Harnessing Large Language Models

3 min read

First published February 13, 2025, and edited for style.

Large language models such as ChatGPT, developed by OpenAI, are among the most useful tools available for professional development. Used well, they help you work more effectively and learn faster.

A lot of professional time goes into looking things up, studying, and trying to remember what you once knew. These models answer questions, explain concepts and suggest ideas or hypotheses in seconds, drawing on what they were trained on.

You can use ChatGPT to explore a subject you know little about, to go deeper into your own field, or to prepare for a presentation. You can also use it to think a problem through before you take it to anyone else.

Like any tool, it works best in trained hands. A carpenter learns the hammer and the saw; the same applies here. That means knowing what the model can and cannot do, and learning to ask a precise question and read the answer critically.

For those interested in using ChatGPT for professional development, we suggest the following steps

Explore and Understand: Spend some time exploring what ChatGPT can do. Ask it questions, try out different types of prompts and see how it responds. Understand its limitations. At the time of writing, for example, it had been trained on a vast range of data but had no access to real-time or proprietary information.
Be Specific and Thoughtful: When asking ChatGPT for help or insights, be as specific as you can. If you're vague, the model might not provide the precise information you're looking for.
Experiment and Learn: Don't be afraid to experiment and learn from the process. Try using ChatGPT in different professional situations, such as preparing for meetings, brainstorming ideas or exploring new topics.
Reflect and Refine: After using ChatGPT, take some time to reflect. Did the model provide useful information? How can you refine your prompts or questions to get better results next time?

These models are a genuine step forward. The way to find out what they are worth to you is to keep using them on real work.

As Clayton Christensen once said, "It's easier to hold your principles 100 percent of the time than it is to hold them 98 percent of the time." Curiosity works the same way. Apply it all of the time, not most of it.

Use these tools to be better informed than you were. What you do with that is still your decision.

What has changed since this was written

This piece was first published in early 2025. It has been edited for style only, because the reasoning holds and the reasoning is the point. The capability assumptions are the part that has moved.

Read it for its advice on how to work with these tools rather than as a description of what they can currently do. Two things are worth checking against the present rather than taking from the text.

Assume the capability floor is higher than described

The piece notes that these tools had no access to real-time or proprietary information. Many now search the web or connect to an organization's own documents and systems. That widens what they can do, and it makes data governance part of using them.
Treat any limitation the piece mentions as something to re-test. Planning around a limitation that has since gone is as costly as ignoring one that remains.

The advice to be specific and to reflect has not dated

Being specific means supplying your goals, your context and your standards. The tool cannot supply those for you, however capable it becomes, so better capability raises the value of clear framing.
The habit of reflecting on output before using it matters more as quality improves, because confident output becomes easier to accept without checking.

Harnessing Large Language Models

3 min read

First published February 13, 2025, and edited for style.

Large language models such as ChatGPT, developed by OpenAI, are among the most useful tools available for professional development. Used well, they help you work more effectively and learn faster.

For those interested in using ChatGPT for professional development, we suggest the following steps

Explore and Understand: Spend some time exploring what ChatGPT can do. Ask it questions, try out different types of prompts and see how it responds. Understand its limitations. At the time of writing, for example, it had been trained on a vast range of data but had no access to real-time or proprietary information.

Be Specific and Thoughtful: When asking ChatGPT for help or insights, be as specific as you can. If you're vague, the model might not provide the precise information you're looking for.

Experiment and Learn: Don't be afraid to experiment and learn from the process. Try using ChatGPT in different professional situations, such as preparing for meetings, brainstorming ideas or exploring new topics.

Reflect and Refine: After using ChatGPT, take some time to reflect. Did the model provide useful information? How can you refine your prompts or questions to get better results next time?

These models are a genuine step forward. The way to find out what they are worth to you is to keep using them on real work.

Use these tools to be better informed than you were. What you do with that is still your decision.

What has changed since this was written

This piece was first published in early 2025. It has been edited for style only, because the reasoning holds and the reasoning is the point. The capability assumptions are the part that has moved.

Read it for its advice on how to work with these tools rather than as a description of what they can currently do. Two things are worth checking against the present rather than taking from the text.

Assume the capability floor is higher than described

The piece notes that these tools had no access to real-time or proprietary information. Many now search the web or connect to an organization's own documents and systems. That widens what they can do, and it makes data governance part of using them.

Treat any limitation the piece mentions as something to re-test. Planning around a limitation that has since gone is as costly as ignoring one that remains.

The advice to be specific and to reflect has not dated

Being specific means supplying your goals, your context and your standards. The tool cannot supply those for you, however capable it becomes, so better capability raises the value of clear framing.

The habit of reflecting on output before using it matters more as quality improves, because confident output becomes easier to accept without checking.