Why Code Effects Matters More Today

AI Is Automating More Work.
That Makes Decisions More Important, Not Less.

Business automation is entering a new phase.

For years, organizations automated workflows by connecting applications, routing tasks and executing predefined processes. Now AI agents can read documents, understand requests, summarize information, communicate with users and take actions across systems.

But this creates a new problem:

Who makes the actual business decision - and how is that decision controlled?

An AI model can interpret information and recommend an action. A workflow platform can move work from one step to another. Neither, by itself, provides a complete answer to questions such as:

• Is this customer eligible?
• Does this transaction comply with policy?
• Is this application approved under the rules currently in force?
• Which policy version produced this result?
• Why did the system reach this decision?
• Who approved the logic that is now running in production?
• Can we reproduce and investigate a decision months later?
• Can an AI agent propose changes without being allowed to silently change production policy?

These questions are becoming more important as automation becomes more autonomous.

Code Effects Version 7 is designed for this environment.

Version 7 expands Code Effects from rule authoring and execution into a broader decision management platform: a governed, deterministic and explainable decision layer that applications, automated workflows and AI agents can use.

The goal is not to replace AI or workflow orchestration. The goal is to give them something they increasingly need:

A controlled place where business policy becomes executable decisions.

The Architecture of a Modern AI-Automated Decision Process

A typical AI-driven business process may involve several different kinds of technology:

Request or Event → AI Understands → Business Rules Decide → Workflow Acts → Decisions Are Recorded and Explained

Each part has a different responsibility.

AI is good at:

• Understanding natural language
• Reading documents
• Extracting and organizing information
• Classifying and interpreting unstructured data
• Conversational interaction
• Identifying possible next actions

A workflow or application is good at:

• Routing work
• Calling services
• Assigning tasks
• Sending notifications
• Updating systems
• Executing business processes

Code Effects is designed to handle:

• Explicit business policy
• Deterministic decision logic
• Eligibility and compliance decisions
• Thresholds and conditions
• Repeatable calculations
• Testing and approval of decision logic
• Controlled publication to production
• Investigation of individual decisions
• Explainability and auditability

This separation matters. An AI model may determine that a request appears to be a particular type of loan application and extract the relevant information from submitted documents.

Code Effects can then apply the approved business policy to those facts.

The workflow can execute whatever must happen next. This creates a clear architectural boundary:

AI interprets. Code Effects decides. The workflow acts.

From a Business Rules Engine to a Decision Management Platform

Earlier versions of Code Effects provided powerful tools for authoring and executing business rules. Version 7 expands that foundation into a much broader decision lifecycle. The new Rule Repository brings rule development, versions, approvals, testing, production investigation and historical information into one environment. The optional Audit library records how production decisions were reached.

The new MCP server provides an integration channel for external AI agents. The built-in AI Rule Editor and AI Assistant allow authorized users to work conversationally with business logic and decision history.

Together, these capabilities support the full lifecycle of an automated decision:

Define → Test → Review → Approve → Publish → Execute → Record → Explain → Investigate

That lifecycle is increasingly important when automated systems make decisions that affect customers, employees, transactions or business operations.

AI Can Participate in Decision Management
Without Becoming the Decision Authority

One of the biggest opportunities - and risks - of AI is its ability to create and modify software and business logic. Version 7 brings AI directly into the decision management environment.

Users can describe the business logic they need in natural language and ask AI to compose or modify a rule. The resulting rule is not simply accepted as arbitrary text. Code Effects validates generated rules against their source object and immediately reports structural validation results. Users can move between conversational authoring and the standard Rule Editor while working with the same rule.

This gives AI a practical role in rule authoring while keeping the rule itself inside a structured decision environment. AI can help answer:

Create a rule that approves this application when these conditions are satisfied.

The platform can then validate whether the generated rule is structurally compatible with the application's available data and operations.

Structural Guardrails for AI-Generated Logic

AI can generate plausible-looking logic that is incomplete, invalid or incompatible with an application. Version 7 addresses this by using Code Effects' structured Rule XML as a precise format that agents can validate. When AI creates or updates rules, drafts or tests, platform guardrails validate the content against its structure and source. Invalid logic can be rejected with validation feedback.

This distinction is important:

Structural validity means:

• The rule is well formed.
• Its elements are valid.
• It is compatible with its source.
• It can be evaluated within the available application context.

Structural validity does not mean:

• The rule expresses the correct business policy.

Business correctness still depends on requirements, testing and review. That is a realistic approach to AI-assisted automation.

AI can help create the logic. The platform can validate its structure. People and organizational processes remain responsible for determining whether the policy itself is correct.

Human Authority Remains in Control of Production Policy

The most dangerous AI automation is not necessarily an AI system that makes a mistake. It is an AI system that can silently change the logic governing production decisions. Version 7 establishes explicit boundaries around AI authority. AI agents acting on behalf of users, independently acting agents and services that have not been identified as users cannot:

• Delete rules or sources
• Approve rules
• Move rules into production
• Take rules offline

Every AI save requests approval. These restrictions apply alongside normal request permissions. This means an organization can allow AI to assist with decision management without giving AI unrestricted control over production policy.

The architectural principle is simple:

AI may propose and prepare. Authorized humans remain responsible for critical production actions.

This is especially important for organizations experimenting with autonomous or semi-autonomous AI agents.

The Rule Repository Governs the Decision Lifecycle

The Rule Repository is central to the Version 7 platform. It provides an integrated environment for developing, testing, approving and managing the logic behind application decisions. Development teams can add these capabilities to their applications without building an entire rule management system themselves.

Immutable Rule History

Every saved rule change creates an immutable version linked to the source version used to author it. Users can compare versions, clone versions and move rules to newer source versions. This creates a reliable historical record of how business logic evolved.

When investigating a decision, the question is not simply:

What does the rule say today?

It may be:

What rule version was in effect when this decision was made?

Versioning makes that question answerable.

Controlled Approval and Publishing

Version 7 provides an explicit workflow for approval and publishing. A rule can be submitted for approval, approved, declined or revoked. Only one version of a rule is live at a time: the version evaluated by the application.

Actions such as Go Live, Take Offline and approval require deliberate confirmation. Combined with permissions, immutable versions and activity records, these controls establish accountability over which business logic reaches production.

This is a major distinction between simply storing rules and managing production decision policy.

Test Decisions Before They Affect the Business

Automated decisions need more than syntactic validation. Organizations need to ask:

What happens if these values change?
Does the rule produce the expected resulWhich conditions caused this outcome?
Does a proposed change break an existing business case?

Version 7 provides testing throughout the rule lifecycle.

Exploration Tests

Users can vary inputs and examine how the logic responds. This supports:

• What-if analysis
• Scenario exploration
• Investigation of alternative outcomes

Expectation Tests

Expectation tests define an outcome the business logic must produce and report whether the evaluation passes or fails. This supports regression-style verification of business policy.

Decision Paths

Each test exposes information about how the decision was reached, including:

• The decision path
• Resulting data values
• Conditions that were satisfied
• Conditions that prevented a result
• Actions involved in the outcome

This makes testing part of the explainability process rather than simply a pass/fail exercise.

Production Decisions Can Be Recorded, Explained and Reproduced

A rules engine can tell an application what to do. A modern decision platform must also help answer what happened afterward. The optional Audit library records production rule evaluations. Depending on configuration, an evaluation record can include:

• The exact rule version that ran
• The outcome
• The branch taken
• Values read during evaluation
• Conditions that led to the decision
• Actions that contributed to the result

This creates a detailed record of an individual automated decision. A user can then investigate questions such as:

Why was this application declined?
Which conditions determined the result?
What data was used?
Which rule version was active?
Can we reproduce the decision?

Recorded evaluations can be opened in the Rule Repository and replayed step by step. They can also become tests using the actual production data. This creates a powerful connection between production behavior and future decision testing.

A real business decision can become a reproducible test case.

That is particularly valuable when investigating unexpected outcomes or validating changes to business policy.

Explainability Is a First-Class Part of the Decision Process

AI systems are often criticized for making it difficult to explain how they reached a conclusion.

Deterministic business rules offer a different model.

A rule can show:

• What information was evaluated
• Which conditions were true
• Which conditions were false
• Which path was followed
• Which actions produced the result

Version 7 connects this decision-level explainability with:

• Rule versions
• Tests
• Production evaluations
• Activity records
• AI-assisted investigation

The result is a richer answer to the question:

Why did the system make this decision?

For organizations operating automated decision processes, that question can be as important as the decision itself.

The AI Assistant Turns Decision Data into a Conversational Resource

The built-in AI Assistant can work across the Rule Repository using the permissions supplied by the host application. Authorized users can ask it to:

• Find and explain rules
• Create and update rules
• Work with versions and drafts
• Create and run tests
• Query test history
• Investigate recorded evaluations
• Query user activity
• Calculate aggregates
• Support reporting

The Assistant shows its steps as it works and preserves conversation history. This changes how users can interact with decision management data. Instead of manually searching through multiple grids and records, users can ask questions such as:

Why was loan application LN-2048 declined?
How many applications over $100,000 were evaluated last quarter, and how many were declined?
Which versions of these rules were tested before they went live?
Who changed this rule, and what happened afterward?

The value here is not merely AI chat. The value is that AI can investigate structured decision information that includes rules, versions, tests, evaluations and activity records.

Connect Written Policy to Executable Decisions

Business policy often exists in more than one place. Part of it may be documented in PDFs. Part may exist as executable rules. Part may exist in historical decisions and test cases.

Version 7 allows authorized users to bring PDF documents into AI-assisted conversations and compare their contents with existing rules, evaluations, activity records and aggregate results.

For example, a user can ask whether eligibility criteria described in a policy document differ from the rules currently being used by the application. This helps connect:

Written policy → Executable logic → Actual decisions

That connection is especially valuable when organizations need to determine whether production logic still reflects current business requirements.

External AI Agents Can Work with the Decision Environment

Version 7 introduces an MCP server based on the Model Context Protocol. The MCP server allows compatible external AI agents to work with the Repository, Audit or both using the application's own OAuth authentication. This provides an external integration channel for AI agents. Depending on the enabled capabilities and permissions, agents can:

• Inspect sources and rules
• Inspect tests and evaluations
• Create and update rules
• Create rule versions
• Work with drafts
• Create and update tests
• Use historical information for investigation and reporting

The same restrictions that protect the built-in AI features also apply. Agents cannot delete, approve, publish or take decision objects offline. Every save requests approval. Tests that would call production code require confirmation from the person using the agent.

This makes MCP more than a way to expose arbitrary tools to AI. It can serve as a controlled boundary between external AI agents and the organization's decision environment.

Code Effects Can Become the Deterministic Policy Layer for AI Agents

This is one of the most important ways to understand Version 7. AI models are probabilistic. Given the same general task, their reasoning and generated output may vary. Business policy often requires the opposite:

• Consistency
• Explicit conditions
• Repeatability
• Controlled changes
• Version history
• Approval
• Explainability

A modern automated architecture can therefore separate responsibilities.

AI handles ambiguity.

For example:

• Read this document.
• Extract relevant information.
• Determine what kind of request this is.
• Summarize the customer's situation.

Code Effects handles explicit policy.

For example:

• Is the customer eligible?
• Does this request meet the required criteria?
• Is this transaction allowed?
• Does the amount exceed an approved threshold?
• Which action must be taken when these conditions are satisfied?

The workflow handles execution.

For example:

• Route the request
• Create a case
• Notify a person
• Call an external service
• Update a business system

This produces a clear model:

AI interprets. Code Effects decides. The workflow acts.

Version 7 provides significantly more infrastructure around that decision layer than rule execution alone.

More Expressive Rules Give People and AI Better Building Blocks

Version 7 expands the language available to both business users and AI agents. New capabilities include:

• Membership conditions with In and Not In
• Inclusive ranges with Between and Is Not Between
• Richer collection logic
• Collection functions such as exists, all, first, last and count
• String composition
• Date calculations
• Date, time and offset-aware values
• Duration-aware authoring
• Broader support for optional values
• More flexible numeric and list comparisons
• Inline execution rules

These capabilities allow more complex decisions to be expressed directly as readable business logic. They also give AI agents clearer structural building blocks when understanding or generating rules. The goal is important:

The business decision remains a complete, readable rule rather than a collection of opaque AI-generated fragments.

Composable Decisions Support Larger Automated Processes

Version 7 also expands the ability to connect business logic through reusable and inline rules. Rules can call other rules while the Repository protects against circular dependencies and changes that would break calling logic. Inline rules allow an execution rule to run another execution rule from its THEN or ELSE actions.

This makes it easier to build larger decisions from smaller pieces of established logic. A sophisticated automated decision process does not always need to be one enormous rule. It can be composed from reusable decision components while preserving visibility into their relationships.

Production Control and Security Are Part of Decision Automation

Decision automation is not only about evaluating conditions. It is also about controlling who can change the logic that affects real business outcomes. Version 7 supports this through:

• Application-defined permissions
• Request-level authorization
• Server-side enforcement
• Immutable versions
• Approval and publishing workflows
• Activity records
• Restrictions on AI agents and services
• Anti-forgery protections
• Browser security features
• Support for strict Content Security Policy environments

The Rule Repository uses the host application's security context rather than requiring a separate security model for the platform. This allows decision management permissions to reflect the organization's existing identities, roles and access policies. AI access is also governed by those controls and by additional restrictions around critical actions.

Your Data and Infrastructure Remain Under Your Control

Version 7 is designed so organizations can keep their rules and repository data on their own infrastructure. The Rule Repository supports SQL Server and PostgreSQL, with additional database support planned.

Code Effects does not host the organization's rules. Organizations control:

• Their database
• Their hardware and infrastructure
• Their security context
• When database updates are applied
• How changes are deployed
• Which users and systems can access decision management capabilities

For organizations with strict security or data-governance requirements, this is an important part of the platform story. AI capabilities can also be configured using the organization's chosen supported language models, while the application controls model selection and API keys.

Why This Matters More Today

The need for business decisions has not disappeared because AI can generate answers. If anything, AI automation makes controlled decision logic more important. As organizations automate more work, they need to distinguish between:

Interpretation

What does the information mean?

Policy

What is the organization allowed or required to do?

Decision

Given these facts and the approved policy, what is the result?

Action

What should happen next?

AI can be extremely effective at interpretation. Workflow systems can be extremely effective at action.

Code Effects Version 7 is positioned around the policy and decision layer between them.

That layer needs to be:

• Deterministic
• Governed
• Testable
• Versioned
• Explainable
• Auditable
• Accessible to AI
• Protected from uncontrolled AI changes

These are exactly the areas Version 7 expands.

A More Complete Decision Lifecycle

The Code Effects Version 7 platform can support a decision lifecycle like this:

1. Define

Business users and developers create explicit decision logic. AI can assist with conversational authoring.

2. Validate

Generated logic is checked against its structure and source.

3. Test

Users explore scenarios and define expected outcomes. Decision paths reveal how results were reached.

4. Review

Proposed logic and versions can be examined before production use.

5. Approve

Authorized users control approval. AI does not independently approve policy changes.

6. Publish

Only approved logic is deliberately moved into production.

7. Execute

Applications evaluate the live rules as part of automated business processes.

8. Record

Audit can capture what happened during production evaluations.

9. Explain

Users can inspect the rule version, data and decision path that produced an outcome.

10. Investigate

AI and users can explore rules, tests, historical evaluations and activity records.

11. Improve

Real production evaluations can become reproducible tests for future changes. This is substantially more than a rules editor. It is a managed environment for the lifecycle of automated business decisions.

The Strategic Position of Code Effects V7

Code Effects Version 7 should not be positioned as a replacement for every workflow platform or as a generic AI agent platform. Its strongest role is more specific.

• AI systems need a way to understand and work with the business.
• Workflow systems need to know what action to perform.
• Organizations need a controlled way to define the policy that determines those actions.

That is where Code Effects fits. A concise way to describe the architecture is:

AI understands the situation. Code Effects applies the policy. The workflow takes action.

Or:

AI automates the work. Code Effects governs the decisions.

Version 7 strengthens that position with decision management, AI-assisted authoring and investigation, controlled agent integration, testing, approvals, production auditing and explainability.

Code Effects V7: A Decision Layer for the Age of AI Automation

AI is making it easier to automate increasingly complex work. That does not eliminate the need for explicit business decisions. It increases the importance of making those decisions:

• Consistent
• Controlled
• Understandable
• Reproducible
• Governed

Code Effects Version 7 brings those capabilities together around the business rules that drive automated decisions. The platform allows people and AI to work with decision logic while maintaining structural guardrails, permissions, approvals and human authority over production changes.

It connects rule authoring with testing. It connects production decisions with explainability. It connects AI agents with controlled access to business policy. And it gives organizations a place to manage the deterministic logic inside increasingly AI-driven automated processes.

The future of automation is not simply AI making every decision. It is AI, applications and workflows working together with explicit, governed business policy. Code Effects Version 7 provides that decision and policy layer.