AI Chatbots Are Sending Paying Customers to Businesses That Can't Track Them

·1 views

An Israeli AI-automation agency, Automaziot AI, discovered that AI assistants like ChatGPT and Perplexity had been quietly referring customers to their business since mid-May 2026, generating at least nine tracked web leads plus additional phone inquiries. Two of those leads converted into paying clients worth a combined ₪35,000 (roughly $9,300), yet the company's CRM had misclassified nearly all of them as 'website' or 'unknown' traffic. A key example involved a window-cleaning business owner who phoned the agency after an AI assistant recommended them, closed a deal, and paid — all within the same day, leaving no digital attribution trail. Standard CRM attribution systems, built around paid-click identifiers and form submissions, are structurally unable to capture referrals that originate from AI assistants, especially when the next step is a phone call. The agency found that AI-referred leads also showed the highest inbound-to-outbound message engagement ratio of any acquisition source in their CRM, suggesting meaningful buyer intent rather than casual browsing.

Read the full story at DEV Community

This is an AI-generated summary. ShortSingh links to the original source for the complete article.

Discussion (0)

Bimaaji brings structured, sovereignty-aware AI mutations to Waaseyaa apps

Bimaaji is a new Waaseyaa package designed to give AI agents a safe, structured way to modify application state beyond simple file edits. Instead of directly altering source files, agents submit high-level task definitions that Bimaaji compiles into coordinated changes across entities, migrations, routing, admin panels, and JSON:API resources. The package works in three stages — introspect, propose, and patch — ensuring no filesystem changes occur until a reviewable PatchSet is accepted by a human or automated workflow. A sovereignty validator checks every proposed mutation against a community's deployment rules before any patch is generated. The goal is to prevent the inconsistencies that arise when an agent edits one subsystem without accounting for the rest of the application.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Why Standard HTTP Error Handling Fails for LLM APIs

Backend engineers typically handle HTTP errors using generic retry logic with exponential backoff, but this approach breaks down when applied to large language model APIs. LLM providers reuse standard status codes like 429 and 500, yet the underlying causes vary widely — from temporary rate limits and model overload to hard quota exhaustion and billing issues — each requiring a different response. Blindly retrying timed-out LLM requests can duplicate side effects in agent workflows, inflate token costs, and degrade user experience rather than improve reliability. Operations such as tool-calling agents, streaming chats, and structured output generation each carry different retry risks that a one-size-fits-all handler cannot address. Developers are advised to build retry logic that accounts for the specific LLM operation type and the precise error category, not just the HTTP status code.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

From Passwords to Tokens: How Web Authentication Has Evolved Over Time

Modern authentication methods each emerged to solve specific limitations of their predecessors, tracing a clear line from simple credential passing to stateless token systems. HTTP Basic Authentication, the earliest approach, transmitted base64-encoded passwords with every request, exposing credentials to interception, scaling issues, and an inability to revoke access selectively. Session-based authentication improved on this by verifying credentials only once at login, issuing the client a session ID via a cookie for all subsequent requests. However, sessions introduced server-side state management challenges, synchronization problems across multiple servers, and vulnerability to cross-site request forgery (CSRF) attacks. Sessions also proved ill-suited for non-browser environments like mobile apps and server-to-server communication, paving the way for token-based authentication approaches.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Microsoft ISE shares how it solved context handoff in multi-agent AI systems

Microsoft's Industry Solutions Engineering team has published a detailed account of how it managed conversational context across multiple independent AI agents using the Agent2Agent (A2A) protocol, an open standard now maintained by the Linux Foundation. The core challenge was passing coherent conversation history between coordinator and domain agents that share no common infrastructure or persistent memory. After evaluating three approaches — shared storage, per-agent stateful memory, and inline payload delivery — the team chose to embed summarised conversation history directly within each message. To control payload size, history is condensed every ten conversational turns, a tunable threshold the team acknowledges carries a risk of losing detail or introducing inaccuracies. The method also provides a security benefit, allowing the coordinator to selectively filter sensitive information before passing context to any given domain agent.

0 comments Read more at DEV Community