WEBVTT

1
00:00:00.000 --> 00:00:08.812
FusionWall is designed around a simple idea: an
important action should not disappear between a
button click and an AI provider.

2
00:00:08.812 --> 00:00:13.848
The approval checkpoint shows you the exact
protected request before it leaves.

3
00:00:13.848 --> 00:00:16.785
The private receipt records what happened
afterward.

4
00:00:16.785 --> 00:00:23.080
Let me show you how to read both without needing
to be a security engineer.

5
00:00:24.680 --> 00:00:30.322
At the checkpoint, start with the basics. Is
this the question you meant to ask?

6
00:00:30.322 --> 00:00:32.203
Is the selected context appropriate?

7
00:00:32.203 --> 00:00:36.716
Is it going to the provider and model you
intended to use?

8
00:00:36.716 --> 00:00:39.726
Then read the policy result and any warnings.

9
00:00:39.726 --> 00:00:46.120
The screen is there to support a real decision,
not to become a form you click through.

10
00:00:47.720 --> 00:00:49.660
Now read the protected request itself.

11
00:00:49.660 --> 00:00:53.862
Confirm that names, account details, and custom
terms have been replaced where expected.

12
00:00:53.862 --> 00:00:58.064
Make sure the remaining text still contains
enough meaning for a useful answer.

13
00:00:58.064 --> 00:01:01.297
Detection is not perfect, so stop if anything
private remains.

14
00:01:01.297 --> 00:01:03.560
Canceling is a normal and safe outcome.

15
00:01:05.160 --> 00:01:10.700
You have three reasonable choices. Send it when
the protected view is correct.

16
00:01:10.700 --> 00:01:14.535
Revise it when the question or context needs
work.

17
00:01:14.535 --> 00:01:18.796
Cancel it when the request should not leave at
all.

18
00:01:18.796 --> 00:01:22.631
The approval applies to this visible action and
route.

19
00:01:22.631 --> 00:01:26.040
It is not blanket permission for future
questions.

20
00:01:27.640 --> 00:01:30.621
After the answer, open the linked receipt.

21
00:01:30.621 --> 00:01:38.287
You can see the request fingerprint, rules that
applied, chosen model, approval, status, and
connections to related records.

22
00:01:38.287 --> 00:01:42.972
The ledger is tamper-evident, which means
unexpected changes can be detected.

23
00:01:42.972 --> 00:01:49.360
It is evidence of the workflow, not a
certification that the AI answer is correct.

24
00:01:50.960 --> 00:01:55.504
If a result surprises you later, start with the
receipt.

25
00:01:55.504 --> 00:02:01.411
Compare it with the conversation, selected
source boundary, active policy, and recorded
result.

26
00:02:01.411 --> 00:02:07.318
That gives a founder, contractor, operator, or
auditor a common set of facts.

27
00:02:07.318 --> 00:02:13.680
You can correct the work without relying on
somebody's memory of what they clicked.

28
00:02:15.280 --> 00:02:20.708
A useful approval is specific enough to
understand and limited to one action.

29
00:02:20.708 --> 00:02:24.465
A useful receipt connects that decision to the
outcome.

30
00:02:24.465 --> 00:02:31.145
Read the request, choose deliberately, and use
the proof when you need to explain what
happened.

31
00:02:31.145 --> 00:02:35.320
That is how human authority stays real instead
of ceremonial.