The productivity tool graveyard is a familiar sight for millions of professionals. It is the digital wasteland where meticulously configured Notion workspaces, complex Asana boards, and color-coded calendars go to die. For most, these tools fail because of a lack of discipline. For the neurodivergent, however, the failure is structural. The friction does not lie in the tool's features, but in the executive function required to maintain them. When a developer faces a screen of 50 unread emails, the mental energy required to prioritize those tasks often exceeds the energy required to actually perform the work. This is the wall of decision paralysis, and for those navigating the intersection of Autism and ADHD—often termed AuDHD—this wall can feel insurmountable.
The Architecture of Cognitive Assistance
To break this cycle, a developer has engineered a system that shifts the burden of maintenance from the human brain to an AI-driven pipeline. At the core of this setup is Amazon Quick, an AI-powered desktop and web assistant that serves as the orchestration layer. While Quick manages the conversation memory and coordinates various tools, it leverages Amazon Bedrock as its underlying reasoning engine. This decoupling is strategic; by using Bedrock, the user can upgrade the underlying model version to improve reasoning capabilities without needing to rewrite the entire operational workflow.
To bridge the gap between high-level reasoning and actual data, the system utilizes the Model Context Protocol (MCP). MCP is a standardized framework that allows AI models to interact seamlessly with external data sources and tools. The developer built specific MCP servers using Kiro, AWS's AI-integrated development environment, to create live links between Amazon Quick and the user's professional ecosystem, specifically Outlook mailboxes, calendars, and Asana task boards. This transforms the AI from a chatbot into a system with eyes and hands, capable of reading the current state of a workspace in real-time.
Crucially, the logic governing the AI's behavior is not buried in a static system prompt. Instead, the developer externalized the decision-making criteria into Markdown (.md) files. These files contain the classification rules, priority logic, and communication patterns. Because the AI reads these files at the start of each session, the user can alter the system's behavior by simply editing a text file. If a new priority emerges—such as a specific project deadline—the user updates the Markdown file, and the AI immediately adopts the new logic without requiring a redeployment of the code.
For repetitive, structured tasks, the system employs the Quick skills framework. This ensures that workflows like email formatting, context logging, and daily summaries follow a rigid, predefined pattern. By removing the randomness inherent in LLM generation, the developer reduced the cognitive load associated with verifying the AI's output, ensuring that the system remains a tool for efficiency rather than another source of noise.
From Generative Writing to Executive Function
The true value of this system is revealed in the transition from decision paralysis to execution. In a typical morning, the AI scans 50 unread emails and applies a strict set of filters to produce a concise briefing: three tasks that must be handled today and two items awaiting a response from others. Noise and low-priority notifications are filtered out entirely. This process eliminates the 20-minute window of decision paralysis where the user would typically struggle to decide where to start, allowing them to move immediately to the first item on the list.
This prioritization is governed by three explicit rules defined in the external logic files: Is someone waiting for me? Is this immediately actionable? Is there a hard time limit? Only when a task satisfies these conditions does it receive the `Do First` tag. Furthermore, the system tracks the elapsed time of all pending tasks, triggering follow-ups within a week to prevent the psychological anxiety that stems from missed deadlines.
Beyond prioritization, the system addresses the high cost of context switching. When moving between complex technical tasks, the AI provides a summary of previous discussions and the current state of progress, slashing the time spent recovering background knowledge. It also acts as a social buffer; the AI learns the user's natural communication style and refines it into a professional business tone for drafts, reducing the social energy expenditure often required in corporate correspondence.
This represents a fundamental shift in how AI is deployed. Most users treat AI as a writing assistant—a tool to generate text based on a prompt. This system, however, treats AI as cognitive infrastructure. It is a pipeline of observation, classification, action, and reporting. By moving the AI from a passive role to an active orchestrator of the user's environment, the developer has essentially outsourced the executive function of the brain to a digital system.
For AI practitioners, the lesson is clear: the next frontier of productivity is not the quality of the output, but the removal of cognitive friction. When AI is designed to fill the gaps in human executive function—planning, prioritizing, and maintaining—it ceases to be a novelty and becomes a necessity. The goal is no longer to write better emails, but to build an environment where the user never has to wonder which email to write first.
Designing AI as a bridge for cognitive deficits is the only way to truly escape the tool graveyard. By leveraging Amazon Quick, MCP servers, and externalized logic, the role of AI expands from a digital secretary to a cognitive prosthetic that makes professional stability possible for the neurodivergent mind.
