The employees of the future must develop strategies to leverage Generative AI, benefiting their organizations while maintaining high safety standards.
In this guide, we outline best practices for using AI in professional settings such as companies and educational institutions. We will teach you how to use ChatGPT more effectively and present strategies that a human can manually apply using the chat interface. However, they can also be automated through a programming approach.
How does AI impact the job market and employee productivity?
AI is transforming the job market and boosting employee productivity. A McKinsey report predicts that 60-70% of jobs could be automated through AI, potentially displacing up to 800 million individuals by 2030.
Harvard research (Dell'Acqua et al., 2023) shows AI's positive impact on productivity. Consultants using AI completed 12% more work in 25% less time than those without AI, with performance improvements of 43% for below-average and 17% for above-average performers. These findings demonstrate that AI can significantly enhance workplace performance.
Meanwhile, the concept of "a jagged technological frontier" exposes a crucial challenge. While AI excels at some tasks, it struggles with others that seem similar in difficulty. For tasks beyond AI’s current capabilities, consultants using AI were 19 percentage points less likely to produce correct solutions.
By understanding AI's strengths and limitations, we can maximize both technological and human potential.
AI's impact is already happening, and preparing the workforce for this evolution is key to leveraging its benefits in education and business.
Models of human-AI collaboration: centaurs and cyborgs
The Harvard University article introduces two models for human and AI collaboration: centaurs and cyborgs. These concepts represent different approaches to integrating human work with AI, significantly improving efficiency and innovation.
Centaur model
Centaurs divide tasks between humans and AI based on specific abilities. Users determine which tasks AI can best perform and which require human intervention.
In our view, centaur-style cooperation involves humans and AI complementing each other. AI supports humans by automating certain tasks, allowing more efficient project execution. When humans and AI complement each other, they bring unique skills, enhancing collaboration.
Here's an example of a centaur practice: in a content creation team, AI generates initial drafts and performs grammar checks, while humans focus on creative writing and final edits. This collaboration leverages AI’s speed and human creativity.
Cyborg model
Cyborgs represent a more advanced collaboration model where humans and AI are tightly integrated into every aspect of a task. This continuous and harmonious collaboration maximizes the strengths of both parties, creating a synergy that combines AI and human capabilities.
To put it in context, imagine a different content creation team where AI and humans work seamlessly together throughout the entire writing process. AI suggests topics, drafts content, and optimizes SEO in real-time, while human writers refine ideas, add creativity, and ensure the content aligns with brand voice and strategy. This tight integration enhances the quality and efficiency of the content creation process.
Which one is more efficient?
Centaurs and cyborgs offer different approaches to human-AI collaboration. Centaurs use AI to support and automate specific tasks while retaining human control. Cyborgs integrate AI into every aspect of work, creating a close symbiosis of human and artificial intelligence.
These approaches are crucial for maximizing efficiency and innovation across industries. They enable companies and individuals to leverage AI's potential, such as ChatGPT fully. By understanding these models, we can better navigate the evolving landscape of AI and human collaboration.
However, we believe that cyborg practices benefit companies and educational institutions more. In the next chapter, we’ll explore leveraging this strategy to use AI efficiently.
How can you use cyborg practices efficiently?
Assigning a persona
Large Language Models (LLMs) like ChatGPT can assume various roles, making them versatile tools. Each word we input guides AI in a specific direction, refining its responses based on the assigned persona.
Clearly defined roles help maximize AI's effectiveness in education, business, or other fields. By assigning personas, we can tailor AI interactions to achieve specific outcomes, ensuring that the technology supports our objectives in a more impactful way.
For example, if we assign ChatGPT the role of Marie Skłodowska-Curie, the interaction becomes educational – AI provides scientific insights from the perspective of a Nobel Prize winner. Conversely, if ChatGPT adopts a Loony Tunes character to teach kids math, it delivers quirky yet accurate guidance.
This approach makes learning more engaging and fun, helping students grasp complex subjects through memorable interactions.
In the workplace, assigning roles to AI can enhance productivity and innovation. ChatGPT can act as a marketing specialist, IT expert, cybersecurity consultant, or startup mentor. By defining specific roles, we can align AI's capabilities with our objectives, optimizing its utility.
Requesting editorial changes
Requesting editorial changes involves asking AI to refine generated content to meet higher quality standards. This process helps tailor content to specific needs and goals.
We suggest creating precise prompts to generate satisfactory results. Define the style, tone, and target audience clearly.
This clarity ensures the AI understands your expectations.
At the same time, think of editing as an ongoing interaction with ChatGPT. Adjust your prompts based on the generated content. For instance, you can assign a different role to the AI, refine the objectives further, or ask ChatGPT for specific improvements.
Additionally, you can ask ChatGPT for guidance on how to improve the prompt. This dialogue helps refine your requests, ensuring the AI provides the best possible output. You can leverage AI to produce high-quality, tailored content by continually adjusting and specifying your needs.
Teaching through examples
By providing concrete, well-executed examples, you can teach the model the desired style and structure of an answer. Essentially, we show the AI a good answer before asking it to generate a new one.
Highlight why these examples are excellent — perhaps due to their structure or tone. Then, ask the AI to produce a similar report. This method helps the AI recognize key elements such as introduction, development, conclusion, and appropriate style.
In our experience, this approach saves significant time when making editorial corrections. It allows AI to learn from quality examples and apply that learning to new tasks.
Modularizing tasks
Task modularization is a highly effective strategy in AI projects. It involves breaking down projects into smaller, more manageable parts. This approach allows AI, like ChatGPT, to focus on individual modules. Users benefit by evaluating each part of the task, iteratively improving the quality of the results.
This method is particularly useful for Project Managers (PMs) who use ChatGPT to automate and optimize processes.
We believe AI's capabilities can revolutionize project management, since it can streamline processes, reduce human error, and complete tasks efficiently. By leveraging AI in a structured way, PMs can focus on strategic decisions, leading to better project outcomes.
For example, ChatGPT can create schedules, assign tasks, send deadline reminders, and generate progress reports. It can even act as a virtual assistant (just like ChatGPT 4-o), especially in its mobile app version, which supports modalities such as voice.
Role Assignment
Combining this strategy with other practices mentioned in the article is recommended. For example, an important aspect of this process is assigning a persona to ChatGPT, which we have already explored above.
Start with asking ChatGPT to act as an experienced PM for a specific project type to ensure its advice and assistance are relevant.
This persona assignment helps the AI to deliver contextually appropriate and expert-level support, enhancing the overall project management process.
Applying task modularization
Let's use the example of creating a project backlog. By applying this practice, a PM should divide the backlog creation into smaller parts:
1️⃣ Identification of tasks. Start with a list of requirements. ChatGPT can help identify and list tasks based on these requirements.
2️⃣ Creating user stories. ChatGPT can draft clear and concise user stories, ensuring they are aligned with the project's goals.
3️⃣ Estimating tasks. ChatGPT can analyze past data to provide accurate estimates. Use historical project data to estimate the effort required for each task.
4️⃣ Assigning tasks. Allocate tasks to team members, considering their availability and skills. ChatGPT can suggest optimal task assignments to balance workloads and leverage team strengths.
Each step should be verified by a human to ensure relevance and accuracy.
Validation and verification of sources
Validating and verifying sources begins with critical thinking.
Never take AI-generated answers for granted. Remember, AI can sometimes produce incorrect information, a phenomenon known as a "hallucination”.
A good practice is to ask ChatGPT for specific sources of information. Make your prompts precise, asking for specific pages of reports or subpages of websites, not just the main links. Prioritize reliable sources such as Google Scholar or reports from reputable companies.
Then, check these sources manually. Be aware that AI can sometimes provide outdated links, so always verify through a search engine like Google.
Manual checking is essential to avoid spreading false information.
This is particularly important for maintaining brand reputation. For example, a New Zealand supermarket used AI to generate meal plans. The AI suggested recipes for harmful items like chlorine gas and "poison bread sandwiches" (Forbes).
In our experience, consistently verifying sources helps maintain accuracy and trustworthiness.
Elaborating and requesting a deep dive
To get the most out of tools like ChatGPT, it's essential to use elaboration and deep dive practices. These methods help obtain detailed, accurate, and valuable information.
Elaborating involves asking AI to add more information to already generated content. This is useful when the initial answer is too general. For example, if ChatGPT gives a basic sales forecast, you can request more detail by typing:
This prompt encourages AI to refine its answer, incorporating specific details that make the information more practical and actionable.
Deep diving means asking AI to analyze a well-defined topic thoroughly. This approach is beneficial when you need a comprehensive examination of a complex issue. For instance:
This method ensures the AI provides a detailed topic exploration, integrating specific data to enhance the analysis. Both practices aim to enrich the information provided, making it more useful for decision-making and strategic planning.
In our experience, using elaboration and deep dive requests with AI leads to more insightful and valuable outputs. It transforms generic responses into detailed, context-rich information that better supports your needs.
These practices maximize the effectiveness of AI tools like ChatGPT by focusing on specific data and clearly defined tasks.
Adding your own data
Adding your own data to AI-generated content enhances the accuracy and relevance of the output. This iterative process refines results by incorporating new information from the user.
For example, imagine you use ChatGPT to create financial forecasts for your company. AI generates a forecast based on your provided reports in the first iteration. However, you notice it lacks the latest quarterly results, which could impact the forecast significantly. You then add this new data and ask AI to re-analyze. ChatGPT processes the updated information and generates a more accurate forecast, reflecting the company’s current financial situation.
Here's how it works:
1️⃣ Initial analysis. AI generates an initial response based on the provided data.
2️⃣ Data addition. You provide additional data not included in the initial set.
3️⃣ Re-analysis. AI incorporates the new data and updates the response accordingly.
Adding new data iteratively allows AI to refine its outputs. This method ensures the final product is as accurate and comprehensive as possible.
However, please note that ChatGPT is a pre-trained model that learns from interactions. We recommend being cautious about data security - it's always a good practice. For example, you can turn the training mode off or use the API (like OpenAI playground). This will ensure the ChatGPT isn't trained using your conversations with the model.
Demanding a logical explanation, exposing contradictions, and disagreeing with the output
When AI generates a recommendation that seems unclear or incorrect, it's essential to address it directly. According to research from Harvard University, users should point out inaccuracies, ask for clarifications, or reject results.
Example #1. Pointing out an incorrect answer and asking for a logical explanation
Example #2. Rejection of the output
Consistently requesting explanations and rejecting incorrect answers helps maintain high information quality standards. It ensures that AI outputs are accurate and reliable, particularly in business contexts.
This practice enhances decision quality and helps identify potential errors, leading to more responsible and effective AI use.
It's important to remember that ChatGPT allows you to evaluate answers. Mark inappropriate answers with a downward thumb. If a chat contains offensive content (e.g., discrimination, violence, hate speech), report it immediately.
Does ChatGPT remember your conversations?
Before, each new session with ChatGPT was independent and did not carry over information from previous conversations. For example, if you provided historical data in an earlier conversation, ChatGPT wouldn't have access to it in subsequent sessions.
Now, ChatGPT is equipped with both context and memory. If you ask it to remember something, it will recall it in future conversations. Additionally, it has the ability to autonomously remember certain information (i.e. you prefer bullet points to full paragraphs of text).
There's also a separate panel to manage what ChatGPT remembers about you. You can also turn the memory off in the settings.
How to ensure security and privacy when using ChatGPT?
Follow these guidelines to protect your data:
Use official sources
Access ChatGPT through the web app link. Download OpenAI apps from Google Play and the App Store (beware of the fakes - there are a lot of them!). Use strong passwords to minimize the risk of hacking. Regularly delete chat history to reduce potential impacts from breaches.
Turn off model training
Chat GPT is a pre-trained model, meaning it learns continuously from user interactions and data. Many ChatGPT beginners are unaware they can disable model training on your data in the settings.
Anonymize data
Even with model training off, avoid entering personal and sensitive data. Remove unique identifiers like names, addresses, and identification numbers. Use pseudonyms, which you can later replace with actual information manually.
Ensure files and data do not contain sensitive information before uploading them to ChatGPT.
Never provide passwords, credit card numbers, or other sensitive details in chat.
Read Privacy Policies
Always read the privacy policies of AI tools. For example, ChatGPT anonymizes data and doesn’t sell user data to third parties. Choose AI tools that comply with legal standards, and be aware of your rights and responsibilities concerning personal data.
Please note that your country or region may have specific AI regulations. For example, on May 21, 2024, the European Union drafted the final version of the AI Act - it will come into force 20 days after the announcement.
Avoid illegal use
Do not use ChatGPT for illegal purposes, such as creating violent, racist materials or promoting banned political movements. The platform may delete your account for such activities.
Account deletion
Deleting your OpenAI account will prevent access to services like ChatGPT, API, and DALL-E. Once deleted, you cannot create a new account with the same email address. OpenAI will delete your data within 30 days, retaining only a limited set of data where required by law.
Conclusion
In this guide, we explored various advanced strategies to enhance the use of AI tools like ChatGPT. These strategies help refine AI outputs, ensure data security, and maximize the utility of AI in professional settings.
But did you know only 21% of companies surveyed had policies in place to regulate employee use of AI technologies? If you are a Manager or CEO, we encourage you to establish clear guidelines for AI usage in your team or organization.
Do you want to learn even more about implementing AI safely and cost-efficiently? Tune in to our podcast, Product Odyssey. Alternatively, consider booking a free consultation with us to get expert guidance for your business.
- 1Image source: https://openai.com/index/memory-and-new-controls-for-chatgpt/
- 2Image source: https://bgr.com/tech/how-to-stop-openai-from-using-your-chatgpt-chats-to-train-ai-in-just-3-minutes/
category
Vazco
We build digital product fundamentally better, helping startups, scale-ups and enterprises multiply efficiency.
We specialize in innovative projects and build them fundamentally better. We help startups, scaleups and enterprises multiply efficiency by combining technological excellence with business experience. We specialize in JavaScript tech stack (React, React Native, Node, TypeScript, GraphQL), including niche technologies like Three.js. Moreover, we provide teams and processes focused on delivering value.