Want to create interactive content? It’s easy in Genially!

Get started free

AI Systems Investigation

Marko Jerkovic

Created on May 7, 2026

Start designing with a free template

Discover more than 1500 professional designs like these:

Winter Presentation

Hanukkah Presentation

Vintage Photo Album

Nature Presentation

Halloween Presentation

Tarot Presentation

Vaporwave presentation

Transcript

Mission-based learning from UML to Formal Proof

AI Systems Investigation

START INVESTIGATION

MISSION BOARD

Debrief
Process Logic
Sysrem Intake
UC Investigation
State Control
Reflection
Class Structure
Formal Spec
Badge Wall
Sequence of Events
Final Boss
Branching
m1 BRIEFING

Users interact with the system daily.Yet no one agrees on what the system actually does.

CONTINUE
M1 OBJECTIVES

Identify system actorsDefine core use cases Clarify system boundaries

START ANALYSIS

M1 INTEL & ai trap

πŸ“‚ Legacy Evidence

πŸ€– AI Advisory

M1 ACTION

⭐ +XP

πŸ… BADGE

+50 XP

UML Explorer

Create a Use Case Diagram with a short justification.

BACK TO BOARD
m2 BRIEFING

The system functions exist.But structure defines responsibility. Hidden assumptions live in the structure.

CONTINUE
M2 OBJECTIVES

Identify core system entitiesDefine responsibilities Model relationships clearly

START ANALYSIS

M2 INTEL & ai trap

πŸ“‚ Legacy Evidence

πŸ€– AI Advisory

M2 ACTION

⭐ +XP

πŸ… BADGE

+75 XP

Structural Analyst

Create a UML Class Diagramand justify key design decisions.

BACK TO BOARD
m3 BRIEFING

The structure looks correct.Failures still occur at runtime. Something happens between components. Find out what.

CONTINUE
M3 OBJECTIVES

Reveal runtime behaviorModel component interactions Focus on one critical scenario

START ANALYSIS

M3 INTEL & ai trap

πŸ“‚ Legacy Evidence

πŸ€– AI Advisory

M3 ACTION

⭐ +XP

πŸ… BADGE

+75 XP

Behavior Analyst

Create a sequence diagramfor the withdrawal scenario.

BACK TO BOARD
m4 BRIEFING

Not all failures are timing-related.Some are decision-related. Follow the logic. Find the weak point.

CONTINUE
M4 OBJECTIVES

Identify decision pointsModel alternative flows Reveal hidden logic errors

START ANALYSIS

M4 INTEL & ai trap

πŸ“‚ Legacy Evidence

πŸ€– AI Advisory

M4 ACTION

⭐ +XP

πŸ… BADGE

+75 XP

Process Analyst

Create an activity diagramfor the withdrawal process, including alternative paths.

BACK TO BOARD
m5 BRIEFING

Some system states should never be reachable.Yet incidents suggest otherwise. Focus on states. Not on flows.

CONTINUE
M5 OBJECTIVES

Identify valid system statesDefine allowed transitions Detect forbidden states

START ANALYSIS

M5 INTEL & ai trap

πŸ“‚ Legacy Evidence

πŸ€– AI Advisory

M5 ACTION

⭐ +XP

πŸ… BADGE

+100 XP

State Guardian

Create a state machine diagramfor the Account lifecycle, including forbidden transitions.

BACK TO BOARD
M6 BRIEFING

Diagrams describe behavior.They do not enforce correctness. Rules do.

CONTINUE
M6 OBJECTIVES

Identify system invariantsDefine preconditions and postconditions Express rules precisely

START ANALYSIS

M6 INTEL & ai trap

πŸ“‚ Legacy Evidence

πŸ€– AI Advisory

M6 ACTION

⭐ +XP

πŸ… BADGE

+125 XP

Formal Reasoner

Define formal rules for the system,including invariants and preconditions, using precise logical statements.

BACK TO BOARD
M7 BRIEFING

Architecture looks correct.AI claims the system is consistent. Incidents continue. Prove the truth.

CONTINUE
M7 OBJECTIVES

Detect contradictionsValidate system consistency Challenge AI conclusions

START ANALYSIS

M7 INTEL & AI Trap

πŸ“‚ Legacy Evidence

πŸ€– AI Advisory

M7 ACTION

⭐ +XP

πŸ… BADGE

+200 XP

System Integrity Guardian

Determine whether the system is consistent. Identify violated invariants, explain why AI is wrong, and propose a minimal fix.

BACK TO BOARD
MISSION DEbRIEF

You investigated a real-world system. You learned that: β€’ diagrams are not proofs β€’ AI suggestions must be validated β€’ correctness requires formal reasoning

CONTINUE

REFLECTION

β€’ Which assumption turned out to be wrong? β€’ Where did AI mislead you? β€’ Which rule was most critical?

FINALIZE

ENGINEER’S DECISION

You cannot fix everything. If you had time for only ONE change: β€’ what would you fix β€’ and why?

achievements unlocked

M1

UML Explorer

M2

Structural Analyst

M3

Behavior Analyst

M4

Process Analyst

M5

State Guardian

achievements unlocked

M6

Formal Reasoner

M7

System Integrity Guardian

WHAT NEXT?

πŸ”

πŸ”

Revisit a Mission

Review AI Traps

🏁

πŸ“˜

Read Formal Rules Again

End Investigation

CASE CLOSED

Do not trust diagrams. Do not trust AI. Trust reasoning.

MISSION COMPLETE

THANKS!

Got an idea?

Use this space to add awesome interactivity. Include text, images, videos, tables, PDFs... even interactive questions! Premium tip: Get information on how your audience interacts with your creation:

  • Visit the Analytics settings;
  • Activate user tracking;
  • Let the communication flow!

If a state is reachable,it must be safe. Forbidden states must be explicit.

AI may remove or weaken problematic rulesto make the system appear correct. Never remove a rule without justification.

Sequence shows order.Activity shows decision logic. Missing branches hide failures.

Got an idea?

Use this space to add awesome interactivity. Include text, images, videos, tables, PDFs... even interactive questions! Premium tip: Get information on how your audience interacts with your creation:

  • Visit the Analytics settings;
  • Activate user tracking;
  • Let the communication flow!

AI assumes atomic operations.Concurrency may break this assumption.

Got an idea?

Use this space to add awesome interactivity. Include text, images, videos, tables, PDFs... even interactive questions! Premium tip: Get information on how your audience interacts with your creation:

  • Visit the Analytics settings;
  • Activate user tracking;
  • Let the communication flow!

AI states:"The system is consistent under normal assumptions."

Got an idea?

Use this space to add awesome interactivity. Include text, images, videos, tables, PDFs... even interactive questions! Premium tip: Get information on how your audience interacts with your creation:

  • Visit the Analytics settings;
  • Activate user tracking;
  • Let the communication flow!

AI often invents unnecessary actors.Validate each one.

Structure enables behavior,but does not guarantee correctness.

AI often proposes over-detailed class models.Prefer clarity over completeness.

Got an idea?

Use this space to add awesome interactivity. Include text, images, videos, tables, PDFs... even interactive questions! Premium tip: Get information on how your audience interacts with your creation:

  • Visit the Analytics settings;
  • Activate user tracking;
  • Let the communication flow!

Got an idea?

Use this space to add awesome interactivity. Include text, images, videos, tables, PDFs... even interactive questions! Premium tip: Get information on how your audience interacts with your creation:

  • Visit the Analytics settings;
  • Activate user tracking;
  • Let the communication flow!

A1: Account balance is never negativeA2: Withdrawal allowed if balance β‰₯ amount A3: Concurrent withdrawals are supported A4: Transactions are eventually consistent A5: Overdraft is not allowed

β€œAI is used as a learning catalyst β€” not as an authority.”

If a rule is not explicit,it can be violated. Invariants define what must always hold.

Focus on WHO interacts with the system, not how it is implemented.

AI often models the happy path only.Check error and alternative flows.

Correct structure does not guarantee correct behavior.Look at the order of interactions.

Got an idea?

Use this space to add awesome interactivity. Include text, images, videos, tables, PDFs... even interactive questions! Premium tip: Get information on how your audience interacts with your creation:

  • Visit the Analytics settings;
  • Activate user tracking;
  • Let the communication flow!

AI often omits forbidden transitions.Explicitly define what must never happen.