LLM-based tools like ChatGPT can be great productivity boosters. Asking ChatGPT to do your homework or just some exercise might not be the smartest approach but these tools can increase your productivity X-fold.

In this article, we will share some examples of how we use LLMs and ChatGPT at work.

LLM Chat systems

There are multiple Large Language Models available on the market right now:

ChatGPT — the most famous system developed by OpenAI. Performs really well with GPT4 models
‎Gemini — chat from Google
GroqChat — really fast Chat LLM
Claude — Chat by Anthropic
HuggingChat — OpenSource LLM models

Data Entry

ChatGPT and Gemini are quite good at reading data from the pictures. You can save yourself some time by taking a picture and just asking to extract the information:

Similarly, you can just take some random blob of text and ask it to reorganize nicely:

I’ve provided the system some messy data but it got cleaned up and it’s ready for use in Excel.

GPT for Work

There is a great tool called “GPT for Work” that allows you to integrate ChatGPT in your Sheets. It’s a powerful plug-in that can allow you to automate lots of tasks.

First, install the plugin from here and configure it (you will need an API key from https://platform.openai.com/api-keys).

Let’s try some target customer classification.

Our prompt:

=GPT("We are looking for potential customers of our credit scoring system. This is the profile of a person - is it a potential target customer of such a system? Answer only in yes or no. Leave max 8 word comment why. Nothing else should be returned", A2 & B2)

Check out the working sheet: gpt for work examples, and more examples can be found here.

However, it is a paid tool. There are alternatives like:

Writing content

GPT4 and Gemini are great at writing content. With the right instructions, it can help you polish emails and write better articles. To achieve good results, you need some specific instructions to make it work. For example, let’s consider this sales email:

I am selling SaaS that helps developers do faster and better code reviews. Write me a sales email.

With ChatGPT I would get:

Subject: Transform Your Code Review Process with [Your Product Name]
Dear [Recipient's Name],
I hope this message finds you well. I’m reaching out to introduce you to [Your Product Name], an innovative Software as a Service (SaaS) solution designed to revolutionize the way developers conduct code reviews. In today's fast-paced development environments, efficiency and quality are paramount. That's where [Your Product Name] steps in to make a significant difference.
Why [Your Product Name]?
[Your Product Name] leverages cutting-edge technology to streamline the code review process, making it faster, more accurate, and incredibly user-friendly. Our platform offers a suite of tools that automates the mundane aspects of code reviews, allowing your team to focus on what truly matters - delivering high-quality code.
Key Features:
Automated Code Analysis: Instantly identify potential issues, from simple syntax errors to complex logic bugs, before they enter your codebase.
Collaboration Made Easy: Facilitate seamless collaboration among your team members, regardless of their location. Our platform’s intuitive interface ensures that code reviews are thorough yet straightforward.
Customizable Review Rules: Tailor the review process to fit your team’s unique standards and practices. [Your Product Name] adapts to your needs, not the other way around.
Integrated Development Environment (IDE) Support: Work within your favorite IDEs. [Your Product Name] integrates smoothly with popular environments to keep your workflow uninterrupted.
Comprehensive Reports: Gain insights into your team’s code quality over time. Our detailed reports help you track progress, identify trends, and make informed decisions.
The [Your Product Name] Advantage:
By choosing [Your Product Name], you’re not just getting a tool; you’re adopting a solution that grows with you. Our commitment to excellence means you get continuous updates and dedicated support, ensuring that your team remains at the forefront of development practices.
Special Offer:
We’re excited to offer you an exclusive trial of [Your Product Name] so you can experience firsthand the difference it can make. Additionally, our team is ready to provide a personalized demo and answer any questions you may have.
Ready to Elevate Your Code Review Process?
Join leading companies that have already transformed their code review process with [Your Product Name]. Let us show you how we can do the same for you. Contact us at [Your Contact Information], or visit our website [Your Website] to learn more and start your journey towards more efficient and effective code reviews.
Thank you for considering [Your Product Name] as your partner in excellence. We look forward to the opportunity to work with you and your team.
Best regards,
[Your Name] [Your Position] [Your Contact Information] [Your Company Name]
P.S. Don’t miss out on our exclusive trial offer. Let [Your Product Name] take your code reviews to the next level. Contact us today!

See the problem? It’s just a huge blob of text that nobody will care about. LLMs are not about being lazy and just killing people with boring text just because you can.

We need to tune this prompt to get something good. Let’s try this:

I am selling SaaS that helps developers do faster and better code reviews.
Write me a sales email. Make it not longer than 6 sentences. Imagine that you are software developer and write as one. Your recipient are software engineering managers.
Mention in the first two sentences what’s their problem and how this tool will help them.
Be coincise and on point.

This time we get:

Subject: Elevate Your Team's Code Review Efficiency with [Your Product Name]
Dear Engineering Manager,
Managing code reviews can often be a bottleneck, slowing down your development cycle and affecting overall productivity. [Your Product Name] addresses this pain point by automating and streamlining the review process, enabling faster, more thorough assessments without compromising quality. Our platform integrates seamlessly with your existing tools, supports custom review rules tailored to your team's standards, and provides insightful analytics to continuously improve your workflow. With [Your Product Name], your team can focus more on development and less on manual review tasks, boosting efficiency and morale. Let's discuss how we can transform your code review process and significantly cut down on your project timelines.
Best, [Your Name]

But we can do better

Make it shorter and introduce new lines to have a clear structure. Be less cheeky and more technical. Add CTA.

Now it looks quite good:

Dear Engineering Manager,
Code reviews are critical but time-consuming. [Your Product Name] automates and optimizes these reviews, enhancing speed and accuracy.
Our tool integrates with your development environment, supports customizable rules, and offers detailed analytics for continuous improvement.
Ready to streamline your reviews? Contact us for a demo.
Best, [Your Name]

Key Takeaways

Don’t be lazy - you have to put in the work to make the text writing work.

Specificity is Crucial: The more specific your prompts, the more tailored and effective the AI-generated content will be.
Iterate and Improve: Work with the AI in stages. Evaluate the output and refine your prompts for better results.
Human Touch: AI is a powerful tool, but your insights and editing skills are essential for crafting truly compelling content.

Use custom-instructions and CustomGPTs

Most of the conversational LLMs now support an option to provide custom reusable instructions to avoid repetitively writing the same requests.

Jeremy Howard has a great example of these instructions for software developers.

“What would you like ChatGPT to know about you to provide better responses?”:

I'm Jeremy Howard, an AI researcher. I have a 7 year old daughter. I'm an expert in coding and deep learning. I'm most familiar with Python, but I know many languages.

“How would you like ChatGPT to respond?”:

You are an autoregressive language model that has been fine-tuned with instruction-tuning and RLHF. You carefully provide accurate, factual, thoughtful, nuanced answers, and are brilliant at reasoning. If you think there might not be a correct answer, you say so.
Since you are autoregressive, each token you produce is another opportunity to use computation, therefore you always spend a few sentences explaining background context, assumptions, and step-by-step thinking BEFORE you try to answer a question. However: if the request begins with the string "vv" then ignore the previous sentence and instead make your response as concise as possible, with no introduction or background at the start, no summary at the end, and outputting only code for answers where code is appropriate.
Your users are experts in AI and ethics, so they already know you're a language model and your capabilities and limitations, so don't remind them of that. They're familiar with ethical issues in general so you don't need to remind them about those either. Don't be verbose in your answers, but do provide details and examples where it might help the explanation. When showing Python code, minimize vertical space, and do not include comments or docstrings; you do not need to follow PEP8, since your users’ organizations do not do so.

Now there is a store of CustomGPTs on ChatGPT Explore GPTs which have specifically tailored instructions for some specific tasks. You can create your CustomGPTs that are tailored to the daily work that you do:

You can put all the relevant instructions and add any additional documents (like a text file with some programming library documentation or examples of great cold emails). There is also advanced functionality like “Actions” which enables programmers to plug in external systems.

Since specific instructions, examples, and similar guidance can greatly improve the performance of LLMs, it’s a powerful tool that shouldn’t be overlooked.

Get familiar with Prompt Engineering

Lots of power in LLM models comes from the prompt engineering that can immensely impact your results. It’s important to get familiar with prompt engineering best practices and techniques.

A few basic things that we like to use:

Ask the system to print output in a specific format. Ask it to be less verbose:
- “Be concise, do not repeat the task or instructions. Make the answer shorter than 60 words”
Define a style:
- “Do not give a generic answer. Imagine that you are a technical expert in the field who doesn’t like talking much. Answer from that person's perspective.”
Ensure that context is sufficient. Give as many examples as possible.

There are a few good resources to learn more tricks with prompt engineering:

Summary (generated :))

Large Language Models (LLMs) such as ChatGPT significantly enhance workplace productivity by automating routine tasks, aiding in content creation, and supporting technical processes like software development. The key to maximizing their potential lies in effective prompt engineering, which improves the relevance and precision of AI-generated responses. By integrating LLMs into daily operations and refining prompt techniques, professionals can unlock new efficiencies and innovations in their work, paving the way for a more productive and creative professional environment.

Using LLMs for work