Rex Signal to Action Engine

An AI business partner that turns live business and competitor signals into complete owner ready actions while keeping approval with the business owner.

Comment

Inspiration

Business owners rarely suffer from a lack of data. The real problem is that their information is scattered across sales tools, inventory systems, supplier messages, campaigns, financial records, and market research.

Even when the information exists, the owner still has to determine what changed, what matters, what is at risk, and what should happen next.

We created the Rex Signal to Action Engine to close that gap.

Rex is an AI business partner designed to understand the company’s context, connect signals across the business, and prepare complete owner ready actions instead of returning another generic AI response.

For this Build Week project, we focused on one specific workflow:

Turning competitor intelligence into a complete, business specific counter campaign that is saved for review and cannot be published or funded without owner approval.

What it does

The Rex Signal to Action Engine transforms a market signal into a structured business action.

The demonstrated workflow works like this:

  1. Rex reads the current demo business profile and available workspace context.

  2. The Competitor Scanner researches the market and returns up to three relevant competitors.

  3. Each competitor report can include source evidence, confidence, audience signals, estimated advertising pressure, directional CPC, positioning, strengths, weaknesses, and a recommended response opportunity.

  4. The owner selects a specific competitor and chooses Build Counter Campaign.

  5. The selected competitor context is passed into Rex instead of being reduced to a generic campaign request.

  6. GPT 5.6 generates a complete campaign draft shaped around the business, the selected competitor, the owner’s goal, the intended audience, the offer, the channels, the budget, and the campaign duration.

  7. The result is saved as a structured campaign and remains available after the page is refreshed.

  8. Owner Approval Required appears as a safety status. It is not used as the campaign title or as a substitute for the generated campaign.

A completed campaign can contain:

  • The competitor it was generated against
  • Why Rex selected that competitor
  • Campaign goal
  • Target audience
  • Main offer
  • Positioning angle
  • Recommended platforms
  • Suggested budget
  • Budget reasoning
  • Primary social copy
  • Headline and call to action
  • Short form video angle
  • Instagram and Facebook direction
  • SMS version
  • Email subject and body
  • Seven day launch plan
  • Follow up sequence
  • Metrics to track
  • Rex recommendation
  • Limitations or missing information that still needs owner confirmation

Nothing is automatically published, sent, or funded. Rex prepares the action, and the owner makes the final decision.

Why it is different

Most AI campaign generators begin with a blank prompt and return generalized marketing language.

The Rex Signal to Action Engine begins with a real business signal.

It knows which competitor was selected, what that competitor appears to be doing, which audience they may be targeting, where the opening exists, and how the current business is positioned to respond.

The project is not simply a chatbot writing an advertisement.

It is an end to end signal to action workflow:

Competitor evidence

Business context

Structured reasoning

Complete campaign generation

Persistent campaign record

Owner review

Approval before external action

Rex is designed to help the owner move from “something changed in my market” to “here is the complete response I can review.”

How we built it

DestinyAI and the broader Rex platform existed before OpenAI Build Week.

This submission is specifically focused on the new competitor aware signal to action workflow created and extended during the Build Week submission period.

The new work includes:

  • Preserving complete competitor evidence throughout the campaign request
  • Passing the selected competitor into Rex as structured context
  • Preventing counter campaigns from being generated without a selected competitor
  • Generating competitor specific campaign titles and strategy
  • Replacing generic fallback copy with structured limitation states
  • Adding one controlled campaign repair pass when generation is incomplete
  • Persisting structured campaigns through the existing campaign create and list flow
  • Preserving the selected competitor after refresh
  • Displaying complete campaign sections inside the customer portal
  • Keeping owner approval as a safety gate
  • Adding focused campaign validation tests
  • Improving scanner to campaign routing inside the portal

Codex helped us inspect the existing codebase, trace where competitor context was being lost, implement the request normalization and campaign fallback changes, create focused tests, validate the frontend and backend builds, and safely integrate the new workflow without changing unrelated systems.

GPT 5.6 powers the reasoning and structured campaign generation demonstrated in the submission.

The application uses a Next.js and TypeScript customer experience connected to a Python and FastAPI Rex backend. The workflow uses structured context, persistent campaign records, approval controls, and focused test coverage rather than relying on a single unstructured prompt.

Challenges we faced

Preserving the selected competitor

The most important challenge was making sure the competitor selected in the scanner remained attached to the campaign through every step.

The context had to survive:

Scanner result

Campaign wizard

Rex request

Generated response

Campaign persistence

Campaign list

Page refresh

Without that continuity, the system could create a campaign but fail to explain who it was generated against.

Preventing generic fallback campaigns

An incomplete model response could previously be replaced with broad marketing copy that looked populated but was not useful to the owner.

We changed the fallback behavior so Rex first maps structured fields, then performs one controlled repair attempt, and finally returns an honest limitation if required information is still missing.

The system no longer treats vague copy as a complete campaign.

Separating approval from content

Owner approval is an important safety control, but it should not replace the campaign itself.

We separated approval state from campaign content so the owner receives the complete strategy while still retaining control over anything that could be published or funded.

Preserving existing functionality

DestinyAI already contains multiple connected modules. The new workflow had to be added without disrupting Rex Chat, supplier automation, inventory, authentication, billing, or other customer tools.

Focused tests and scoped changes were important throughout the build.

Accomplishments that we are proud of

We are proud that the submission demonstrates a complete business workflow rather than a standalone model response.

The system can:

  • Identify multiple competitors
  • Preserve the selected target
  • Generate a complete response campaign
  • Explain what the campaign is responding to
  • Create channel specific content
  • Produce a controlled seven day execution plan
  • Save the result
  • Restore it after refresh
  • Keep the owner in control
  • Reject generic placeholder campaigns
  • Report missing context honestly

The workflow demonstrates the larger direction for Rex:

AI that does not merely answer a business owner, but understands the signal, prepares the next move, and turns it into something the owner can actually review and use.

What we learned

We learned that model quality alone is not enough.

Reliable business AI also requires:

  • Strong context contracts
  • Structured outputs
  • Persistent state
  • Clear approval boundaries
  • Honest fallback behavior
  • Business specific user interfaces
  • Testing across the entire workflow

The biggest improvement did not come from asking the model for a longer answer. It came from preserving the correct business evidence and giving Rex a clear structure for turning that evidence into action.

We also learned that approval controls do not make the system less capable. They make the system more useful for real businesses because the owner can delegate preparation without surrendering final authority.

What is next

Our goal is to continue developing Rex into a business specific AI partner that understands the company, connects activity across departments, prepares the next move, and helps the owner run the business with the full operation in view.

Built With

Share this project:

Updates