1
00:00:00,000 --> 00:00:04,100
Why use Jev, an AI that doesn't write?

2
00:00:04,280 --> 00:00:07,520
Jev is TypeSafe's decision model.

3
00:00:07,520 --> 00:00:11,240
Give Jev text, a question, and a list of answers.

4
00:00:11,240 --> 00:00:14,660
It picks one and gives probabilities: numbers that

5
00:00:14,660 --> 00:00:17,320
show how likely it thinks each answer is.

6
00:00:17,500 --> 00:00:20,660
The customer says, I paid twice.

7
00:00:20,660 --> 00:00:22,830
This server code asks who should

8
00:00:22,830 --> 00:00:27,020
help: payments, tech support, or someone else.

9
00:00:27,200 --> 00:00:31,960
So if it picks payments, do we just trust it?

10
00:00:32,140 --> 00:00:32,420
No.

11
00:00:32,420 --> 00:00:34,175
These are demo numbers.

12
00:00:34,175 --> 00:00:37,800
Point nine six means ninety-six percent.

13
00:00:37,800 --> 00:00:40,360
We chose ninety percent as our limit.

14
00:00:40,360 --> 00:00:44,060
Below it, or for other, our code asks a person.

15
00:00:44,240 --> 00:00:47,120
Jev answers questions together when one

16
00:00:47,120 --> 00:00:49,080
does not need the other's answer.

17
00:00:49,080 --> 00:00:52,120
It does not write a reply piece by piece.

18
00:00:52,120 --> 00:00:55,200
This can save time, but check the speed yourself.

19
00:00:55,380 --> 00:00:58,720
A chat model can choose from a list too.

20
00:00:58,720 --> 00:01:01,280
Use one if you also need text.

21
00:01:01,280 --> 00:01:05,040
Test your choice and your limit on real messages.

22
00:01:05,040 --> 00:01:07,860
An allowed answer can still be wrong.

23
00:01:08,040 --> 00:01:11,760
So it can send my problem to the wrong team.

24
00:01:11,760 --> 00:01:12,760
Great.

25
00:01:12,760 --> 00:01:14,600
Now two teams can ignore me.