WEBVTT

00:00:01.000 --> 00:00:06.010
This explanation accompanies our article on ChatGPT for business.

00:00:06.440 --> 00:00:10.330
How do individual prompts become a shared practice that colleagues can

00:00:10.360 --> 00:00:13.250
use, review and improve together?

00:00:13.720 --> 00:00:18.090
A fictional service team summarises the same working instruction.

00:00:18.120 --> 00:00:20.810
Each colleague takes a different approach.

00:00:20.880 --> 00:00:25.970
One summary loses an important exception, although the wording sounds convincing.

00:00:26.360 --> 00:00:30.730
The team selects one task: summarising an instruction for internal

00:00:30.760 --> 00:00:31.530
use.

00:00:31.800 --> 00:00:36.255
It defines the source, expected structure, boundaries and the point at

00:00:36.280 --> 00:00:38.490
which someone reviews the result.

00:00:38.920 --> 00:00:43.610
First agree which information people may use, who manages access and

00:00:43.680 --> 00:00:45.130
where results can be shared.

00:00:45.720 --> 00:00:50.090
Check the features actually available in your workspace and configuration.

00:00:50.520 --> 00:00:54.570
The instruction requests main steps and exceptions using only the

00:00:54.600 --> 00:00:55.930
supplied text.

00:00:56.200 --> 00:00:58.810
Missing information must remain visible.

00:00:59.080 --> 00:01:02.730
The output is a draft, rather than an approved instruction.

00:01:03.560 --> 00:01:06.970
A colleague checks the draft against the original instruction.

00:01:07.240 --> 00:01:08.970
Are conditions preserved?

00:01:09.000 --> 00:01:10.250
Has anything been added?

00:01:10.840 --> 00:01:15.050
A missing exception is corrected before anyone uses the summary elsewhere.

00:01:15.880 --> 00:01:18.970
Ask several colleagues to work through the same examples.

00:01:19.400 --> 00:01:22.930
Record errors, questions and correction effort.

00:01:23.240 --> 00:01:27.290
Polished wording or one successful example does not establish

00:01:27.320 --> 00:01:28.970
usefulness across the team.

00:01:29.800 --> 00:01:33.370
Choose an initial user group and an owner for instructions and

00:01:33.400 --> 00:01:34.090
changes.

00:01:34.680 --> 00:01:39.290
After the trial, decide which tasks continue, need adjustment or

00:01:39.440 --> 00:01:41.450
remain outside the agreed practice.

00:01:42.200 --> 00:01:46.170
Start with working agreements, rather than a collection of prompts.

00:01:46.600 --> 00:01:51.690
Our source article on ChatGPT for business explains how configuration,

00:01:51.720 --> 00:01:54.890
everyday use and evaluation fit together.