The 9-Step Decision Framework
Boringdots is a business-first workflow operating system for reliable automation.
The core belief:
Automation should not automate chaos. It should automate a business process that is understood, standardized, measurable, reliable, and maintainable.
The framework map
The 9 steps
01 · Business goal
Define the outcome the business cares about. A workflow can be clean and still not worth automating. Before you touch a tool, ask which business or department goal this workflow supports, and what metric proves it matters.
If the workflow doesn't support a real goal, simplify, merge, or remove it instead of automating it.
02 · Workflow
Map the work as it actually happens, not how it's documented in the handbook. Identify who starts the workflow, who receives the output, what happens before and after, what data moves between steps, and which workflows are connected.
A workflow can be core, supporting, compliance/risk, or shared-data. Each tier deserves a different level of investment.
03 · Data
Confirm where the source of truth lives. Is the data clean, complete, and consistent? Are required fields always available? Are field names and status values standardized? Is the data confidential?
Dirty input creates unreliable automation. Data readiness must come before workflow reliability.
If data is messy, build a preparation pipeline before the main workflow.
04 · People
Check that the humans involved are ready to operate the workflow. Will the automation change their daily habit? Can they test it, give feedback, and adapt? Is there a manual fallback? Who owns the workflow after launch, and who gets notified when it fails?
If people aren't ready, start with a small test group, clear feedback loop, manual fallback, and gradual rollout.
05 · ROI and safety
Evaluate value and risk as separate decisions. ROI is a money question: manual time × frequency, minus automated human time, minus build and maintenance cost.
Safety is a risk question: what happens when this workflow fails silently? Who notices, how long until anyone catches it, what's the recovery cost?
High ROI doesn't justify low safety. Both gates must clear.
06 · Input and output
Classify what comes in and what must come out. Structured or unstructured? Single value, list, or document? Confidential or public? What format is expected by the next system, and what guarantees do you need on the output?
This classification drives the next decision: RAG or deterministic.
07 · RAG or deterministic
Decide whether retrieval-augmented generation is the right fit, or whether rule-based logic is cleaner. RAG is correct when input is genuinely fuzzy (natural language, partial matches). Deterministic is correct when input is structured and the output must be exact.
Privacy is a second axis: confidential data points away from RAG by default.
08 · Error handling and retry
Design failure paths before the happy path. What does the workflow do when an API returns 500? When a credential expires? When two runs collide? When a webhook fires twice?
Every reliable automation has explicit error workflows, retry rules, fail outputs, notifications, logs, and locks.
09 · Documentation and maintenance
Write for the humans and future AI agents who will inherit the workflow. Every important section needs notes, tags, expected input/output, failure behavior, retry rules, and ownership.
Every workflow should be understandable by the business user, the automation builder, the future maintainer, and a future AI agent that may inspect, modify, or troubleshoot it.
