Lefty's New AI Capabilities

AI-Powered Features in Lefty: A Brief Timeline

At the Beginning: Computer Vision for Visual Analysis

Lefty began its AI journey by using Convolutional Neural Networks (CNNs) to analyze influencer photos. This enabled accurate detection of:

• Categories — Recognizing relevant themes and topics.
• Gender and Age — Identifying demographics for both influencers and their audiences.

2017: Statistical Models for Audience Quality & Engagement

In 2017, our AI models evolved to estimate key performance metrics & audience insights of influencers:

• Detecting bots, fake accounts, and inactive followers.
• Assessing audience quality and improving targeting accuracy.
• Predicting post performance to optimize campaign outcomes.

2018: Language Models for Text Analysis

With the integration of BERT (Bidirectional Encoder Representations from Transformers), we expanded our capabilities to text-based analysis:

• Enhancing detection of interests and categories.
• Identifying geolocation for both influencers and their followers.

2022: Large Language Models (LLMs) for Advanced Classification

LLMs brought a new level of precision to Lefty’s AI, giving us the ability to:

• Detect sensitive topics in influencer content to ensure brand safety.
• Refine influencer classification for client-defined custom categories.

How do I Know if Something is AI Powered?

To help users easily spot AI-powered insights, Lefty displays an AI Badge next to relevant features:

Up Close: Brand Safety & Country Detection with AI*

• Brand Safety: Lefty analyzes the last 200 posts from an influencer using LLMs to detect potentially sensitive content and ensure alignment with brand values.
• Country Detection: AI scans the influencer’s last 10 posts, along with username and bio text, to determine their primary country of residence accurately.

*both Influencer Brand Safety & Country Detection are available at cost, on-demand

Up Close: Audience Reports in Lefty with AI

Sample Size & Audience Metrics:

• Audience reports are based on a rotating monthly sample of 10,000 to 20,000 profiles, ensuring consistent accuracy. The sample is refreshed monthly to maintain relevance and minimize error.

Determining Demographics & Interests:

• Age & Gender: Photo analysis and bio descriptions help identify age and gender.
• Categories: Captions, @mentions, #hashtags, and location tags from the last 150 posts reveal key interests like Sports, Beauty, and Travel.
• Location: A time-map is created from recent posts to determine the most consistent country or city, accounting for frequent travelers.
• Language: Bio descriptions and captions help identify primary language.

Areas of Exploration for AI

Looking ahead, Lefty will look to further leverage AI capabilities to make influencer discovery and campaign insights even smarter - some areas of exploration include:

• Text-Based Search: Moving from a traditional filter-based search to a natural language tool, enabling queries like: “Find influencers who talk about beauty and perfumes.”
• Automated Insights: Using LLMs to generate deeper insights into influencer content and detect brand-associated topics for better campaign reporting.

With its continuous innovation in AI, Lefty will seek to sett new standards in influencer marketing—ensuring that brands can find the right voices and make data-driven decisions with confidence.