work you can assess.
AI can help produce the work. A project manager makes it the right work: setting the scope, connecting the tasks, checking the evidence and deciding what is ready.
01 / WHAT THE PM ADDS
A useful tool still
needs a clear direction.
Jack’s value is in shaping, coordinating and challenging the work, rather than treating generated output as evidence of completion.
Delegation & dependencies
Give each task a clear owner, input and acceptance condition. Sequence work so one missing decision does not hide behind activity.
Independent review
Separate producing an answer from challenging it. Check assumptions, routes and claims against evidence rather than confidence.
Risk & acceptance
Make the trade-offs visible. Accept what meets the agreed conditions, request rework where it does not, and keep unresolved risks on hold.
02 / CASE STUDY
Jaggy.
Helps, Not Creates.
A music companion built around the musician’s creative control. A website built around the same principle: useful help, clear limits and honest status.
Jack manages the priorities and acceptance. Tiggy, Mother and Cassius are credited as AI collaborators in the wider Jaggy work. The collaboration supports the work; it does not transfer responsibility for decisions away from Jack.
JackHuman project manager & acceptance ownerTiggy · Mother · CassiusAI collaborators
A usable resource, a bounded promise.
The site links to a directory of over 600 plugin and FX entries. Responsive checks and keyboard navigation inform website review. The established Jaggy logo and a clear development status are preserved.
Explore the directory resourceWebsite UAT reported passed.
On 12 October 2026, Jack reported that the existing website behaved as expected in his UAT. The formal acceptance record is to follow. That report concerns the earlier website; this new Portfolio page was added afterwards and requires its own review.
Testing is not a completed release.
The demonstrated reference adapter is Windows/REAPER. Broader DAW independence is a design direction. Installer and recovery acceptance remain on hold; this portfolio makes no production download or verified recovery claim.
03 / MAKE THE PROCESS TANGIBLE
Artifacts you can inspect.
A structure you can reuse.
These are generic, public-safe examples and blank templates. They illustrate a working method; they are not live project records, private reviews or proof of completed product acceptance.
EXAMPLE 01 / PLAN
Download example as plain textA task board with dependencies
| Task | Owner role | Depends on | Acceptance condition |
|---|---|---|---|
| A · Agree the brief | Human PM | Goal and constraints | Scope and exclusions are explicit |
| B · Prepare the draft | AI collaborator | A | Draft addresses the agreed brief |
| C · Independent review | Separate reviewer | B | Claims and behaviour have evidence |
| D · Acceptance decision | Human acceptance owner | C | Accept, rework or hold is recorded |
A → B → C → D. Rework loops back to the relevant task; activity alone does not close the dependency.
EXAMPLE 02 / VERIFY
Download template as plain textAn acceptance template
Before testing
Version, scope, environment and agreed acceptance criteria.
For each test
Action, expected result, observed result and evidence reference.
At the decision
Pass, fail or blocked; open defects; accept, rework or hold; human owner and date.
Blank by design. A template is not a completed acceptance record.
EXAMPLE 03 / EXPLAIN
Download template as plain textA decision & evidence record
Illustrative decision
Hold a change when its acceptance evidence is incomplete. Record the question, options, evidence quality, unresolved risk and what would allow the decision to change.
Useful decisions retain uncertainty. They distinguish an observation, an inference and a claim that still needs checking.
04 / TIME & COST
Show the value.
Measure it honestly.
There is no measured savings figure or ROI claim here yet. A fair comparison needs an equivalent scope and quality bar, a defensible human-only baseline and the total assisted effort, including supervision, review and rework.
Human-only effort
Record the person-hours a comparable task would need without AI, with the basis and uncertainty of that baseline.
Not measuredTotal assisted effort
Count human planning, prompting, coordination, review, testing and rework. Record AI run time separately; overlapping runs are not human hours.
Not measuredActual money spent
Record actual tool charges and other attributable paid costs separately. Unpaid person-hours are effort, not cash expenditure or cash saved.
Not reportedComparable effort difference
Human-only baseline person-hours − total assisted human person-hours.
Keep the comparisons separate
Report elapsed time separately from effort. Any cash saving needs its own comparable paid-cost baseline. Disclose estimated baselines and the review/rework included.
The result may be positive, zero or negative. A lower number only means something when the scope and accepted quality are comparable.