Advanced Prompt Engineering for Hyper-personalised Sales Outreach
The Challenge: improving an AI outreach model
RevvedUp.ai, a startup specialising in hyper-personalised B2B sales messaging, faced a significant challenge with their AI outreach model.
Their initial idea was to create an AI-powered outreach system: after being given a list of businesses to target, the AI would look for relevant individuals to reach out to, scrape through LinkedIn and Websites to learn about these people, and create a personalised message and landing page to use for outreach.
Unfortunately, the output was below their expectations, with content sounding AI-generated, lengthy landing pages and generic messaging. This was affecting email opening rates and conversation rates for their clients, and ultimately, the success of their startup.
Despite their best efforts trying to improve their AI model, they soon realised they had hit a wall.
Our Solution: advanced prompt engineering techniques
The team needed to know how to make AI-generated content sound more natural, how to make AI messaging sound human-generated, and overall how to produce quality outputs for AI Sales Outreach.
To help them, we began by conducting a thorough analysis of the AI model they were using and potential alternatives. Their model was originally relying on ChatGPT, but we identified that Claude, with a similar cost but more efficiency* would be better suited to their needs (*note this comparison between ChatGPT & Claude is relevant to this specific case at this specific time and is not general advice).
To further enhance the tone and relevance of the messages, we used advanced prompt engineering techniques. We fine-tuned the prompts for the AI model to make better use of the information it had gathered, allowing it to link personal information with relevant sales information and trained the LLM on how to create content that sounds human - and not obviously AI-generated.
The Results:
The results that the company have been able to generate for their users speak for themselves, with skyrocketing open rates and increased revenue per sale. After our intervention, it even got a new round of funding for their startup - something that makes us incredibly happy.