AI-Powered Email Management

Managing emails, especially for customer support or campaigns, can feel like juggling too many balls simultaneously. It's easy to get overwhelmed. But what if you could make that whole process smoother?

Managing emails, especially for customer support or campaigns, can feel like juggling too many balls simultaneously. It's easy to get overwhelmed. But what if you could make that whole process smoother?

We've worked with several clients to build AI-powered email management tools for various use cases: invoice follow-ups, marketing campaign analysis, customer support analytics and streamlining, team efficiency analysis, and more. This tutorial shows you how to get started.

Why AI-Powered Email Management Matters

In today’s fast-paced business environment, the ability to manage emails effectively is more than just a convenience—it's a competitive advantage. AI-powered tools not only help reduce the manual workload but also extract valuable insights from email data that can drive business decisions.

For instance, in invoice follow-ups, AI can automatically track unpaid invoices and send timely reminders, improving cash flow without the need for manual intervention. In marketing campaign analysis, AI can sift through responses to gauge customer sentiment, helping you refine your messaging in real time. Customer support analytics becomes more robust with AI, allowing teams to identify recurring issues, track resolution times, and improve overall customer satisfaction. Finally, by analyzing team efficiency, AI can highlight bottlenecks in communication and suggest areas for process improvement.

These use cases show how integrating AI into your email management can lead to not just operational efficiency but also strategic insights. It's about turning everyday email tasks into opportunities for growth and optimization.

Getting Started with AI-Powered Email Management

This guide demonstrates how to build an AI-powered email Management Tool that integrates Gmail with Zapier AI Actions and OpenAI's GPT-3.5. The result? A more efficient way to sort, respond to, and prioritize emails.

By the end of this tutorial, you'll have a comprehensive tool that simplifies email management using AI, with the potential to adapt it to various business needs.

Key Features

  • Natural Language Understanding: This tool gets what you're saying, letting you search through emails just by asking a question.
  • Email Summarization: It fetches details, gives you quick summaries, and tells you why certain emails are important, helping you decide what needs attention.
  • Automated, Polished Replies: Drafts well-structured replies using LlamaIndex and GPT-3.5, ready for review and sending.
  • Seamless Gmail Integration: Draft responses show up directly in your Gmail drafts, making it easy to review and send without leaving your inbox.
  • Intuitive User Experience: Built with Streamlit for simplicity, the interface guides you through managing emails with ease.

The Tech Behind It

  • OpenAI’s GPT-3.5: A powerful language model that drives the email summaries and replies.
  • Document Handling: Uses SimpleDirectoryReader to pull text from PDFs, which is crucial for processing and indexing documents.
  • LlamaIndex: Connects the language model with data, handling summarization and querying.
  • Streamlit: A straightforward tool for building user-friendly UIs.
  • Zapier Integration: Automates email retrieval and reply drafting, linking effortlessly with Gmail for a smooth workflow.

Why Zapier AI Actions?

Zapier AI Actions are powerful for managing integrations and automating workflows. They seamlessly connect Gmail with advanced AI features, making your email management process smooth and efficient.

The workflow begins by creating vector indexes from documents using LlamaIndex. This step not only organizes documents but also generates summaries, making your search results more relevant and context-aware.

A Look at the Code

Fetching Emails:

This function pulls emails based on a search prompt using Zapier AI Actions.

def get_emails(search_prompt):
    url = "https://actions.zapier.com/api/v2/ai-actions/01J06T8Q6TM4EVE9F3NYW1Z292/execute/"
    payload = {
        "instructions": search_prompt,
        "params_constraints": {},
        "preview_only": False
    }
    headers = {
        "X-API-Key": os.getenv('ZAPIER_NLA_API_KEY'),
        "Content-Type": "application/json"
    }

    response = requests.request("POST", url, json=payload, headers=headers)
    response_json = json.loads(response.text)
    return response_json["full_results"]
Crafting Email Responses:

This function drafts email replies using Zapier AI Actions.

def draft_email_reply(thread, body, to):
    url = "https://actions.zapier.com/api/v2/ai-actions/01J09PBZ62KMPFG55J4CFZC0MR/execute/"
    payload = {
        "instructions": 'Draft a reply with the given body',
        "Thread": thread,
        "Body": body,
        "To": to,
        "preview_only": False
    }
    headers = {
        "X-API-Key": os.getenv('ZAPIER_NLA_API_KEY'),
        "Content-Type": "application/json"
    }
    response = requests.request("POST", url, json=payload, headers=headers)
    return response
Interacting with LlamaIndex:

This function queries LlamaIndex to generate accurate, context-aware replies.

st.title("Email Search and Query with LlamaIndex")

search_prompt = st.text_input("Enter search prompt:", value="Get all emails in inbox with 'Customer Query' in the subject")
if st.button("Search Emails"):
    with st.spinner("Searching emails based on the search prompt"):
        if search_prompt:
            emails = get_emails(search_prompt)
            if emails:
                email_data = []
                for email in emails:
                    if email['_zap_search_was_found_status']:
                        subject = email["subject"]
                        body_plain = email["body_plain"]
                        from_email = email["from"]["email"]
                        email_data.append({"Subject": subject, "Body": body_plain, "FromEmail": from_email})
                    else:
                        email_data.append('No emails match the search criteria')

                st.session_state.email_data = email_data

if 'email_data' in st.session_state:
    st.table(st.session_state.email_data)

    if st.button("Generate Reply Using AI"):
        with st.spinner('Querying the LLM to generate replies'):
            st.subheader("LlamaIndex Query Results")
            for email in st.session_state.email_data:
                response = query_llamaindex(f"""Reply to a customer query as Samsung Customer Care by drafting a well-formated email reply based on the provided information. The email should have a proper format with greeting, main body, and signoff. Here’s the customer query: {email["Body"]}""")
                st.markdown(f"**Customer Email Subject:** `{email['Subject']}`")
                st.markdown("**Response:**")
                st.code(response.response, language='markdown')
                draft_response = draft_email_reply(email['Subject'], response.response, email['FromEmail'])
                if draft_response.status_code == 200:
                    st.success("Draft email successfully created in the 'Drafts' folder")

Check out our Github page for more code snippets and a complete app implementation. Share comments and thoughts with us at hello@cohorte.co.

Wrapping Up

And that's it! This should help you build your AI-powered email management tool that can handle the heavy lifting of customer support emails. With the integration of GPT-3.5, LlamaIndex, and Zapier, your team can now process emails more efficiently, leaving more time for what truly matters—providing top-notch support.

This tool is just the beginning. Whether you're managing customer queries, handling campaigns, or sorting through any large volume of emails, the principles and code we've walked through can be adapted and expanded to fit your needs. The possibilities are wide open—now it's your turn to take this foundation and build on it.