Release 30 · Draft

Your AI opportunity, connected from discovery to delivery planning.

Agentic AI Canvas Release 30 brings the planning journey around a shared Agentic Brain. Discover the operation, challenge its readiness and value assumptions, then prepare connected outputs for the people deciding what to build.

Release 30 is the draft release currently being built. These release notes will be updated as it lands, and the live site runs what these release notes say is shipped.

Start with the situation you actually have

Instant Chat helps gather the business problem, workflow, people, systems, desired outcomes and constraints before creating a first Canvas. You can start with an early idea, a proposal under review or a stalled attempt. Unknowns can remain explicit instead of becoming invented facts.

Prior-attempt context gives a restart a useful starting point: what was tried, where it broke down and what has changed. The goal is a better next decision, not a more confident retelling of the same proposal. Explore the restart guide.

Build one shared operating model

The eleven authored Canvas sections connect business need, proposed solution, value, implementation and challenges. The Agentic Brain adds systems context, open questions, conversation history and reviewed source summaries to that continuing planning process.

You can refine the model as evidence arrives. Filled sections, AI Readiness and value confidence answer different questions: having text in every section does not prove the opportunity is ready to implement. Explore the Canvas sections.

Choose how closely to steer

• Chat: discuss the model and review individual proposed changes.
• Guided: work through one saved gap at a time and review its proposed improvement before applying it.
• Agent: start a visible, bounded run that can work through changes and can be stopped.

Most assessment and output conversations explain what is present without editing the Canvas or regenerating artifacts. ROI chat can save explicit answers to its active baseline question and apply baseline proposals you accept; it cannot edit authored Canvas sections. Return to an artifact’s owning page to generate or refresh it. Learn the Brain controls.

Inspect what the Brain knows

What I know provides a dedicated view of the Brain’s memory. Search entries and filter by kind and confidence to examine what is shaping the project. Where an entry can be resolved back to an editable source, use its confirmation or correction action; available actions depend on the entry.

Source research has a review step. Read the proposed summary before accepting it into the model; a link alone is not verified evidence. Relevant source and memory context can support continuity as you return to the work. Understand sources and memory.

Separate readiness from the value case

AI Readiness examines data readiness, process readiness, decision clarity, stakeholder alignment and governance. Evidence gaps and recommended next steps help focus preparation. A score is not a probability of success or authorization to build. Sector comparisons depend on sufficient readiness cohort data; sparse samples can offer only directional context.

Agentic ROI examines the confidence behind the value case and uses supplied or explicitly estimated baseline inputs for financial scenarios. Guided baseline questions help collect volume, handling time, loaded cost and other inputs. Annual value and payback ranges expose their calculation basis, while run costs and implementation costs shape the case. Revenue upside remains a separate expected scenario, not committed savings. Missing inputs and estimated evidence stay visible. Read the value-assessment guide.

Prepare the output the next decision needs

The output journey is grouped into four stages, keeping the purpose of each deliverable clear:

• Understand: review the Canvas in Overview, prepare a stakeholder Summary and Slides, and use the available export, copy and print options.
• Prove: examine AI Readiness, Agentic ROI and Gap Assessment to decide which assumptions need attention.
• Design: generate a Solution blueprint, then the solution diagram and agent definitions. Use Security Review to organize questions for qualified review; it is not certification.
• Build: prepare an Implementation plan from reviewed design inputs, with phases, deliverables and success criteria for the delivery team.

A solution may need no autonomous agents. The blueprint can recommend conventional integrations, guardrails or human review where they fit the problem. The plan should reflect that choice. Explore Agentic Outputs.

Refresh deliberately as the model changes

Generated artifacts can show a Stale badge when their inputs change. A dependency-aware refresh can propose upstream and downstream work; review the scope before confirming. Completed content is retained if a later step fails, and unrefreshed dependents may remain stale. A change to the Brain does not automatically rewrite every document.

Hand off the right material with the right access

Export brings downloads, copy and print options together, including audience-focused bundles. Review the selected content and its freshness before using it with a stakeholder or delivery team.

Where sharing is available, Public preview creates a read-only link to selected results that recipients can open without an account. Invite people provides a separate path for signed-in collaborators, with view or edit access. Choose the intended audience and permissions before sharing. Read the Canvas guide.

Optional execution previews remain separate

Run the agents and n8n deployment are gated previews, disabled by default. Where enabled, in-app runs are simulated and use no live tools. Evaluation cases and supplied fixtures support rehearsal; they do not prove that an integration works. Deployment targets an n8n instance the user controls and requires its own actions and prerequisites.

The app is not a hosted production runtime. The current Brain is descriptive and updateable; a live operational digital twin remains a future direction. Generated plans, assessments and outputs require human review.

Find your first useful workflow

Try the industry planning examples, read the missing middle between AI ambition and implementation, or start a Canvas. For the product’s earlier milestone, see Release 20, which introduced Blueprint View.