San Francisco, CA – November 26, 2025 – Markopolo AI today announced the release of ATHENA-709M, the first behavioral foundation model that enables businesses to recover 30-40% of abandoned revenue through individually personalized customer journeys. The model represents a fundamental breakthrough in understanding human behavior across digital experiences.

Understanding what customers will do next

ATHENA solves a problem that has limited marketing technology for decades: predicting what individual customers will do next with enough accuracy to create personalized experiences at scale. The model achieves 72.67% accuracy in predicting the next action a customer will take, compared to less than 1% random chance and 32% from existing approaches.

Markopolo's CTO Rubaiyat Mostofa explains the significance: "At 0.97 AUC-ROC, ATHENA outperforms Google's published recommendation benchmarks by a margin the industry considers transformational. This is the first behavioral foundation model, trained across 603 independent businesses spanning e-commerce, streaming, SaaS, and mobile applications."

The model processes 2.45 million unique customer behaviors and normalizes them into 90 universal behavioral patterns. This enables ATHENA to understand whether someone is comparison shopping, needs social proof, responds better to urgency or validation, and which communication channel they prefer.

How ATHENA powers Markopolo's revenue agents

Markopolo uses ATHENA as the intelligence layer behind its autonomous revenue agents. When a customer visits a website, Markopolo's MarkTag system captures every micro-interaction: hesitation patterns, scroll behavior, product comparisons, cart additions, and time spent on specific pages.

ATHENA transforms these behaviors into a 384-dimensional mathematical understanding of that specific individual. This behavioral fingerprint reveals the customer's intent, their position in the buying journey, price sensitivity, need for validation, preferred communication style, and optimal engagement windows.

The AI then creates a completely unique recovery strategy for that individual customer, answering three critical questions:

Which channel should we use? 

ATHENA's training across 43.7 million customer interactions enables it to predict whether someone responds better to email, SMS, WhatsApp, push notifications, or voice calls based on their behavioral patterns and past engagement.

When should we reach out? 

The model identifies optimal intervention points by analyzing behavioral momentum and engagement patterns. It knows when someone needs immediate contact versus space to consider their decision.

What should we say? 

ATHENA determines whether a customer needs social proof, price comparison information, technical specifications, or limited-time urgency based on their browsing behavior and decision-making patterns.

Real-world performance: from 10% to 40% recovery

Traditional marketing automation treats all customers the same, sending identical recovery sequences to broad segments. This approach achieves 10-15% cart recovery rates because it ignores individual differences in motivation, timing preferences, and communication styles.

“We built the first behavioral foundation model because we believe technology should understand humans as individuals, not as segments in a spreadsheet,” said Tasfia Tasbin, co‑founder and CEO of Markopolo AI. “ATHENA doesn't just predict what customers might do; it understands why they hesitate, when they're ready to buy, and how they want to be reached.”

Markopolo's ATHENA-powered approach creates millions of unique recovery journeys. The results demonstrate the power of true behavioral understanding:

A price-sensitive customer who has compared multiple competitor sites receives price-match guarantees and value comparisons through SMS at their historically optimal engagement time.

An impulse buyer who browsed premium products receives immediate outreach about limited stock availability, with no discount that might devalue the product.

This means if a technical validator who spent time reading specifications would receive detailed comparison guides via email, followed by access to expert reviews and technical specialists through voice calls.

Each journey is created in real-time by AI based on that specific individual's behavioral fingerprint. This approach achieves 30-40% recovery rates, a 3-4x improvement over traditional methods.

The technical breakthrough

ATHENA's breakthrough comes from three technical innovations:

Cross-domain training

Unlike existing models that train on single datasets, ATHENA learned from 603 independent businesses. This 120x greater domain diversity enables the model to recognize universal behavioral patterns that transfer across different industries and business types.

Universal behavioral vocabulary

The model normalizes millions of unique URLs and interactions into 90 semantic event types, enabling it to understand behavior patterns regardless of specific product catalogs or website structures.

Production-grade inference

ATHENA makes predictions in 0.01 milliseconds, 100x faster than typical recommendation systems. This enables real-time decision-making as customers interact with websites.

Industry impact

The release positions Markopolo as the first company to successfully deploy a behavioral foundation model in production. While companies like Meta have built trillion-parameter models for internal use, and academic research has explored multi-domain approaches with 5 datasets, ATHENA is the first model trained across hundreds of independent organizations and deployed for cross-domain generalization.

The model's Expected Calibration Error of 0.065 indicates well-calibrated probability estimates, meaning businesses can trust ATHENA's predictions for automated decision-making. This technical reliability enables Markopolo to autonomously orchestrate revenue recovery campaigns without human intervention.

Availability

ATHENA powers Markopolo's revenue orchestration platform, available to businesses through Markopolo AI. The company projects it will recover $100 billion in lost revenue by 2030 by ensuring no customer is treated as a segment again.

About Markopolo AI

Markopolo AI is a revenue intelligence platform that understand billions of customers as individuals, not segments. By orchestrating AI agents that track, predict, and optimize every customer journey in real-time, Markopolo helps brands recover lost revenue and transform how they connect with customers. Founded in Dhaka and now operating across Saudi Arabia, and Silicon Valley, Markopolo is on a mission to recover $100 billion lost revenue by 2030.