AI glossary Models & Architecture
What is task decomposition in AI agents?
← All glossary termsBreaking a goal into smaller subtasks that an AI agent, a set of agents or a team can carry out and check one by one.
Explained
Task decomposition turns a high-level goal into smaller steps, ordered or parallel, each with clear inputs, outputs and a way to tell when it is done. In AI systems, it is what lets a language model or a group of agents handle work too large or varied for a single prompt. The decomposition can be written by people, as in a defined workflow, or produced by the model itself as a plan. It connects specification work, use cases and acceptance criteria, with step-by-step reasoning, tools and orchestration, and it is one of the main ways to keep agents from drifting beyond their scope.
How it works
A complex request is split into subtasks, each small enough to be solved reliably. The subtasks are then executed in order, in parallel or in a mix, and their results are combined.
Research on prompting showed the effect early: least-to-most prompting first has the model break a problem into simpler subproblems and then solve them in sequence, each building on the previous answers, which helped on problems harder than the examples it was shown.
Common patterns
- Prompt chaining: a fixed sequence of steps, where each step's output is the next one's input.
- Parallelisation: independent subtasks run at the same time and the results are merged.
- Orchestrator and workers: a coordinating model breaks the task down on the fly and hands pieces to worker models or agents.
- Plan and execute: the agent writes a plan first, then carries it out and revises it as results come in.
Keeping it verifiable and bounded
- Give each subtask a clear definition of done that can be checked.
- Set stop conditions and limits on steps, time and cost.
- Prefer fixed decompositions for well-known processes; let the model plan only where steps truly vary.
- Log the plan and each step so a person can review what happened and why.
Frequently asked questions
Why is task decomposition important for AI agents?
Because language models are more reliable on small, well-defined steps than on large, open-ended requests. Splitting the work also makes it possible to use the right tool or model for each step and to check results along the way, instead of only at the end.
What is the difference between task decomposition and chain-of-thought?
Chain-of-thought is the model writing out its reasoning within one answer. Task decomposition splits the work into separate subtasks that may be run as different calls, by different agents or with different tools. Decomposition often uses chain-of-thought for the planning step.
Should the model or a person decompose the task?
For stable, well-understood processes, a person-defined workflow is more predictable and easier to test. For open-ended tasks where the steps depend on what is found, letting the model plan is more flexible. Many systems combine a fixed outer structure with model-planned steps inside it.
What can go wrong with task decomposition?
Subtasks can be split badly, missing steps or overlapping. Errors in early steps can carry through to the end. Model-generated plans can grow beyond the original goal. Clear completion criteria, limits and checks between steps reduce these risks.
Decision guides that use this term
How to put it to work
Where it shows up at Thinkia
Further reading
Related terms
Sources
- Zhou et al. · Least-to-Most Prompting Enables Complex Reasoning in Large Language Models
- Anthropic · Building effective agents
- Khot et al. · Decomposed prompting: a modular approach for solving complex tasks (arXiv)
Updated: 2026-10-21