CampaignCraft AI

• 

  logo

Inspiration

Marketing specialists often need to brainstorm ideas for email campaigns. AI agents can help by finding the products based on sales report data, and crafting example emails and generate HTML email code that can be integrated with existing campaign workflow (ex. using Salesforce).

What it does

CampaignCraft AI is an AI assistant designed to streamline your email marketing workflow. It integrates with your sales data to analyze trends and suggest products with high sales potential for your next campaign. Additionally, it recommends target audiences most likely to convert for those products. Beyond product selection, CampaignCraft assists with crafting email content. Once you define campaign goals, target audience, and chosen products, the AI suggests subject lines and helps generate email body. It also outputs the corresponding HTML code, allowing for seamless integration into your email marketing platform such as Salesforce. This allows users – the marketing specialists – to focus on campaign strategy and optimization while CampaignCraft handles the initial brainstorming and content creation.

How we built it

We used the Vertex AI Agent for the entire design. The app has 3 agents: Email campaign assistant – used to steer user to sales data assistant for finding product promotion ideas, or to email designer for writing the email contents. The sales data assistant uses 1 custom tool – a Data Store containing PDF documents on sales trends and are termed as sales reports. The example used in this prototype is generated using Gemini, using a sports equipment retailer as the company. The document is generated from a list of claims extracted from data, such as the popular sport among teens, the most popular items in summer or the general interest of customers based on demographics. In the real-world environment, these claims can also be extracted from business analysis using Gemini. They are then transformed into Q&A format, by prompting Gemini to brainstorm example Q&A pairs using the claims. The example doc Q&A is added at the end.

Challenges we ran into

It was hard to grasp what the appropriate text data should be for effective querying in a data store tool. In the end, I found that structuring into Q&A is the most effective format. In real world agent usage, this could mean that another step should be added prior to importing sales reports: transforming reports into Q&A pairs.

In this demo, the report details are for a sports retail company, and Q&A pairs are created from Gemini, prompted with:

You are a business analyst for a sports supply retail company. Generate some summaries about sales for the company. The summaries need to include:

- preferences of different groups customers: which sports is favored by customers in which demographic/age group?
- Popularity of products: which products / sports is the most popular in current time of year?

Be diverse and conclusive in the summary, and include as many sport types as possible. For example, sports can be basketball, volleyball, golfing, baseball, hockey etc.

Generate 50 question-answer pairs regarding these claims. Each point should support an answer for the proposed question. 

Built With

• gemini
• vertex-ai-agent
• vertex-ai-tools