Why AI may be the missing infrastructure for neurodivergent tech talent

The quiet talent tax in technology workplaces

Estimates from Birkbeck University of London suggest that between 15 and 20 percent of UK adults are neurodivergent. In technology roles that prize analytical depth and pattern recognition, that percentage is likely higher. Yet most productivity platforms still assume a standard set of cognitive preferences: reliable working memory, effortless task switching, and sustained organisational discipline.

For those whose brains operate differently, the daily cost appears in email triage, context recovery, and the endless reconstruction of order. The technical work itself often feels easier than the administrative overhead required to stay visible inside conventional systems. This imbalance helps explain why many neurodivergent engineers and architects report ending their workdays with no cognitive capacity left for life outside the job.

When internal conflict meets rigid tools

People with co-occurring autism and ADHD frequently describe an exhausting internal standoff. One part of the mind demands structure, predictability and flawless systems. Another rejects repetition the moment novelty fades. The result is a familiar cycle: an elegant new workflow is designed, used with enthusiasm for days, then abandoned when maintaining it requires too much effort. The subsequent disorder creates fresh distress, restarting the loop.

Commercial applications such as Asana, Notion and Todoist have largely failed to break this pattern for many users. The tools require exactly the sustained executive function their owners lack. Paper planners and physical whiteboards meet the same fate. What looks like inconsistency from the outside is often the predictable outcome of competing neurological preferences colliding with inflexible software.

From compensation to self-maintaining systems

A growing number of neurodivergent technologists have begun treating large language models and desktop agents as more than writing assistants. Instead they construct persistent digital scaffolds that watch, classify, remind and report without demanding constant user direction. One solutions architect described launching a local AI assistant each morning that then manages priorities, surfaces forgotten follow-ups and maintains context across fragmented tasks. Security constraints prevent full autonomy, yet the arrangement shifts the cognitive burden from constant supervision to occasional oversight.

This approach differs markedly from generic prompts to draft emails. The goal is not to automate creative output but to externalise the parts of executive function that drain capacity fastest. Early adopters report preserved energy for deep analytical work and, crucially, for relationships outside office hours.

Risks and dependencies that deserve scrutiny

Reliance on privately built AI scaffolds introduces new vulnerabilities. If the system observes work patterns to function effectively, questions of data privacy and employer access immediately arise. Most organisations lack clear policies on employee-created AI agents that process internal information. There is also the longer-term risk that individual solutions mask deeper organisational failures to accommodate cognitive variation in the first place.

Uncertainty remains about whether these tools improve outcomes over years or simply create new forms of dependency. If an AI assistant becomes indispensable, what happens during system outages, job changes or model updates? Speculation that widespread adoption could reduce demand for better-designed collaborative software is plausible but unproven.

Policy and design implications

Technology leaders and regulators have spent considerable energy debating AI safety and copyright. Far less attention has gone to AI as infrastructure for cognitive accessibility. Procurement standards, accessibility guidelines and workplace regulations lag behind the creative ways professionals are already deploying these systems.

Thoughtful integration could help organisations retain talent whose strengths in systems thinking and creative problem-solving are often offset by struggles with administrative theatre. Yet without guardrails, the burden stays on individuals to build and maintain their own accommodations. The more interesting frontier may lie in platforms designed from the start to support variable cognitive styles rather than requiring users to compensate for their shortcomings.

Questions that remain open

Can AI agents be made truly portable across employers without compromising sensitive context? How should performance evaluation change when part of a worker's executive function is effectively outsourced? And at what point does personalised scaffolding become a reasonable accommodation rather than an invisible workaround?

The answers will matter not only to the substantial minority of neurodivergent professionals but to anyone whose working life includes more coordination overhead than the job description ever suggested. The experiment now underway in personal AI systems may ultimately force a broader conversation about what genuine cognitive inclusion in technology workplaces should look like.