Skip to content

Business-First AI Framework

Most AI adoption starts with the technology — “we have ChatGPT, where should we use it?” This leads to shallow adoption and misses the workflows where AI creates real leverage. And even when people identify the right workflow, there’s a gap between the idea and making it real — without a repeatable process to follow, they get stuck and stop. The Business-First AI Framework closes that gap: audit your workflows, identify where AI creates the most value, deconstruct those workflows into building blocks, design the architecture, build the components, test them, deploy, and improve over time.

The framework is facilitated by seven skills — reusable AI routines that guide you through each step interactively. You don’t need to figure out the right questions to ask or remember the methodology — the skill drives the conversation, probes for details, and produces structured outputs you carry to the next step. You bring the domain knowledge about your work; the AI handles the process.

StepSkillWhat it guides you through
1. AnalyzeanalyzeIdentify and prioritize the workflows worth reimagining with AI
2. DeconstructdeconstructMap the workflow’s process — or define the outcome it should produce
3. DesigndesignArchitect how AI building blocks will power your workflow
4. BuildbuildBuild the AI building blocks your design specifies
5. TesttestTest your workflow’s output quality and fix what’s not working
6. RunrunDeploy and operationalize your workflow
7. ImproveimproveMonitor quality and innovate your workflow over time

Get the skills: See Get the Skills for installation instructions across Claude Code, Cowork, Claude.ai, Cursor, Codex CLI, Gemini CLI, and VS Code Copilot. The plugin name is business-first-ai.

Seven steps, each building on the last:

Find which workflows are candidates for AI.

Before you can apply AI to anything, you need to know where it fits. Step 1 is a structured audit of your workflows that produces a prioritized list of opportunities classified on two dimensions:

  • Autonomy — How much decision-making does the AI have? Deterministic (follows fixed rules), Guided (makes bounded decisions within guardrails), or Autonomous (plans and adapts independently)
  • Human Involvement — Is a human in the loop during execution? Augmented (human reviews and steers) or Automated (AI runs solo)

The audit starts by determining which lens to use — individual or organizational — then scans what AI already knows about your work, interviews you with lens-specific questions to fill gaps, and produces a classified report with specific opportunities and actionable first steps.

Deliverable: Opportunity Report (outputs/ai-opportunity-report.md) — prioritized opportunities with a Workflow Candidate Summary listing the workflows you’ve chosen to pursue.

Facilitated by the analyze skill. See Analyze Workflows for details and Get the Skills for installation on any supported platform.


Define what the business process does — every step, decision, and handoff — before deciding how to implement it with AI.

How you enter depends on what you’re starting with:

  • You know the process. You can describe the steps, decisions, and handoffs — the model interviews you to surface hidden details and capture it all in a structured format. This is the most common path for workflows you already do manually.
  • You have a problem, not a process. You know what’s broken or slow, but there’s no defined workflow yet. The model proposes a candidate workflow for you to react to, then decomposes it collaboratively.
  • You know the outcome, not the process. You know what you want produced but don’t want to prescribe how — the model captures your goal, constraints, quality criteria, and what the agent system needs to be good at, producing an outcome-driven definition that feeds into agent-oriented design.

For the first two paths, the model uses the six-question framework to break down each step:

  1. Is this step actually multiple steps bundled together?
  2. Are there decision points, branches, or quality gates?
  3. What data flows in and out?
  4. What context, documents, or reference materials does this step need?
  5. What happens when this step fails?
  6. Can the AI access, interpret, and persist the data this step needs?

This is purely the what — it captures the process without prescribing how AI will handle it. The how comes in Step 3 (Design), where the Workflow Definition becomes the input for architecture decisions.

Deliverable: Workflow Definition (outputs/[name]-definition.md) — either a step-decomposed breakdown (refined steps with decision points, data flows, context needs, failure modes, and a context shopping list) or an outcome-driven definition (goal, inputs, outputs, constraints, quality criteria, capability domains, and human gates).

Facilitated by the deconstruct skill. See Deconstruct Workflows for details and Get the Skills for installation on any supported platform.


Decide how the workflow should be built — before you build it.

The Design step takes your Workflow Definition and produces a complete blueprint for your AI workflow. The skill confirms your platform, extracts tool integrations and constraints from your Workflow Definition, assesses the workflow’s autonomy level (Deterministic, Guided, or Autonomous), recommends an orchestration mechanism (Prompt, Skill-Powered Prompt, or Agent) and involvement mode, classifies each step on the autonomy spectrum, maps AI building blocks, identifies skill candidates, and documents agent blueprints when needed. The spec must be approved before moving to Build.

Deliverable: AI Building Block Spec (outputs/[name]-building-block-spec.md) — architecture decisions, autonomy assessment, orchestration mechanism, per-step classifications, skill candidates, agent blueprints, context inventory, and implementation order.

Facilitated by the design skill. See Design Your AI Workflow for the full guide with autonomy assessment, orchestration mechanism decision flow, and output format.


Generate platform-appropriate artifacts from the approved spec.

The Build step starts with a Prepare Context phase — systematically resolving the context needs identified during Deconstruct and Design. Then the model researches your platform using a curated platform registry, integration registries, and web search to generate the actual artifacts — context, skills, prompts, agents, MCP connections — in whatever format your platform requires. Only the components your orchestration mechanism needs are built.

Deliverable: Platform artifacts — prompts, skills, agents, and configs generated for your specific platform, plus resolved context artifacts (style guides, reference materials, examples).

Facilitated by the build skill. See Build for mechanism-specific build paths and platform research details.


Structured testing against the evaluation criteria from Design.

Your first run is a test, not a deployment. The Test step walks you through a quick smoke test (does it run at all?), then a full eval suite where you run each test scenario from the Building Block Spec and score the output on the quality dimensions defined during Design. You also test individual building blocks in isolation and establish a baseline for future comparison.

Most workflows need 2-4 iterations between Build and Test before they produce reliably good output. When something is off, the skill helps you diagnose which building block to fix and sends you back to Build with a clear target.

Deliverable: Test Results (outputs/[name]-test-results.md) — eval scorecard with scores per scenario and dimension, baseline averages, diagnosed issues, and a readiness assessment.

Facilitated by the test skill. See Test for the full evaluation process and troubleshooting guide.


Deploy and operate your tested workflow.

Once your workflow passes testing, Run helps you put it into production. The skill produces a Run Guide tailored to your platform and technical comfort level, then helps you choose the right run pattern — from simple paste-and-run to fully automated schedules. For organizational workflows, Run also covers sharing with your team and setting up ongoing operations.

Deliverable: Run Guide (outputs/[name]-run-guide.md) — artifact inventory, setup steps, first production run instructions, run pattern, and operationalization guidance.

Facilitated by the run skill. See Run for run patterns, deployment options, and operationalization guidance.


Evaluate and evolve running workflows.

Workflows are not set-and-forget. Over time, business context changes, new tools become available, and quality can drift. The Improve step teaches you when and how to revisit a running workflow — watch for quality signals, re-run your eval suite to catch regressions, and assess whether the workflow should graduate to a more capable orchestration mechanism. Four outcomes: no changes needed, tune it, redesign it, or evolve it.

Deliverable: Improvement Plan (outputs/[name]-improvement-plan.md) — current vs. baseline scores, quality signals, graduation assessment, decision outcome, and specific next actions.

Facilitated by the improve skill. See Improve for the full evaluation and graduation framework.


Quick reference for the framework’s vocabulary:

Every AI workflow is classified on two dimensions — autonomy and human involvement — producing six archetypes. See the AI Workflow Design Matrix for the full 3x2 matrix with descriptions, examples, and guidance on choosing the right archetype.

AutonomyDescriptionExample
DeterministicAI follows fixed rules — no decisions, no judgmentFormatting reports, processing forms
GuidedAI makes bounded decisions within guardrailsCo-writing, brainstorming, research summaries
AutonomousAI plans, decides, and adapts independentlyCompetitor monitoring, research → report pipelines
InvolvementDescription
AugmentedHuman is in the loop — reviews, steers, or decides at key points
AutomatedAI runs solo — human reviews only the final output

Used to decompose each step in a step-decomposed workflow:

  1. Discrete steps — Is this one step or multiple bundled together?
  2. Decision points — Any if/then branches, quality gates, or judgment calls?
  3. Data flows — What goes in? What comes out? Where from and where to?
  4. Context needs — What documents, files, or reference materials are required?
  5. Failure modes — What happens when this step fails?
  6. Data readiness — Can the AI access, interpret, and persist the data this step needs?
BlockWhat It Is
ModelThe AI engine that processes inputs and generates outputs
PromptA well-crafted instruction that tells the model what to do
ContextBackground information, reference docs, or examples the model needs
SkillA reusable routine the model can invoke — give it inputs, it follows a defined process, it produces consistent outputs
AgentAn autonomous AI that plans, uses tools, and executes multi-step work
MCP (Model Context Protocol)A connector that lets AI access external tools, services, or databases on your behalf
ProjectA persistent workspace grouping prompts, context, skills, and agents

For detailed definitions and cross-platform examples, see Agentic Building Blocks.

When classifying opportunities from Step 1, it helps to know what type of AI work each one involves. The six use case primitives — Content Creation, Research, Coding, Data Analysis, Ideation & Strategy, and Automation — provide a classification system for the opportunities you identify.

Used to classify each workflow step during Design:

LevelDescription
HumanRequires human judgment, creativity, or physical action
DeterministicFollows fixed rules; AI executes reliably with no decisions
GuidedAI makes bounded decisions within guardrails; human reviews at key checkpoints
AutonomousAI plans and executes end-to-end, including decisions and tool use

The orchestration mechanism answers: who drives the workflow? The right mechanism depends on the workflow’s autonomy level and what it actually needs:

MechanismDescriptionSignals
PromptHuman follows structured instructions step by step, all logic inlineSequential steps, human provides inputs and makes decisions
Skill-Powered PromptHuman invokes reusable skills in a defined sequenceRepeatable sub-routines, moderate complexity
AgentAgent orchestrates the flow, invoking skills and making sequencing decisionsTool use required, autonomous decisions, multi-step reasoning

Single-agent vs. multi-agent is an architecture detail decided during agent configuration — not a top-level choice.

  1. Analyze Workflows — identify your best AI candidates
  2. Deconstruct the workflow — break it into discrete steps
  3. Design your AI workflow — assess autonomy, choose an orchestration mechanism, classify steps, map building blocks
  4. Build the components your orchestration mechanism requires
  5. Test your workflow against evaluation criteria and establish a quality baseline
  6. Run — deploy, choose a run pattern, and operationalize
  7. Improve — periodically evaluate, catch regressions, and evolve