Step 01
Design the process
Break the work into a sequence of named steps. For each step, assign an agent or a person, write the instructions, and add reviewers where oversight is required.
Flows
Turn complex processes into repeatable, reliable work.
Flows coordinate people and agents across multi-step processes from start to finish. Build your best-known approach once, run it consistently, and let it improve with every execution.
Flow · run #128in progress
step 01 · researchagent
complete
step 02 · draftagent
running
step 03 · reviewhuman
waiting
step 04 · publishagent
queued
The coordination gap
Complex work still depends on memory, habit, and heroics.
Most organizations run their most important processes the same way they always have: through emails, spreadsheets, meeting notes, and institutional knowledge held in people's heads. When something changes or someone leaves, the process breaks. Quality becomes inconsistent. Accountability becomes unclear. And every run starts from scratch.
The problem is not a lack of effort. It is a lack of infrastructure.
01
Processes live in people's heads
Emails, spreadsheets, meeting notes, and institutional knowledge carry your most important work. When someone leaves, the process leaves with them.
02
Quality varies run to run
Without a defined sequence and built-in review, outcomes depend on who is running the process and how much experience they have.
03
Every run starts from scratch
Nothing carries forward. Lessons from the last execution are lost, and change mid-process means rebuilding rather than adapting.
The Flows feature
A living SOP that runs itself.
A Flow is a sequence of steps that coordinates people and agents through a process from beginning to end. Each step has a clear assignee, instructions, and an optional review stage. The whole Flow has shared context so every step runs with a unified understanding of the objective.
Flows are not one-time automations. They are version-controlled, continuously improving, and built to handle change without starting over.
- 01
Step-by-step structure
Break any process into a clear sequence of steps. Assign each step to an agent or a person, add instructions, and set review requirements where oversight matters.
- 02
Built-in review at every step
Add an agent reviewer, a human reviewer, or both to any step. Each reviewer gets their own instructions. Quality and accountability are part of the process, not added on afterward.
- 03
Adaptive re-runs
When requirements change mid-process, reset and re-run with new context. Opal reworks only the steps the change affects and skips the rest. No starting over.
- 04
Continuous learning
With Continuous Learning enabled, each successful run compiles observations and lessons into a document that improves future runs. Flows get better the more you use them.
- 05
Full traceability
Every action across every step is captured in a Run Log. Work is auditable, decisions are traceable, and nothing is lost.
- 06
Step-level resources
Every step draws on the knowledge and resources available in its Space and from its assigned agents. When a step needs something more, attach additional Knowledge, Skills, Connectors, Tools, and Guardrails at the step level.
Flow · anatomylive run
Knowledge · input
Space knowledge, context, and guardrails feed every step
step 01 · gather contextagent
complete
step 02 · build the workagent
running
human reviewhuman-in-the-loop
step 03 · sign-offhuman
waiting
step 04 · deliveragent
queued
Output
Completed work, reviewed and logged
Lessons learned
Fed back into the next run
How Flows work
Build once. Run reliably. Improve continuously.
Step 02
Set the context
Define the overarching goals and purpose of the Flow at the Flow level. Every step runs with a shared understanding of what the work is for.
Step 03
Run with confidence
Start a run. Agents execute, reviewers check, and humans are notified when their input is needed. Pause or reset at any point. Every action is logged.
Step 04
Adapt without starting over
If something changes, reset the current run, add the new context, and re-run. Opal reviews each step, reworks what needs to change, and skips what does not.
Step 05
Capture and apply what you learn
With Continuous Learning enabled, each successful run generates captured insights that sharpen future runs. Your processes improve over time without manual updates.
What Flows deliver
Processes that hold up when things change.
Consistency at scale
Every run follows your best-known approach. No variation based on who is running it, what day it is, or how much experience someone has.
Oversight where it matters
Human and agent reviewers are built into the process at the step level. You decide where quality checks happen and what each reviewer is responsible for.
Resilience to change
When requirements shift, a Flow does not break. Reset, re-run with new context, and only the affected steps are reworked. The rest carry forward.
Continuous improvement
Flows learn from every successful run. Lessons are captured, reviewed, and applied so each execution is better than the last.
Full auditability
A complete Run Log records every step, every decision, and every action. Work is traceable and evidence is always available.
Flows in action
See it in practice.
Three shapes of coordinated work, from transformation programs to client delivery.
Case 01
Operations
Transformation project management
An operations lead builds a Flow to manage an internal transformation project end to end. A Research Agent handles current-state analysis. An Analyst Agent drafts recommendations. A human reviewer approves each stage before work proceeds. Midway through, new requirements arrive. The operations lead resets the current run, adds the updated context, and re-runs. Opal reworks only the steps the change affects and skips the rest. The project continues without losing progress.
- Steps
- Research · analysis · approval
- Change handling
- Reset and re-run
- Outcome
- Progress never lost
Case 02
Content
Content production pipeline
A content team builds a Flow for their monthly publication cycle. A Research Agent handles topic validation, a Writing Agent drafts the piece, and an editorial reviewer approves before the content moves to publishing. Each run follows the same process. Continuous Learning captures what worked and what to improve, and each production cycle runs more smoothly than the last.
- Cadence
- Monthly publication cycle
- Review
- Editorial approval gate
- Outcome
- Each cycle runs smoother
Case 03
Professional Services
Client onboarding and delivery
A professional services team builds a Flow for client onboarding. Each step covers a defined stage of the engagement, assigned to the appropriate agent or team member with review checkpoints at key handoffs. New client details are passed in as run-level context, so the same Flow adapts to each client without rebuilding the process each time.
- Structure
- Stage-by-stage engagement
- Inputs
- Run-level client context
- Outcome
- One Flow, every client
Enterprise-ready by design
Governed processes, end to end.
01
Role-Based Access Control
Granular permissions determine who can create, edit, run, and review Flows within each Space.
02
Audit Log
Every action in every Flow run is captured in a complete, tamper-evident log. Full traceability, always.
03
Versioning
Flows maintain version history. Review how a process has changed over time and restore previous versions when needed.
04
Zero Data Retention
By default, your data is not used to train AI models. Your processes and their outputs remain yours.
05
Certifications
ISO 27001, ISO 42001, and SOC 2 Type II certifications are underway.
Flows work with the rest of Opal
How Flows fit the platform.
01SpacesFlows run within a Space, which provides operational context, access control, and the shared Knowledge every step can draw on.02AgentsAgents are assigned as step executors and reviewers, carrying out and checking the work at every stage.03TasksTasks handle a single unit of work. Flows orchestrate an end-to-end process across multiple steps, assignees, and reviewers.04KnowledgeSpace-level Knowledge is available to every step by default. Additional Knowledge can be attached at the step level when a step needs it.05SkillsSkills can be attached at the step level to give a specific step a defined capability or approach.06ConnectorsConnectors can be attached at the step level to give a specific step access to external systems and data.07ToolsTools can be attached at the step level for actions a specific step needs to perform.08GuardrailsInput and output Guardrails can be set at the step level to enforce quality and compliance standards for that step.
Coordinate complex work with built-in oversight.
Build processes that run reliably at any scale.
Flows give your organization the infrastructure to coordinate complex work consistently, with built-in oversight and the ability to improve with every run.