Your cart is currently empty!
CAAP + AI AHEAD SIGNATURE SHOWCASE 04 OF 06
How will I learn?
Experience how CAAP turns a business challenge into a capability you understand, a decision you can explain and a method you can reuse.
Created using CAAP capabilities
CAAP learning: Independent reuse, testing and AI-assisted workflow design.
Tool used: OpenAI Codex for source-grounded writing and interactive webpage development. Revised 29 October 2026.
The examples are staged from fictional case evidence. No live AI calls or installed Claude skill execution take place here.
Learn through decisions, demonstrations and reuse
Take part in the learning.
CAAP’s trainer-led business cases introduce useful AI capabilities and make the professional decisions around them visible. You consider an approach, predict or inspect a result, and discuss why it works. Follow-up tools help you recreate the method afterwards — without needing to programme.
The formats below serve different purposes. Choose your AI route opens up a decision. Predict → Demonstrate → Explain → Retry reveals a capability and its limits. Spot → Improve → Apply develops judgement. The Skill Handover Builder supports reuse after training. A case uses the formats that suit its objective.
One shared example · W5-1B
The Weekly Client Report
At fictional Kettlewick Digital, the Account Lead has prepared Tessaly Home’s one-page report every Friday at 3 p.m. for fourteen months. Now a newer colleague must take over during three weeks of leave. What must you leave behind so she can produce it without calling you?
Non-coding · Six independent weekends · 24 cases · 24 live hours. Explore one example or try them all; there is no required order.
Select a learning method to see an example.
Open a method below to read its example. The selector activates when the page is ready.
01 · Choose your AI route
Start by choosing an approach to a realistic work challenge. The discussion makes the assumptions behind each route visible.
01 · Choose your AI route
What would you leave behind?
It is Wednesday, 4:10 p.m. You are the Account Lead. From Monday, the Account Executive covers the account. She joined six weeks ago and has never made the report.
Choose a route for feedback, or open the explanation below.
Read the handover principle
C is the stronger handover, once checked. A precise prompt can carry rules, but the chat history mixes current rules, superseded rules and one-week exceptions. Past reports show format and examples, not every decision. A skill packages reusable instructions; it becomes dependable only with verified rules, a cold run, tests, boundaries, an owner and review triggers.
A downloadable skill’s popularity does not establish safety. Read unfamiliar skill files as documents before anyone enables them. Do not let a reporting skill read unrelated files or send the report itself.
02 · Predict → Demonstrate → Explain → Retry
When a capability is unfamiliar, predict what it will do, inspect a demonstration, explain the result and retry with a changed request. This staged example lets you explore that sequence.
02 · Predict → Demonstrate → Explain → Retry
The surprising capability is reuse without the original conversation.
A Claude Custom Skill can carry named instructions and supporting files for repeated work. Here the important move is writing down what “the usual” means, so another person can invoke the method with fresh inputs.
Staged walkthrough, not a live Claude demonstration. The instructor guide reports Claude-model dry runs outside the Claude app, including reading a skill folder as files. It explicitly says no installed skill was run in Claude or Gemini. The panels below reproduce documented observations and checked fixture targets; clicking them does not execute a model or upload your data.
Predict
In a new chat with only the three exports, will “please do the usual Friday report” recover this client’s current reporting rules?
Make a prediction, then inspect the comparison. You can also read it directly.
Demonstrate · inspect the documented baseline and checked target
Baseline · guide’s reported observation
New chat, no report skill
Prompt: “Please do the usual Friday report for this week from the three exports attached.”
The reported dry run acknowledged it did not know “the usual”, then drafted on stated assumptions. A real run might instead stop to ask questions; that is a useful response.
Reusable route · checked case target
Current rules + template + fresh inputs
Extract current rules from history and emails, confirm gaps with the owner, and package them as a skill. A different user then runs it in a new chat with the exports and this week’s notes.
The guide’s file-based cold runs matched the reference figures. Native app installation and invocation still require rehearsal.
| Decision | Baseline reported in guide | Client’s checked rule and target |
|---|---|---|
| Trading week | 11–17 December 2026 | Thursday 10–Wednesday 16 December |
| Revenue source | Web analytics, three paid channels | Client sales sheet, Online rows, net of returns |
| Spend | ₹19.15 lakh; compared with prior week | ₹18.8 lakh against ₹18.0 lakh plan |
| Return | 2.19, labelled ROAS | ₹3.8 back for each ₹1 spent |
| New customers | 422 | 412, from client sales sheet |
These are source-reported outcomes, not guaranteed model behaviour. The plan is absent from the three exports, so the exports alone cannot establish spend against plan.
Explain
A skill carries a method; it does not make the method correct. The owner must remove replaced instructions, confirm open questions and test with known answers. The agency template supplies five headings: Headline; The numbers; What we did; What is next; Decisions needed. This week’s notes supply actions, reasons and decisions — the numbers alone cannot invent them.
Inspect the rules that make this report reusable
- Use Thursday–Wednesday, ignoring later dates in an export. This replaces Monday–Sunday.
- Use Online gross revenue minus returns from the client sales sheet, replacing platform revenue. New customers also come from the client sheet; site visits are analytics sessions, all channels.
- Use the ₹18.0 lakh weekly plan through 31 December 2026: Search ₹8.0 lakh, Shopping ₹6.0 lakh, Social ₹4.0 lakh. This replaces ₹16.0 lakh.
- Judge absolute spend variance on unrounded values: within 5% is Green; above 5% through 10% is Amber; above 10% is Red. Write the status word. Red needs a reason and action.
- One page, money and percentages to one decimal; site visits unrounded. Use plain language. Actions and next steps come only from notes; up to three dated decisions with an owner, or “None this week”.
- Require all three exports and this week’s notes. Tessaly Home online reporting only. Draft only; a person checks and sends, including on public holidays.
Retry · test what happens when the request changes
Choose an edge case, then choose the response you would build into the skill.
Try each of the three tests. A correct refusal can be a successful result.
Read all three expected test results
- Missing export: stop and name the missing web analytics export. Ask for it; do not estimate sessions or silently omit them.
- Expired plan: stop and ask the Account Lead for the January plan. ₹18.0 lakh expires on 31 December 2026; a week crossing that boundary is not covered. Supplied exports and notes must also cover the requested week.
- Wrong client: decline direct reuse for Nordby Kitchens. Its trading week, plan, thresholds and revenue rule need an owner-approved specification and new tests.
If a cold run fails, fix the source skill and run again in a fresh chat. Patching the answer in the test chat leaves the next user exposed to the same missing rule.
03 · Spot → Improve → Apply
Professional judgement grows through finding a consequential weakness, improving the approach and applying the lesson in a different situation. Try it on a convincing-looking report figure.
03 · Spot → Improve → Apply
₹5.1 looks persuasive. Is it the right return?
Deliberately flawed draft · fixture-derived
“Tessaly Home earned ₹5.1 back for every ₹1 spent this week.”
The trading week and spend are correct. The draft used ₹96.4 lakh of ad-platform revenue and ₹18.8 lakh of spend. This is a separate wrong-source example, not the baseline dry-run quotation above.
Spot · where is the consequential error?
Choose the error that changes the meaning of the report.
Improve · make the source rule explicit
The client’s email of 16 April 2026, recorded in the working history, says finance counts online orders from its own sheet and takes returns off. The 17 April instruction confirms Online rows only, excluding Store rows. The source is selected by the client’s definition of the measure, not by whichever system displays the largest number.
(₹77.5 lakh − ₹6.3 lakh) ÷ ₹18.8 lakh = ₹3.8 per ₹1 spent
Unrounded: ₹71.2 lakh ÷ ₹18.8 lakh = 3.7872…; the client requires one decimal. The ad-platform total is ₹96,39,900 (₹96.4 lakh when displayed): dividing that by ₹18.8 lakh produces 5.1276…, or ₹5.1. Correct arithmetic cannot rescue the wrong measure.
Apply · the next owner changes the reporting week
A new client trades Monday–Sunday and uses a different sales definition. Which part should transfer?
A reusable learning method does not make every client’s rules interchangeable.
Read the answer and checked report
Spot: wrong revenue source. Improve: specify the client sales sheet, Online rows, net of returns. Apply: transfer the verification method and establish new client rules.
Reference report · 10–16 December 2026
Spend ₹18.8 lakh against ₹18.0 lakh plan: 4.4% over, Green. Search ₹7.8 lakh, 2.5% under, Green; Shopping ₹6.4 lakh, 6.7% over, Amber; Social ₹4.6 lakh, 15.0% over, Red. Overall Green does not hide a Red channel.
Online revenue net of returns ₹71.2 lakh; return ₹3.8 per ₹1 spent; 412 new customers; 118,600 site visits.
Social’s Red has a reason: the ₹0.9 lakh Weekend Sale was not offset by reducing Prospecting. Actions and decisions come from this week’s notes. A person checks the Amber and Red lines before sending.
04 · Recreate after training
After a live case, a verification tool can help you reconstruct the approach for a similar task. Here you classify what transfers and build a handover checklist for your own recurring work.
04 · Recreate after training
Skill Handover Builder
Use this practice tool to reconstruct the approach for a recurring task of your own. Classify what travels, check what your handover needs and build a brief you can copy. This is a planning aid, not an executable skill or proof that your workflow is safe.
Everything stays in this browser page. No storage, analytics, uploads or AI calls are added by this component. Entries disappear on reload. Use generic task labels and avoid confidential details.
Try the method in a fresh situation
Choose a situation to rehearse a next step.
Read the complete follow-up answer key
Classification: keep cold runs and the tests/boundaries/owner/review method; replace the trading week and plan for a new client; leave out the one-week sale instruction.
Renamed column: pause, ask the owner to confirm the mapping, update the method and rerun both known-answer and refusal tests. Do not guess from a similar heading.
Another client: reuse the handover design; obtain its approved definitions, plan, sources, boundaries and examples, then test independently.
Outside skill: read it as a document before enabling it. Reject instructions that read unrelated drives, transmit data elsewhere, hide actions, approve for the user or send without review. Ratings are not permission or validation.
Case test pack: 10–16 December: ₹18.8 lakh spend, ₹71.2 lakh net revenue, ₹3.8 return, 412 new customers, Social Red. 3–9 December: ₹18.2 lakh, ₹66.0 lakh, ₹3.6, 388, all statuses Green. Add refusals for a missing export, a post-December week and another client.
How the six-weekend programme works
The live programme
Practical AI application. No programming required.
CAAP — Certified Applied AI Professional — uses business cases to connect consequential AI capabilities with worthwhile work. Trainer-led demonstrations are the primary mode. You predict, debate, make decisions, inspect results and consider how you would adapt the method.
6
Independent weekends
Each weekend is self-contained, supporting rolling entry. Complete missing capability areas as the cycle returns.
24
Business cases
Two sessions per weekend. Two cases per session: an accessible case and a more demanding case.
24
Live hours
Two hours per session × two sessions × six weekends. Four cases and four live hours per weekend.
The core spans three domains — Understand, Decide and Execute — and six capability areas. Weekends 1–6 cover Research, Knowledge & Evidence; Data, Analysis & Insight; Problems, Options & Decisions; Communication, Influence & High-Stakes Situations; AI Workflows & Productivity; and AI Applications & Agentic Work.
Discover
Encounter a worthwhile AI capability you may not yet have appreciated.
Experience
Make a prediction, observe a demonstration and develop professional judgement.
Recreate
Use follow-up verification tools to reconstruct and adapt the approach to your work.
These are complementary learning formats, chosen for the objective; every case need not use every format. Case titles and tool routes can evolve. This preview does not establish that all 24 cases have completed technical rehearsal.
From this case to your own work
You learn the capability, the judgement and the way to reuse it.
The skill is the application. Discovering the missing rules, checking a source, improving the skill and testing it cold are the professional work around it. CAAP’s live cases make those decisions visible; follow-up tools help you recreate the approach afterwards.
This page’s builder is an example of that follow-up design. It is not a promise that every case has an identical tool, and completing it is not a CAAP assessment. Certification requires Orientation completion and passes in six AI-permitted online knowledge and scenario checks, one per capability area; those checks do not independently establish workplace deployment mastery.
Use the product page for current dates, fees, availability and terms.