AI assistants like ChatGPT and Claude AI can help automate tasks and save time. However, there are two common problems when using them - maintaining privacy when sharing confidential data, and sharing prompts across teams. This post summarizes a tutorial on how to use these AI tools while addressing both issues.
What is a "Prompt" anyway?
A prompt is the text that a user provides to an AI system to guide it towards a desired response. The prompt acts as the input that shapes the AI's output. For chatbots like Claude.AI and ChatGPT, prompts are the messages that users send to initiate a conversation and direct how I should respond.
Prompt engineering is the practice of carefully crafting prompts to get better results from an AI system. Prompt engineers experiment with different phrases, styles, and formulations to coax the AI into producing more coherent, relevant, and helpful responses.
A poorly constructed prompt can lead to nonsensical or unhelpful AI responses. Here's an example of a bad prompt for asking an AI to generate a marketing email:
"Write me an email to send to customers about our new product offering."
This prompt lacks key details the AI needs to create a coherent email, like what the product is, who the customers are, or any messaging for the email. The AI is likely to struggle generating a sensible email from this prompt.
A more concise, detailed prompt would be:
"Act as a content marketer for a pet food company. Write a friendly, conversational 150-word email to existing customers announcing a new line of organic dog treats. Use the PAS framework - Problem, Agitate, Solve. First, explain the Problem that many dog treats use artificial ingredients. Then, Agitate by noting how those ingredients can be unhealthy for dogs long-term. Finally, present our new line of organic treats as the Solution. Describe the treats in 1-2 sentences, emphasizing their natural ingredients. Offer a 15% discount code for customers to try the new treats. Use an enthusiastic, customer-focused tone. Mention how our company values pet health and nutrition. Close by recapitulating the discount code."
This prompt provides clear guidance on the context, length, messaging and call-to-action the AI needs to generate a marketing email that fulfills the desired goal. Prompt engineering is the art of guiding the AI with just the right amount of detail - not too vague but not overly prescriptive. Crafting precise, well-structured prompts takes experimentation but leads to much better results from AI systems.
What are the 2 big problems business face when using Chatbots like ChatGPT and Claude.AI?
Artificial Intelligence Chatbots are indeed remarkable, yet numerous enterprises have already banned staff from using them. Even in instances where their deployment is permitted, these businesses still grapple with the intricacies of formulating a consistent approach to prompting them.
Let's look in more detail:
Problem 1: Free Chatbots generally train on your private data
When you use ChatGPT, it trains on the data you provide, potentially exposing confidential information to people outside your business. Alternatively, you could consider using OpenAI API (the same company behind ChatGPT) which doesn't train from your inputs and outputs. OpenAI takes privacy very seriously with GDPR, CCPA and HIPAA compliance. Using the API requires an account and key but allows utilizing ChatGPT privately.
Alternatively, Microsoft have also deployed a separate instance of OpenAI into Azure, which is also another viable approach if you're worried about data privacy.
The point here, if you're worried about privacy then there are solutions out there.
Problem 2: It's hard to reuse and share Prompts across your team
When first using chatbots, it becomes clear that thoughtfully crafted prompts are key to getting good responses from the AI. However, it's tedious to type out detailed prompts from scratch every time you want to ask the chatbot something. So you end up storing your best prompts in documents to reuse them.
The problem is, these prompts are trapped in silos on your personal device. It would be much better if you could easily share your best prompts with teammates to enable consistency.
One simple solution is to create a shared repository of prompts that everyone can access. For example, you could store prompts in a SharePoint list. Then integrate that list into chatbot interfaces like Power Apps so users can select from preset prompts to prepend to their queries.
This makes it easy to reuse your best prompts and aligns everyone on how to properly frame questions to the AI. The chatbot gets consistently clear, detailed prompts that lead to better responses. And users save time by not re-inventing prompt wording. A shared prompt repository unlocks the power of your best prompts across your team.
The video below demonstrates how to build a Power App that enables easy reuse and sharing of prompts with a chatbot. By integrating a SharePoint list of prompts, the app allows you to select predefined prompts to prepend to your queries. This eliminates the need to manually copy and paste prompts from documents each time.
With this Power App, you get the best of both worlds - the ability to have private conversations with the chatbot while also leveraging shared prompts created by your team. The end result is a more streamlined and consistent way to interact with your chatbot. Structured prompts lead to better responses from the AI. And not having to reinvent prompt wording saves you time while aligning your team's approach.
Watch the video for a step-by-step walkthrough of constructing this handy Power App for managing and integrating prompts with your chatbot. You'll gain a better technique for prompt reuse that makes chatting with your AI assistant smoother and more effective.
- Maintain privacy with AI assistants using the OpenAI API
- Share prompts centrally for re-use across teams
- Tools like Power Platform enable quick automation with AI
- The same process could be used to connect to Azure OpenAI as well.
This approach helps organizations tap into AI capabilities while addressing common challenges around privacy and sharing of prompts. With some thoughtful planning, AI can be used securely and consistently across business teams.