Todoist Prioritizes Predictability Over AI Autonomy in New Strategy

Doist is steering its flagship productivity tool, Todoist, away from the industry trend of total AI autonomy in favor of a more predictable user experience. The company is implementing a strategic framework that limits AI's role to interpreting intent rather than controlling the final output.

According to Doist CTO Gonçalo Silva, the company utilizes AI specifically during the "generation phase." In this stage, AI determines what a user is requesting and translates that natural language into specific execution steps. However, once the intent is established, the resulting automation is designed to be predictable and stable. Silva noted that while AI handles the initial interpretation of what a task should actually do, the execution itself remains deterministic.

The Challenge of AI Volatility

This cautious approach is a direct response to the current state of the artificial intelligence market. The rapid pace of evolution within the field has created a volatile environment for software developers. This volatility has made it difficult for Doist to commit to a single long-term roadmap, a specific model provider, or a rigid way of working. By decoupling the interpretation of user requests from the execution of the task, Doist avoids over-committing to any one technology that may become obsolete or unstable.

A Middle Path for Productivity

This strategy represents a "middle path" in the broader AI arms race. While many productivity software companies are rushing to integrate Large Language Models (LLMs) into every possible feature, Doist is betting that reliability is more valuable to the end user than the novelty of fully autonomous agents. The industry is seeing a growing trend toward these hybrid systems, where AI manages the ambiguity of human language but deterministic code handles the actual output to ensure consistency.

Future Outlook

As Doist continues to refine this balance, the company remains flexible regarding its technical stack. Because they have avoided locking themselves into a single provider, they can pivot as more efficient models emerge. The primary focus remains on maintaining user control and system stability, though the company continues to evaluate how the "generation phase" can be further optimized without sacrificing the predictability of the final action.