AI Workflow Design Matrix¶
Every AI workflow can be described by two dimensions: how much decision-making the AI has and whether a human is in the loop during execution. These two dimensions combine into a 3x2 matrix of six workflow archetypes — a shared vocabulary for classifying, comparing, and designing AI workflows.
Two Dimensions Define Every AI Workflow¶
Dimension 1: Autonomy — How Much Decision-Making Does the AI Have?¶
Autonomy describes how much latitude the AI has to make decisions during execution. This is about the AI's role, not the human's.
| Level | AI's Role | What It Looks Like | Example |
|---|---|---|---|
| Deterministic | Follows fixed rules — no decisions, no judgment | Input → predefined steps → structured output. Same input produces same output every time. | Format a report from a template, extract data from a form |
| Guided | Makes bounded decisions within guardrails | AI chooses how to accomplish a step, but the human sets direction and reviews output. | Draft an email based on meeting notes, research a topic and summarize findings |
| Autonomous | Plans, decides, and adapts independently | AI determines what to do, uses tools, adjusts its approach based on what it finds. | Research agent that finds sources, evaluates quality, and produces a report end-to-end |
Key question: How much does the AI decide on its own?
- Deterministic — nothing; it follows your script exactly
- Guided — some; it makes choices within boundaries you set
- Autonomous — a lot; it plans its own approach and adapts
Dimension 2: Human Involvement — Is a Human in the Loop During Execution?¶
Human involvement describes whether a human participates while the workflow is running — not before (design) or after (review), but during.
| Mode | Human's Role | What It Looks Like | Example |
|---|---|---|---|
| Augmented | Human is in the loop — reviews, steers, or decides at key points | AI pauses for human input, feedback, or approval before continuing. Human and AI collaborate in real time. | Co-writing a document, reviewing AI research before it continues |
| Automated | AI runs solo — human reviews only the final output | Workflow executes end-to-end without human intervention during the run. May be triggered manually or on a schedule. | Weekly report generated overnight, prospect list compiled on a schedule |
Key question: Does a human participate during the workflow run, or only see the final result?
The Matrix¶
Combining these two dimensions produces six distinct workflow archetypes:
| Augmented (human in the loop) | Automated (AI runs solo) | |
|---|---|---|
| Deterministic | Human triggers a fixed sequence and reviews output before it's used. AI follows rules; human confirms results. | Fixed sequence runs on a schedule or trigger with no human involvement. Same input → same output, every time. |
| Guided | Human and AI collaborate — AI researches, drafts, or analyzes; human steers, refines, and decides. | AI handles research, drafting, or analysis on its own, applying bounded judgment. Human reviews only the final output. |
| Autonomous | AI plans and executes multi-step work; human reviews at defined checkpoints before the workflow continues. | AI runs a full pipeline end-to-end — plans, executes, adapts — with no human intervention until the deliverable is complete. |
The Six Archetypes¶
| Archetype | Autonomy | Involvement | Description | Example |
|---|---|---|---|---|
| Deterministic + Augmented | Deterministic | Augmented | Human triggers a rules-based process and confirms results before use | Run a data extraction template, review output, then forward |
| Deterministic + Automated | Deterministic | Automated | Rules-based process runs unattended on a trigger or schedule | Nightly report formatting, scheduled form processing |
| Guided + Augmented | Guided | Augmented | AI drafts, researches, or analyzes; human steers and decides in real time | Meeting prep research, co-writing, competitive analysis |
| Guided + Automated | Guided | Automated | AI applies bounded judgment independently; human reviews the final output | Draft personalized outreach emails overnight for morning review |
| Autonomous + Augmented | Autonomous | Augmented | AI executes a multi-step pipeline with human approval gates | Multi-agent research → write → edit pipeline with a review checkpoint |
| Autonomous + Automated | Autonomous | Automated | AI runs a full autonomous pipeline end-to-end without human involvement | Continuous monitoring agent that detects, analyzes, and alerts |
Worked Examples¶
These three worked examples from the Build section illustrate different matrix positions:
| Example | Archetype | Why |
|---|---|---|
| Deterministic Automation | Deterministic + Automated | Fixed rules, structured input/output, runs the same way every time with no human steering |
| AI Collaborative | Guided + Augmented | AI researches and drafts; human reviews, refines, and decides what to use |
| Autonomous Agent | Autonomous + Augmented | Multi-agent pipeline executes autonomously with one human review gate before publishing |
Choosing Your Archetype¶
Two Questions¶
- How much should the AI decide on its own? → Determines your autonomy level (Deterministic / Guided / Autonomous)
- Does a human need to be involved during the run? → Determines your involvement mode (Augmented / Automated)
Common Progressions¶
Most workflows evolve along predictable paths as you build confidence:
- Deterministic + Augmented → Deterministic + Automated — You start by running the process and reviewing output. Once you trust it, you schedule it to run unattended.
- Guided + Augmented → Guided + Automated — You collaborate with AI on drafts and research. Once the AI consistently produces good output, you let it run independently and review only the final result.
- Guided + Augmented → Autonomous + Augmented — The workflow grows more complex. You add tool access and multi-step reasoning, but keep a human review gate for high-stakes output.
Start simple, upgrade when needed
If you're new to AI workflows, start with Deterministic + Augmented — the lowest-risk archetype. Move to Deterministic + Automated once you trust the process. Explore Guided and Autonomous levels when you're ready for more AI decision-making.
How This Maps to Framework Concepts¶
Execution Patterns¶
The execution pattern spectrum describes what you build. The matrix describes how the AI and human interact. They're complementary:
| Execution Pattern | Typical Archetypes |
|---|---|
| Prompt | Deterministic + Augmented, Deterministic + Automated |
| Skill-Powered Prompt | Deterministic or Guided, either involvement mode |
| Single Agent | Guided or Autonomous, either involvement mode |
| Multi-Agent | Autonomous + Augmented, Autonomous + Automated |
Architecture Patterns¶
The seven workflow architecture patterns provide implementation blueprints within each archetype:
| Archetype | Common Architecture Patterns |
|---|---|
| Deterministic | Augmented LLM, Prompt Chaining |
| Guided | Prompt Chaining, Routing, Evaluator-Optimizer |
| Autonomous | Orchestrator-Workers, Autonomous Agents |
Related¶
- Business-First AI Framework — the full three-step methodology
- Design Your AI Workflow — choose an execution pattern and map building blocks
- Build Workflows — worked examples across the matrix
- Workflow Architecture Patterns — implementation blueprints for each pattern