Build Guide Decision Tables Business Rules V2

Using Decision Tables and Workato Decision Models in Agentic Workflows

A Workato Decision Model maps input conditions to an output using a rule-based table evaluated top-to-bottom. Each row is one rule; the first matching row wins. Decision Models replace brittle nested IF/ELSE logic with a reviewable, auditable ruleset that business users can verify without reading code.

What Is a Decision Table?

A decision table is a structured mapping of conditions to outcomes. Each row represents one rule — a set of input conditions that, when all matched, produce a specific output. Rows are evaluated in order; the first row where all conditions are satisfied determines the result.

In the context of Workato's Decision Models: columns are variables, cells are conditions for that variable in that rule, all conditions in a row use AND logic, and rows themselves use OR logic relative to each other.

How Workato Decision Models Work

  • Each row defines one rule with an associated output
  • Evaluation is top-to-bottom; the first matching row wins
  • Conditions within a row are ANDed — all must match
  • Rows with OR logic are expressed as separate rows with the same output
  • End with a catch-all row (no conditions) as a safe default or "flag for human review"

When to Use Decision Models in Agentic Workflows

Data table
Use Decision Models WhenUse Other Approaches When
Business rules have multiple contributing signals (e.g., usage + contract tier + region → license recommendation)Logic is purely deterministic with a single condition (use a simple IF or formula)
Logic needs to be reviewable and maintainable by non-engineersRules are stable, simple, and unlikely to change (hardcode or use Properties)
Nested IF/ELSE logic is growing beyond 3–4 levelsLogic requires complex statistical or probabilistic reasoning (use LLM judgment)
Business logic changes frequently and needs a non-code review pathConditions can't be expressed as discrete, enumerable values

Decision Model Best Practices

  1. Use atomic input signals. Avoid pre-combined conditions as inputs — pass the raw signals and let the model combine them. "Usage tier: High" is better than "high-usage-enterprise-customer: true".
  2. Order most-specific rules first. More specific conditions should appear before broader ones. The catch-all row always goes last.
  3. Split OR conditions into separate rows. If two different condition sets produce the same output, write them as two rows with identical outputs.
  4. Always end with a catch-all. A row with no conditions that produces a safe default or "flag for review" output prevents unhandled cases from failing silently.
  5. Treat it as living documentation. Each row should read as: "If X and Y, then Z, because [business reason]." The table doubles as a spec non-engineers can audit.

Example: License Tier Recommendation

Data table
Usage ScoreUser CountContract Type→ Output
High (>80%)>100EnterpriseRecommend upgrade to Premium
High (>80%)AnyAnyFlag for account review
Low (<20%)AnyAnyRecommend downgrade evaluation
AnyAnyAnyNo action required

Decision Models vs. Properties vs. Hardcoded Logic

Data table
ApproachBest ForWho Can Update
Decision ModelMulti-signal business rules that change frequentlyBusiness users with table access
Data TableDynamic reference data (SKUs, thresholds, routing maps)Business users, ops teams
Workato PropertiesEnvironment-specific configuration (endpoints, flags)Admins / engineers
HardcodedTruly static logic that never changesEngineers only

Put these cookbook patterns to work

Get started