Comparison questionMedium~10 minAgents & Workflows / 1 of 18

Interview question

What makes an AI agent different from a deterministic workflow?

Distinguishes model-directed execution from explicit application control and explains when each is appropriate.

1Say itAnswer out loud first

2CompareCompare with Core answer

3RefineRefine with Stronger and Advanced

TL;DR

In a deterministic workflow, application code chooses the steps and transitions, even if one step calls a model. In an agent, the model helps decide which action or tool to use next and when the task is complete. I use that flexibility only when the task cannot be expressed clearly enough as a fixed flow and the actions can be safely bounded.

Full answer below

Practice focus

Agent definition, workflow control, model authority, and architectural restraint.

Answer first

Answer path

Core answer first, then checkpoints toward a stronger answer.

Start
Core answer
Start here. This is the direct spoken answer to practice first.
Overview
Using an LLM inside a workflow does not automatically make the system an agent.
In a deterministic workflow, application code chooses the steps and transitions, even if one step calls a model. In an agent, the model helps decide which action or tool to use next and when the task is complete. I use that flexibility only when the task cannot be expressed clearly enough as a fixed flow and the actions can be safely bounded.
Checkpoint
Stronger answer
Optional upgrade with examples, trade-offs, and sharper phrasing.
A document pipeline that always extracts, validates, and stores fields is a workflow. A research assistant that decides which sources to search, follows evidence, and revises its approach is more agentic. The agent still runs inside application-owned limits for tools, permissions, steps, time, cost, and completion criteria. The distinction matters because model-directed control creates more paths to evaluate, observe, and recover.
Finish
Advanced answer
Optional deeper angle for production judgment and nuance.
I treat autonomy as a spectrum rather than a product label. A workflow may contain one model-directed routing step, while an agent can still have deterministic checkpoints and state transitions. I begin with the least autonomy that improves measured task success, then add freedom only where rules become brittle or unstructured evidence drives the next step. Business authority, irreversible side effects, and terminal outcomes remain explicit application concerns even when the model manages part of execution.

Review signals

Use these to catch weak phrasing, shallow reasoning, and risky claims.

Common mistakesCandidateCommon mistakesMistakes that weaken the answer or signal shallow understanding.Open details

Details

Calling every multi-step LLM feature an agent.
Assuming an agent must control every workflow step.
Choosing autonomy because the framework supports it.

Red flagsCandidateRed flagsSignals that an interviewer may treat as risky.Open details

Details

No one can state what decisions the model controls.
The model decides authoritative business outcomes.
The workflow has no explicit terminal conditions.

Follow-up prompts

Quick interviewer prompts inside this drill.

Follow-up questionsCandidateFollow-up questionsLikely next questions after the main answer.Open details

Details

Treat these as the next interviewer turn. Answer one out loud, then save the gap in your practice notes.

Can a deterministic workflow contain tool calls?
Where would you allow model-directed control?
What remains application-owned?

Continue the section

Next in Agents & Workflows: When not to use an agent

Step 2 of 18

When should you not build an AI agent?

Practice route

Read core answer Review checklist Follow-up prompts Track progress

Finished this drill?

Practice freely. Sign in only when you want to save completion, confidence, notes, or bookmarks.

New to Aporeon? Create account

Drill details

Human editorial direction

Aporeon is shaped by Aleksandar Tomovski, a software developer with experience on both sides of technical interviews. Content is reviewed for accuracy, natural spoken delivery, useful depth, and honest trade-offs.

About the creator Editorial standards

Was this drill useful?

Help improve the interview library.

Found this answer useful?

Say thanks with a standalone, one-time $5 contribution. No account required.

No subscription or perks.Choose another amount

Opening drill

Comparison questionMedium~10 minAgents & Workflows / 1 of 18

Interview question

What makes an AI agent different from a deterministic workflow?

Distinguishes model-directed execution from explicit application control and explains when each is appropriate.

1Say itAnswer out loud first

2CompareCompare with Core answer

3RefineRefine with Stronger and Advanced

TL;DR

Full answer below

Practice focus

Agent definition, workflow control, model authority, and architectural restraint.

Answer first

Answer path

Core answer first, then checkpoints toward a stronger answer.

Start
Core answer
Start here. This is the direct spoken answer to practice first.
Overview
Using an LLM inside a workflow does not automatically make the system an agent.
In a deterministic workflow, application code chooses the steps and transitions, even if one step calls a model. In an agent, the model helps decide which action or tool to use next and when the task is complete. I use that flexibility only when the task cannot be expressed clearly enough as a fixed flow and the actions can be safely bounded.
Checkpoint
Stronger answer
Optional upgrade with examples, trade-offs, and sharper phrasing.
A document pipeline that always extracts, validates, and stores fields is a workflow. A research assistant that decides which sources to search, follows evidence, and revises its approach is more agentic. The agent still runs inside application-owned limits for tools, permissions, steps, time, cost, and completion criteria. The distinction matters because model-directed control creates more paths to evaluate, observe, and recover.
Finish
Advanced answer
Optional deeper angle for production judgment and nuance.
I treat autonomy as a spectrum rather than a product label. A workflow may contain one model-directed routing step, while an agent can still have deterministic checkpoints and state transitions. I begin with the least autonomy that improves measured task success, then add freedom only where rules become brittle or unstructured evidence drives the next step. Business authority, irreversible side effects, and terminal outcomes remain explicit application concerns even when the model manages part of execution.

Review signals

Use these to catch weak phrasing, shallow reasoning, and risky claims.

Common mistakesCandidateCommon mistakesMistakes that weaken the answer or signal shallow understanding.Open details

Details

Calling every multi-step LLM feature an agent.
Assuming an agent must control every workflow step.
Choosing autonomy because the framework supports it.

Red flagsCandidateRed flagsSignals that an interviewer may treat as risky.Open details

Details

No one can state what decisions the model controls.
The model decides authoritative business outcomes.
The workflow has no explicit terminal conditions.

Follow-up prompts

Quick interviewer prompts inside this drill.

Follow-up questionsCandidateFollow-up questionsLikely next questions after the main answer.Open details

Details

Treat these as the next interviewer turn. Answer one out loud, then save the gap in your practice notes.

Can a deterministic workflow contain tool calls?
Where would you allow model-directed control?
What remains application-owned?

Continue the section

Next in Agents & Workflows: When not to use an agent

Step 2 of 18

When should you not build an AI agent?

What makes an AI agent different from a deterministic workflow?

Practice focus

Answer path

Overview

Review signals

Follow-up prompts

Continue the section

What makes an AI agent different from a deterministic workflow?

Practice focus

Answer path

Overview

Review signals

Follow-up prompts

Continue the section