← POETICA.STUDIO WORKGHOST NETSABOUT
SecureGuide

Compliance isn’t the brake.
It’s the steering.

CONCEIVED AT THE AI SUMMER SCHOOL, SAN FRANCISCO — AUGUST 10–14, 2026
A PROGRAM OF SMARTSTART (WORLD ECONOMIC FORUM) & THE DO

open the live prototype →
THE PROBLEM

In defense, aerospace, medical and finance, the question that stalls innovation isn’t can AI do this? It’s am I allowed to? Policy is ambiguous, the penalties are personal, and a compliance answer takes days.

So capable employees do one of two things: nothing, or something quietly risky. Both cost the company — one in speed, the other in exposure.

“I’d like to, but can’t.”

Elena — our persona, a project manager turned team lead at an aerospace company — said it in five words.

WHAT SECUREGUIDE DOES

A personal AI conformance copilot that answers “am I allowed to do this?” in seconds, not days — every verdict cited to a named, signed rule, and every no paired with a compliant path forward.

✓ Green — proceed

The rule that permits it is cited by name.

? Amber — conditions

Proceed, and the copilot shows exactly which conditions apply.

× Red — stop

And here is the nearest compliant alternative.

Ask → Assess → Act

The product promise in one move: turn No into How.

THE TRUST MODEL
01 The AI never grants permission Humans sign the rules in advance. The copilot only applies what a named person has already signed — it cannot invent a verdict, and it never guesses.
02 Two tiers of asking Anonymous questions, so people learn without fear. Identified escalations, only when a real decision needs a human signature.
03 Export control is hard-gated ITAR and EAR questions are never resolved by an automated verdict. They route to a person, every time. The refusal is the feature.
04 Gaps become policy Every unanswerable question is logged as a policy gap — the backlog that tells governance where to write next.
05 An audit trail by default Every ask, verdict and cited rule is recorded. Compliance evidence generated as a side effect of daily work, not a separate project.
WHY IT WORKS

Company-level AI governance exists. Nobody built it for the person.

SecureGuide moves compliance from a gate you queue at to infrastructure you stand on. The employee gets certainty. The company gets speed with an audit trail. The policy team gets a live map of where its rules fall short.

DEMONSTRATED WITH A WORKING PROTOTYPE · A FICTIONAL AEROSPACE & DEFENSE COMPANY
3
PERSONAS
9
POLICIES
12
SIGNED RULES
1
LIVE RULE-SIGNING
POLICY-GAP
BACKLOG
LIVE, FROM UNANSWERED ASKS
THE PEOPLE WHO MADE IT
Saba Khalid, narrative and conversational design lead on SecureGuide
01
Saba Khalid
Concept & pitch production
Christian Wanzek
02
Christian Wanzek
Launch-vehicle & rocket technology
Vivian Burda
03
Vivian Burda
Software architect — Airbus
Carol (Yujing) Jin
04
Carol (Yujing) Jin
Policy research & the governance case

The demo company, its policies, officers and programmes are fictional. Nothing here is drawn from a real client.

THE FILE · WHAT I OWNED, WHAT THE MACHINE DID, WHAT SHIPPED
PROBLEMin regulated industries the blocker isn’t “can AI do this” — it’s “am i allowed?”
MY ROLEconcept and pitch production, one of four founders.
THE WORKa conformance copilot: green / amber / red verdicts, each citing a rule a named human signed.
AI’S PARTdrafts the verdict; never grants permission. export-control questions always route to a person.
STATUSworking prototype, live in the browser — built at the WEF SmartStart summer school, SF 2026.
poetica— SABA KHALID · saba@poetica.studio