# Oxylabs: The Missing Layer in Agentic AI _Giedrius Šteimantas of Oxylabs reveals the 'missing layer' in agentic AI: the inefficient handling of product page data, leading to massive token waste._ **Published:** 2026-08-25 **Source:** https://www.startuphub.ai/ai-news/artificial-intelligence/2026/oxylabs-the-missing-layer-in-agentic-ai --- In the rapidly evolving world of artificial intelligence, agentic AI promises to automate complex tasks by enabling AI agents to interact with the real world. However, a critical bottleneck is emerging, one that threatens to hinder its widespread adoption and scalability. Giedrius Šteimantas from Oxylabs discusses this challenge, identifying the 'missing layer' in agentic AI as the efficient handling of product data. Inefficient Data HandlingDriver AI agents process all product pages, even if only a few contain relevant informationFrom the article 2 mentionsGiedrius Šteimantas from Oxylabs discusses this challenge, identifying the 'missing layer' in agentic AI as the efficient handling of product data.Massive Token WasteOutcomesending ten product pages to the model when only three yield useful contentFrom the articleThis inefficiency comes at a steep cost in terms of computational resources and token usage. Šteimantas points out that "seventy percent of those tokens go to reading CA..." This implies a massive waste, where a significant majority of the processing power and associated costs are spent on irrelevant or redundant data.Oxylabs' Missing LayerCoreGiedrius Šteimantas identifies efficient product page data handling as criticalFrom the article 2 mentionsGiedrius Šteimantas from Oxylabs discusses this challenge, identifying the 'missing layer' in agentic AI as the efficient handling of product data.isThe Token TaxDriverinefficiency leads to steep costs in computational resources and token usageFrom the article 2 mentionsThis 'token tax' makes scaling agentic AI applications prohibitively expensive and slow.Agentic AI PromiseContextFrom the article 8 mentionsIn the rapidly evolving world of artificial intelligence, agentic AI promises to automate complex tasks by enabling AI agents to interact with the real world.facesInefficient Data HandlingDriverAI agents process all product pages, even if only a few contain relevant informationFrom the article 2 mentionsGiedrius Šteimantas from Oxylabs discusses this challenge, identifying the 'missing layer' in agentic AI as the efficient handling of product data.Massive Token WasteOutcomesending ten product pages to the model when only three yield useful contentFrom the articleThis inefficiency comes at a steep cost in terms of computational resources and token usage. Šteimantas points out that "seventy percent of those tokens go to reading CA..." This implies a massive waste, where a significant majority of the processing power and associated costs are spent on irrelevant or redundant data.Oxylabs' Missing LayerCoreGiedrius Šteimantas identifies efficient product page data handling as criticalFrom the article 2 mentionsGiedrius Šteimantas from Oxylabs discusses this challenge, identifying the 'missing layer' in agentic AI as the efficient handling of product data.The Token TaxDriverinefficiency leads to steep costs in computational resources and token usageFrom the article 2 mentionsThis 'token tax' makes scaling agentic AI applications prohibitively expensive and slow.Path ForwardEffectoptimizing data input to models to reduce unnecessary processing and costscreatesScalability BottleneckOutcomecurrent data processing methods hinder widespread adoption and growth of agentic AIFrom the article 2 mentionsHowever, a critical bottleneck is emerging, one that threatens to hinder its widespread adoption and scalability. ## The Token Tax on Agentic AI The core problem, as outlined by Šteimantas, lies in the sheer volume of data that current agentic AI systems must process. When an agent is tasked with analyzing product pages, for instance, it often needs to sift through numerous pages to find relevant information. Šteimantas provides a stark example: pointing an agent at ten product pages might yield useful content from only three. Yet, the system still sends all ten pages to the model for processing. This inefficiency comes at a steep cost in terms of computational resources and token usage. Šteimantas points out that **"seventy percent of those tokens go to reading CA..."** This implies a massive waste, where a significant majority of the processing power and associated costs are spent on irrelevant or redundant data. This 'token tax' makes scaling agentic AI applications prohibitively expensive and slow. ## Oxylabs' Perspective on Data Handling Oxylabs, a company specializing in proxy solutions and data collection, is acutely aware of these data-related challenges. Their expertise lies in efficiently gathering and delivering web data, a capability that is directly relevant to solving the bottlenecks in agentic AI. By understanding the nuances of web scraping and data retrieval, Oxylabs is positioned to offer solutions that can help agents operate more intelligently. The implication is clear: for agentic AI to move beyond proof-of-concept and become a practical tool for businesses, a more sophisticated approach to data ingestion is required. This involves not just fetching data, but also intelligently filtering, pre-processing, and prioritizing it before it reaches the AI model. The goal is to ensure that the AI agent is fed the most pertinent information, thereby maximizing its effectiveness and minimizing wasted resources. ## The Path Forward for Agentic AI Šteimantas's commentary highlights a fundamental hurdle that needs to be overcome. The future of agentic AI depends on developing smarter data pipelines. This could involve AI-powered summarization tools, advanced filtering mechanisms, or specialized data retrieval agents that can identify and extract key information with high accuracy. The goal is to reduce the amount of raw data sent to large language models, allowing them to focus on higher-level reasoning and task execution. As agentic AI applications become more sophisticated, the demand for efficient data handling will only increase. Companies like Oxylabs, with their deep understanding of web data, are poised to play a crucial role in building the infrastructure necessary for this next generation of AI. The 'missing layer' is not just about getting data; it's about getting the *right* data, at the *right* time, in the *right* format, to enable AI agents to perform at their best. --- Original analysis from [startuphub.ai](https://www.startuphub.ai), the #1 AI startup directory.