Teach the Agent.Build the Workflow.Scale Autonomous Operations.
Put manufacturing expertise to work at the equipment. Demonstrate a task and explain how it should run. Builder Agents turn that knowledge into a workflow your engineers can test, qualify and deploy.
Your engineers know the steps, the conditions that matter and when to stop. Builder Agents turn that knowledge into workflow logic. Follow the six stages from the first demonstration to a qualified agent—and on to the next task.
01 / Define the Task
Show the Work You Want to Repeat.
Create a project around an equipment task. Give the agent a clear objective and the operating context it needs.
Your Expertise / The Starting Point
01 / Start with the Task
Show the Work You Want to Repeat.
Create a project around an equipment task and the operating knowledge required to perform it.
Project / Equipment setupExample brief
“Turn our demonstrated setup procedure into a repeatable equipment workflow.”
⌘DemonstrationThe engineer performs the task
▤Operating knowledgeSOPs and task conditions
◇Desired outcomeA qualified repeatable workflow
✳
iWave Builder Agents
Operating expertise → executable workflow
02 / Demonstrate & Explain
Demonstrate the Task. Explain What Matters.
Record the equipment actions and add your operating procedure. The demonstration captures what happened; your explanation gives the Builder Agent the context behind it.
Your Demonstration + Operating Knowledge
Watch an engineer record a recipe re-teach, then stop the recording. The Builder Agent identifies the actions and screen assets used to create the first workflow draft.
iWave AutopilotFrom Demonstration to Workflow
Record the task. Demonstrate the re-teach. Review the captured actions.
03 / Program the Agent
Your Expertise Becomes Workflow Logic.
Builder Agents use the demonstration and operating knowledge to create the workflow. Inspect and refine its Functional Blocks—the actions, conditions and checks that determine how the agent works.
Agent-Program-Agent / Inspectable Workflow
The workflow canvas makes the sequence and its decision paths available for engineering review.
iWave AutopilotWorkflow Canvas
Actions, conditions and checks become an inspectable workflow.
04 / Simulate & Refine
Test the Behavior. Refine the Workflow.
Run the workflow in the simulator and follow each step in the activity log. Use those observations to refine the logic and test again before engineer qualification.
Simulation / Review / Refinement
The simulated machine screen and activity log show what the workflow does, giving engineers evidence for review.
iWave AutopilotWorkflow Simulator
Follow the machine-screen action and its activity trace together.
05 / Qualify & Deploy
Qualified by Engineers. Executed at the Edge.
Your engineers qualify and approve the workflow. The approved Autopilot agent runs locally on InnoConnect for the intended equipment task.
Engineer Qualification / Local Execution
After qualification, select the workflow, review its contents and choose the InnoConnect destination for the approved agent.
iWave AutopilotDeployment Preparation
Select the workflow. Review the package. Choose its destination.
06 / Adapt the Workflow
A New Task. The Expertise Carries Forward.
Start from a relevant workflow. Builder Agents incorporate new demonstrations and procedures, customize the logic and prepare the changed workflow for renewed testing and qualification.
AI Scalability / Automating the Customization
06 / AI Scalability through Customization
A New Task. The Expertise Carries Forward.
Use the relevant workflow as a template. Builder Agents incorporate new demonstrations and knowledge, then refine the changed task.
Compare the customizationIllustrative second use case
Builder Agents automate the customization. The adapted agent is tested for its own task.
Agent-Program-Agent
Builder Agents create and refine the workflow. Engineers qualify and approve it. The deployed Autopilot agent carries out the approved task.
AI Scalability / Automating the Customization
New Tasks. Your Expertise Goes Further.
Scaling automation means adapting it to the next task. Builder Agents use a relevant workflow as a starting point, incorporate new demonstrations and procedures, and refine the logic that needs to change.
Apply the demonstrated setup sequence and check that the equipment reaches the expected state.
→
The Next Task
Post-Maintenance Verification.
Add the verification procedure, result conditions and escalation checks needed after maintenance.
Carry the Knowledge Forward
Start from the relevant equipment context and workflow template. Add the demonstration and SOP for the new task.
Let Builder Agents Adapt the Logic
Customize the workflow steps, conditions and exception paths. Simulate the changed behavior and refine the candidate.
Qualify the New Version
Engineers review and qualify the adapted workflow before deployment. Reuse it across the same tool model and qualified process, with compatibility checked for the intended equipment.
Actual Innowave Tech Implementations
From One Machine to an Entire Operation.
Recover an equipment alarm. Correct a drifting process. Coordinate the decisions and actions across a manufacturing operation. Explore the work behind these implementations.
01 / Alarm Recovery
Restore the Operation. Keep the Engineer in Control.
Besi Datacon Evo Series
A search or data-exchange error interrupts the task. Autopilot checks that control is available, performs the qualified recovery sequence and verifies the resulting state.
Implemented scenarios include Adjust Point Not Found, Post Bond Search Failed and Data Exchange Error. Operator control takes priority; unresolved conditions escalate.
Insight detects drift in the post-epoxy result and determines the adjustment needed. Autopilot applies qualified offsets to the epoxy gun’s pressure and height to keep the process within its required conditions.
Monitoring continues after the correction, connecting the next process result to the next decision.
Throughput depends on how the machine’s functions work together. Inspection and process-health observations inform coordinated adjustments across the nozzle arm, pick-and-place, indexing stage and turret.
After maintenance, a bounded tuning workflow adjusts one step at a time, checks quality and alarms, and holds or rolls back when a limit is reached.
Lots from lower-health process tools receive higher metrology sampling priority. Autopilot automates the appropriate tool workflows, from alarm recovery and recipe reteaching to inspection recipe creation.
Out-of-control action plans connect additional measurement, reteaching and lot disposition. Engineers confirm the disposition action and sign off; the remaining disposition steps are automated.
The qualified Autopilot agent runs locally on InnoConnect at the equipment. It follows the approved workflow, checks the task context and escalates conditions it cannot handle.
A Qualified Operation
01
Check the Context
Confirm the required equipment state and task inputs.
02
Execute the Approved Steps
Run the qualified workflow at the equipment edge.
03
Verify the Result
Check the expected outcome. Escalate an unhandled condition.
Local Execution. Connected Management.
The platform deploys and manages agents and receives edge data. Production execution does not need a continuous platform connection; required live inputs, such as MES context, must still be available.
Engineers Retain Authority.
Engineers qualify and approve each version before deployment. Local control takes priority, and an operator can take control when needed. The agent follows its programmed behavior.
Actual Use Case / Connected with iWave Insight
Die Bonder Self-Healing.
Insight determines the recommended correction for post-epoxy drift. Autopilot applies the qualified offsets to the epoxy gun’s pressure and height.
