How to use GenAI as a Secretary

Day 1 of a hands-on, non-technical guide for executives

Start with the Introduction to series . This is Day 1.

Image generated with Gemini

Take a long piece of business writing and have the model whittle it down to one sentence. As an example, take this paragraph from a 1996 Federal Meeting:

With regard to the information available to the market, I
think there may be some steps we could take. I am somewhat
uncomfortable with the continued use of the customer RP, which is a
bit of an historical artifact. It used to reflect the pass-through of
the repo pool when the repo pool was only $2 or $3 billion or some
similarly small amount. Now the repo pool is $10 billion or more, and
we really do not pass it through to the market in any sense. The
customer RP also was used by my predecessors as the non-signaling
device, which likewise is no longer necessary, but it does have the
tradition of having the amount to be done publicly attached to it.
One thought I have, which I intend to propose when I come back to you
with an outline of the steps we have to go through to conduct our
operations earlier in the day, would be to ask for your guidance on my
desire to eliminate the use of the customer RP as we now employ it.
We would do only System Account RPs. Currently when we do a customer
RP, we announce that we might do, say, a total of up to $1.5 billion;
however, if we tell the market that we are doing RPs for System
Account we never announce how much we are doing.

1. Go to https://chat.openai.com/ and ask the model to rewrite this to be more clear. You might start with the instruction or prompt:

Clarify the following paragraph:

With regard to ...  how much we are doing.

Are you happy with replacing the paragraph with what the model produced? Obviously, you could just edit the model response to fit your needs, but restrict yourself from doing that. You want to learn how to write a prompt that will clarify a piece of dense text into something that you would put in an email.

2. This was my starting prompt (try it out):

I am an executive at a financial firm and would like to explain to
my team why we are considering a policy change.  Rephrase this paragraph
into a single crisp sentence:

With regard to ... how much we are doing.

Notice that I’m specifying a role (“executive at financial firm”), an audience (“my team”), style (“crisp”), and format (“single sentence”). A good LLM textual transformation prompt will have all four of these.

3. Continue experimenting and see if you can find the prompt that works for you on the above paragraph (you can use the Edit pencil-like button to change the prompt and try again).

4. Because these are chat interfaces that have “memory” in the course of a session, you can ask the model to further change the text that it created. For example, I issued the following follow-on instruction:

make the language a bit friendlier

and got a result I was quite happy with:

We're thinking about phasing out customer RPs, since they no longer
reflect how we operate today, and instead using only System Account RPs,
which are more in line with current market practices and give us greater
flexibility without the need for public disclosures.

This fits the way I tend to write emails, but may not be how you write. You should experiment to find what works for you. Perhaps you prefer several short sentences? Maybe you want football metaphors? Maybe you want it to be even more succinct?

5. Ask OpenAI to summarize all your experimentation into a prompt that works for you:

can you combine my prompts above into a single prompt i can use next time?

I now got the following prompt:

I’m an executive at a financial firm and would like to rephrase a
technical or policy-related paragraph into a single, crisp, and friendly
sentence that explains to my team why we’re considering a change.
Please summarize the paragraph clearly, maintain the key reasoning,
and make the tone approachable but professional.

This is the prompt I’d want to use anytime I get an unclear, meandering paragraph that I want to distill into a clear, actionable email.

6. Take a passage (use non-confidential data, unless you are using the paid/enterprise version) from your own domain and see if you can use a prompt to create the different kinds of emails you send — instructions to staff, questions about KPIs, motivational messages, etc. Make sure to start a new session in ChatGPT (use the “New Chat” button the left).

Recommended extensions for Day 1:

7. Repeat this exercise in https://gemini.google.com/ and/or https://claude.ai/. I recommend you do this so that you start to get a feel for what is true of GenAI as a whole vs. what is specific to the model you are using.

8. Load an entire document into Gemini and ask it to summarize the document or URL into a paragraph that avoids jargon and is understandable to people outside the core role. For example, distill the report on the collapse of Enron into an ethics code (see below):

Gemini used to distill a complex report into a 10-point code of ethics

9. As before, try it on a document from your own domain.

10. Save the prompt(s) so that you can use it the next time. You can do this in Gemini by creating a “Gem” (see the menu on the left). On OpenAI, this stored/shareable prompts are called GPTs, but you need the “plus” version for this.

Extra credit work for Day 1:

An executive I know sends out a weekly “Friday Thoughts” to his team. It might be a memorable customer interaction, a great demo they saw, or something that got them inspired or upset. See if you can dump a long email thread on the thing you want to highlight and create a message that you can send to your staff.

Hint: Include 2–3 examples of previous such messages handwritten by you in your prompt. You might be surprised at how well the LLM can capture your voice.

Great! Take a break before you start Day 2.

Links to Full Series

Day 0: Introduction to series

1. Day 1: Use GenAI as a secretary (this page)
2. Day 2: Use GenAI as an executive assistant
3. Day 3: Use GenAI as a librarian
4. Day 4: Use GenAI as an analyst
5. Day 5: Use GenAI as a researcher
6. Day 6: Use GenAI as a sparring partner
7. Day 7: Use GenAI as an engineer