Predicai Academy™ — Course architecture

Five levels.
Built to hold up.

This version of the Academy closes the biggest gaps: role-specific pre-assessment hooks, explicit submission operations, gold-standard benchmarks, role variants at every level, and Momentum Log prompts that feed the final Business Case.

View pre-assessment Jump to levels
Start
15-minute entry
Role-based Monday Morning Panic
Progression
5 levels
One workflow built into proof
Review
Human-reviewed
Pass / revise / unlock flow
Outcome
Recovered capacity
Turn time saved into a business case
Start here

Your path through the Academy

New here? Start with the Panic Test, choose one workflow worth fixing, then build your first real artifact. This page explains the full system. The Notion workspace is where you do the work.

Start the Panic Test →
01
Take the Panic Test

Pick the scenario closest to your work and find the signal under pressure.

02
Choose one workflow

Use Level 0.5 to define one recurring job worth rebuilding.

03
Submit and unlock

Build the artifact, submit it for review, and progress level by level.

Pre-assessment

Monday Morning Panic, now role-aware

Keep the pressure, but make the first 15 minutes feel personal. One core mechanic. Three entry scenarios.

Scenario

Chief of Staff — Board briefing in 15 minutes

It is Monday, 8:45 AM. Your CEO forwards a chaotic weekend thread about a $250k renewal risk. Marketing and Product are fighting. Finance says no refunds. The CEO needs a one-page Board briefing by 9:00 AM.

  • Find why the client is actually mad
  • State the financial exposure
  • Name the immediate next step
Signal

The buried lede

The real issue is not the internal fight or discount pressure. The client wants the API limit increased. The correct move is a zero-cost technical unlock, not a concession-heavy summary.

Scenario

Account Executive — Save the renewal before pipeline call

It is Monday, 8:10 AM. You have a pipeline call in 20 minutes. A major account says they are ready to leave. Your manager wants a save plan, not a story.

  • What is the actual blocker to renewal?
  • What should not be offered?
  • What concrete action could save the account today?
Signal

The commercial read

The signal is a product access issue disguised as pricing tension. The winning summary avoids discount panic and frames the account save around the technical unlock the customer already asked for.

Scenario

Ops / Marketing Manager — Launch risk before leadership standup

It is Monday, 8:30 AM. A campaign or rollout is slipping. Design says the asset is done. Product says the message changed. Finance says the budget cannot move. Leadership wants a 3-bullet decision brief in 15 minutes.

  • What is the real dependency causing the delay?
  • What is noise versus what leadership needs?
  • What single action unblocks the work?
Signal

The operational read

The signal is usually a hidden dependency or ownership gap, not the loudest disagreement. The brief should isolate the true blocker and force a decision, not retell the thread.

Operating model

Submission, gating, and unlocks

The student experience is only half the system. This section defines what happens after “submit.”

1. Submission intake

Student submits a PDF, link, or approved tracker entry. Every submission is tied to one level and one named artifact.

2. Review SLA

Promise within 48 business hours publicly unless you know you can sustain 24. Beat the promise when possible.

3. Review outcome

Outcomes are Pass, Revise, or Escalate. Revise always includes exactly what failed and what to fix next.

4. Unlock message

On pass: “Your artifact holds. Level unlocked.” On revise: “Your system is not ready to scale yet. Fix the flagged gap and resubmit.”

Pass flow

What the student sees

  • Confirmation receipt
  • Status = Under Review
  • Pass note with one strength called out
  • Unlock key / next page / next button
  • One sentence telling them what to prepare before opening the next level
Revise flow

What happens on failure

  • Do not say “failed” without reason
  • Name the exact gate or criterion that broke
  • Return 1 to 3 concrete fixes only
  • Include a Cai rescue prompt
  • Keep the student in motion: “Revise and resubmit when ready”
Step 1

Submitted

The student sends one named artifact for one specific level.

Step 2

Under Review

Status is active. Public SLA: within 48 business hours unless otherwise stated.

Step 3A

Pass

The artifact holds. Student receives the unlock message and next-step direction.

Step 3B

Revise

The student gets 1–3 fixes, a Cai rescue prompt, and a clear resubmission path.

Recommended product decision

Design these HTML pages as hosted reference + guided build pages, not raw editable student forms. Students should build their actual work in Notion, Google Docs, or exported PDFs. The hosted HTML should teach, benchmark, and direct action. That keeps the Proof Checklist style and the course experience aligned.

Course levels

Gold standards, role variants, and task sheets

Every level below includes the missing bridge: what excellent looks like, what weak looks like, how to build it, how Cai helps, and what to log before leaving the page.

0.5
Blueprint
Choose a workflow worth building before you burn time on random tinkering.
Deliverable Project Charter
Gold standard

What “done right” looks like

  • Recurring workflow, not a one-off task
  • Specific trigger and clear definition of done
  • Pain is measurable in time, delay, or friction
  • Small enough to build in this program, large enough to matter
Common failure

What gets rejected

  • “Use AI better”
  • “Fix our whole strategy”
  • Purely creative or purely vague work
  • No recurring trigger or no measurable outcome
Bad

“I want to explore AI tools for marketing.”

Good

“I want AI to help with campaign reporting.”

Gold

“Every Friday at 3 PM, I compile campaign data from 3 sources into a one-page executive report. I want a first draft ready in 10 minutes.”

Role variants
Ops

Weekly project status summary across tools and owners.

Sales

Pipeline review brief from CRM activity, deal risk, and next steps.

Marketing

Campaign performance brief from ad channels, web, and email data.

Founder / Data

Board or leadership snapshot from metrics, risks, and operating notes.

Task sheet
1

Name the workflow. Use the exact recurring job, not a department-wide ambition.

2

Define the trigger. When does this start in the real world?

3

Write the pain. What wastes time, creates risk, or causes delay?

4

Define done. What would success look like in plain language?

5

Stress-test with Cai. Ask if the workflow is frequent, structured, and high-value enough.

6

Submit only when narrow. Broad projects should be revised before Level 1.

Cai prompt
Cai, I am starting the Predicai Academy. My proposed workflow is: [insert workflow]. Grade it 1–5 on frequency, structure, and value. If it scores below 4, tell me exactly how to narrow it down.
Momentum log prompt

Log your win before you close this page

What changed after narrowing your project? Did the work become clearer, smaller, or more measurable? Add one sentence to the Momentum Log.

01
Translator
Stop typing into AI. Start architecting the request.
Deliverable Intent Protocol
Gold standard

What 90%+ looks like

  • Role is specific and useful, not generic
  • Input includes real context, constraints, and stakes
  • Steps force reasoning, not just output generation
  • Expectation defines format, tone, and finish line
  • Output is repeatable across runs
Review criteria

Pass / revise rules

  • Pass when Role, Input, Steps, and Expectation are all explicit
  • Revise when the model still has to guess what matters
  • Revise when the prompt could apply to any job or audience
  • Revise when there is no obvious definition of done
Bad

“Summarize this thread and make it better.”

Good

“Summarize this for my manager in bullets.”

Gold

“Act as a Chief of Staff. My CEO needs a 3-bullet briefing in 10 minutes. Ignore internal blame. Identify the client’s actual demand, the financial exposure, and the immediate zero-cost action.”

Role variants
Ops

Turn meeting notes into decision memo, owners, deadlines, and unresolved blockers.

Sales

Turn deal thread into renewal risk summary, commercial exposure, and save plan.

Marketing

Turn campaign results into a one-page performance brief for leadership.

Founder / Data

Turn mixed updates into board-ready insight, risk, and next decision.

Task sheet
1

Pick the job. What exactly needs to be produced?

2

Write the role. Give the AI a job title and a useful bias.

3

Add the context. Who is this for, what matters, what is risky, what is urgent?

4

Define the steps. Make the model reason in sequence.

5

Define the output. Format, tone, and what “good” looks like.

6

Run and revise. Check if the output still feels generic or misses the signal.

Cai prompt
Cai, here is my draft Intent Protocol using R.I.S.E. Grade it against the Architect Standard. Tell me what context, risk, or reasoning step is still missing.
Momentum log prompt

Log your win before you close this page

What became easier after using R.I.S.E.? Note one moment where you got to a usable result faster than your old method.

02
Operator
See the drudgery before you try to automate it.
Deliverable Time Ledger
Gold standard

What 90%+ looks like

  • 3 real days of work, not guesses
  • Tasks classified as Automate, Augment, or Human-only
  • Hours/week and annual cost calculated
  • Top Zombie Task clearly identified
  • Evidence points forward to Level 3 build target
Review criteria

Pass / revise rules

  • Pass when the ledger shows real repeated work and monetized drudgery
  • Revise when tasks are vague, one-off, or emotionally labeled without evidence
  • Revise when annual cost is missing
  • Revise when no workflow candidate emerges from the audit
Bad

“I do a lot of admin work each week.”

Good

“I spend about 3 hours a week updating CRM and recapping calls.”

Gold

“Weekly reporting = 2 hrs. CRM cleanup = 2.5 hrs. Follow-up drafting = 1.5 hrs. At $60/hr, this is $18,720/year in recoverable drudgery.”

Role variants
Ops

Status rollups, tracker updates, meeting follow-ups, owner chasing.

Sales

CRM hygiene, follow-up emails, call recaps, pipeline prep.

Marketing

Reporting pulls, channel summaries, asset routing, approval follow-ups.

Founder / Data

Metric compilation, investor notes, recurring analysis summaries.

Task sheet
1

Track 3 days. Capture what you actually do, not what you think you do.

2

Name the task. Be specific enough that someone else would understand it.

3

Classify it. Automate, Augment, or Human-only.

4

Calculate cost. Convert hours into annualized drudgery.

5

Find the Zombie Task. Pick the best Level 3 build candidate.

6

Note the buyback. What would you do with those hours if they came back?

Cai prompt
Cai, I am auditing my week. Review this task list and identify my top 3 Zombie Tasks: high repetition, low judgment work that I should target for automation first. Then tell me which one is best for Level 3.
Momentum log prompt

Log your win before you close this page

What was your most expensive recurring task? Add the task name and annualized cost to the Momentum Log.

03
Builder
Break complex work into Reader → Thinker → Writer.
Deliverable Agentic Blueprint
Gold standard

What 90%+ looks like

  • Single workflow broken into Reader, Thinker, Writer steps
  • Each step tested in isolation
  • Inputs and outputs clearly handed off
  • Failure modes named before scaling
  • Human checkpoint included before real-world action
Review criteria

Pass / revise rules

  • Pass when the workflow is auditable and small enough to run
  • Revise when one prompt is still trying to do everything
  • Revise when the Thinker step is missing
  • Revise when no verification or handoff logic exists
Bad

“Read this 40-page file and write me a strategy.”

Good

Reader summarizes, then Writer drafts output.

Gold

Reader extracts facts → Thinker compares, prioritizes, flags risks → Writer formats the executive brief. Each step tested and chained.

Role variants
Ops

Meeting notes → decision extraction → action tracker.

Sales

CRM + call notes → deal risk logic → pipeline brief.

Marketing

Channel metrics → insight synthesis → campaign briefing.

Research / Data

Source docs → finding comparison → report or dashboard narrative.

Task sheet
1

Start from Level 2. Use the highest-cost Zombie Task.

2

Define the Reader. What comes in? What facts matter?

3

Define the Thinker. What logic must happen before output?

4

Define the Writer. What exact deliverable gets shipped?

5

Test each step. Do not chain a broken step.

6

Add the human gate. Decide where review happens before action.

Cai prompt
Cai, act as a Systems Engineer. I identified this workflow as my highest-value automation target: [insert task]. Convert it into a Reader → Thinker → Writer chain. Then tell me where the weak link is and what should be human-reviewed.
Momentum log prompt

Log your win before you close this page

Where did the chain reduce friction? Record one moment where splitting the work into Reader, Thinker, and Writer made the output better.

04
Forecaster
Move from reacting to what happened to seeing what happens next.
Deliverable Signal Map + HITL
Gold standard

What 90%+ looks like

  • Signal is an anomaly, not an obvious lagging metric
  • Forecast includes linear, exponential, and Black Swan paths
  • Decision attached: Watch, Test, or Pivot
  • Confidence threshold and stop condition are named
  • Human-in-the-loop protocol is real, not decorative
Review criteria

Pass / revise rules

  • Pass when a current weak signal is tied to a future decision
  • Revise when the work only summarizes current data
  • Revise when no second-order effects are modeled
  • Revise when no monitor, threshold, or rollback exists
Bad

“Traffic is down this month.”

Good

“Support tickets about integrations are rising.”

Gold

“Integration tickets are up 40% while renewal sentiment is flat. If ignored, this predicts delayed expansion, churn risk, and product trust erosion within 6 months.”

Worked example

How a weak signal is actually found

Role

Ops Manager

Source

Weekly support dashboard, escalation log, and team handoff notes

Observation

Escalations are up 30% while total ticket volume is flat

Signal

This is not “more work.” It suggests a broken handoff or hidden complexity

Decision

Run a 2-week handoff test before service lag becomes a bigger issue

Discovery path: start with one dashboard, look for a mismatch between two related metrics, then use Cai to model what happens if that mismatch continues for 30, 60, and 180 days.

Role variants
Ops

Escalations, handoff lag, and approval bottlenecks as risk signals.

Sales

Deal slippage, stakeholder loss, or pricing pattern shifts before churn shows up.

Marketing

Channel fatigue, conversion pattern changes, or audience quality drift.

Founder / Data

Usage anomalies, cost spikes, or behavioral shifts before the metric breaks.

Task sheet
1

Find an anomaly. Look for unusual movement, not average performance.

2

Separate noise from signal. What is everyone talking about versus what actually matters?

3

Forecast 3 futures. Linear, exponential, and Black Swan.

4

Make a bet. Watch, Test, or Pivot.

5

Define the HITL rule. Who watches it and when does the AI stop?

6

Write the rollback. What happens if the signal model is wrong?

Cai prompt
Cai, act as a Strategic Futurist. Analyze this weak signal: [insert signal]. Give me the linear, exponential, and Black Swan scenario. Then define the human-in-the-loop trigger, monitor, and rollback procedure.
Momentum log prompt

Log your win before you close this page

What weak signal did you notice this week that you would have ignored before? Add it to the Momentum Log in one sentence.

05
Leader
If you do not design the system, shadow AI will design itself.
Deliverable Strategic Doctrine
Gold standard

What 90%+ looks like

  • Mission is specific and tied to business value
  • Architecture defines input, agent, and human loop
  • Governance names red lines, verification, privacy, and ownership
  • Rollout states pilot, scale, and KPI
  • The workflow is teachable to someone else
Review criteria

Pass / revise rules

  • Pass when there are actual operating rules, not slogans
  • Revise when no human owner is named
  • Revise when the red lines are soft or vague
  • Revise when the rollout does not define pilot, scale, and measurement
Bad

“Use AI responsibly and double-check the output.”

Good

“AI drafts. Humans review before sending customer-facing content.”

Gold

“No PII enters public models. Every outbound customer-facing output requires named human approval. All metrics must be verified against source systems before distribution.”

Role variants
Ops leader

Rules for intake, handoff, escalation, and approval across internal workflows.

Sales leader

Rules for customer communication, CRM automation, proposal review, and pricing claims.

Marketing leader

Rules for content generation, brand tone, factual verification, and channel approval.

Founder / Product / Data

Rules for strategic memos, forecasting, data use, and team-level enablement.

Task sheet
1

Set the mission. What exactly is this system meant to improve?

2

Name the anti-goal. What must the system never do?

3

Map the architecture. Input, agent, human loop.

4

Set red lines. Privacy, verification, tone, compliance, commitments.

5

Define rollout. Who pilots it, when it scales, how success is measured.

6

Teach it once. If you cannot teach it, you cannot scale it.

Cai prompt
Cai, audit this workflow against the 7 Gates and help me draft a Strategic Doctrine. I need mission, anti-goal, architecture, governance, rollout, and the human owner at each risk point.
Momentum log prompt

Log your win before you close this page

What rule or red line became clear today that your team did not have before? Add one sentence to the Momentum Log.

Capstone

System Dossier assembly

The final outcome is not a pile of artifacts. It is one connected system with proof of value.

Required components
  • Level 1 — Intent Protocol
  • Level 2 — Time Ledger + annualized cost
  • Level 3 — Agentic Blueprint
  • Level 4 — Signal Map + HITL
  • Level 5 — Strategic Doctrine
  • System Logic slide connecting the whole engine
  • Business Case slide using Level 2 numbers
First slide rule

Business Case first

Start with the math: [Hours/Week] × [Hourly Rate] × 52. Then scale it to the team or department. The Dossier should prove recovered capacity before it proves elegance.

What this page should do

This page should sit above the actual artifact pages as the Academy’s master architecture page. It explains the system, the standard, the student journey, and the review model. The artifact pages do the hands-on teaching. The Notion workspace remains the student operating desk.