Platform without method
automates bureaucracy.
Partners · OutSystems
From methodologyto governedproduction.
Onni defines the what and the why. OutSystems builds and runs the how. Together they form one agentic transformation platform, with method, architecture and governance from the first conversation.
01 · 1 The method · The question
Every company wants AI in production.Few have decided the architecture.
- Shelf agents?
- Everything from scratch?
- One vendor for everything?
- Or your own platform, with method, governance and speed?
01 · 2 The 10-20-70 principle
A platform alone solves the 30%.The offer puts the 70% under engineering.
Source: BCG, The Leader’s Guide to Transforming with AI.
01 · 3 The equation
Agentic transformation needsall four together.
MethodPlatformArchitectureGovernance= Transformation with ROI
Method without platform
stays on the slide.
Speed without governance
becomes technical debt and risk.
01 · 4 The partnership
Two forces.One platform.
Method and work intelligence
- Onni Method, with the 10-20-70 rule
- Fabric · work intelligence
- Builder · redesign and specification
- Atlas · architecture, documentation and governance
Agentic platform and infrastructure
- OutSystems Developer Cloud (ODC)
- Agent Workbench · agents and orchestration
- Mentor · intent-driven development
- Deploy and scale: cloud, hybrid or on-premises
Onni defines the what and the why.OutSystems builds and runs the how.
01 · 5 The pipeline
A governed pipeline,from conception to production.
Onni OutSystems Onni + OutSystems
- 01DiscoverWork, activities and AI potentialOnni
- 02RedesignThe process rethought for AI: the 70%Onni
- 03SpecifyFunctional spec and data dictionaryOnni
- 04BuildApps and agents on ODC, with MentorOutSystems
- 05IntegrateLegacy, APIs and eventsOnni + OutSystems
- 06DeployCloud, hybrid or on-premisesOutSystems
- 07Govern and measureEvidence, audit and ROIOnni + OutSystems
- Security · identity · governance · audit across the whole pipeline
01 · 6 The offer
Three fronts.One platform.
Legacy refactoring
The challenge
Critical systems trapped in old technology, with undocumented rules.
What we do
Atlas extracts rules and data; OutSystems rebuilds them in a governed model.
Indicator
Maintenance cost and time to change
Application modernization
The challenge
A slow portfolio, with a poor experience for users.
What we do
Redesign with the team; web, mobile and chat apps on ODC.
Indicator
Time from idea to production
Agentic automation
The challenge
Manual processes that cross systems and people.
What we do
Orchestrated agents, human in the loop and evidence in Atlas.
Indicator
Capacity per person and quality of evidence
01 · 7 How we co-create
From the first conversationto an MVP in weeks.
- 1DiagnosticAI Transformation Blueprint: where AI transforms the work and the process.Start here
- 2Co-created MVPA critical process redesigned and live on OutSystems.
- 3ScaleThe governed pipeline applied across the portfolio: legacy, new apps and agents.
- 4OperationGovernance, ROI measurement and continuous evolution.
- OnniMethod, redesign, specification and measurement
- OutSystemsPlatform, build, deploy and scale
- Your companyOwner of the architecture, the knowledge and the decisions
Start the co-creation with your diagnostic.
A conversation with an Onni agent maps where AI transforms your work. The complete co-creation offer is in the methodology deck.
02 · 1 The architecture · Today
Everyone already does something with AI.No one governs the whole path.
Who owns it?Claude Code on a laptopProduct team
Which rule applies?Custom GPTCustomer service
Was it tested?Agent in a notebookData
What does it cost?Vendor copilotFinance
Sensitive data?Script with an LLM APIOperations
Who approved it?Gemini in the browserLegal
LegacyERPMainframeIn-house systemsDatabaseDocuments
The problem isn’t the model. It’s the lack of an architecture that governs the whole path, from scope to production, for every system.
02 · 2 Ten dimensions
Ten dimensions without governance. The same ten, governed.
Claude · ChatGPT · Gemini · copilots · loose scripts and agentsOnni intelligence on 100% OutSystems infrastructure
Agentic securityNo governanceGateway · Atlas · OutSystems security
Scope and redesignNo governanceFabric · AI Blueprint
Business knowledgeNo governanceFabric · Brain · Atlas
SDD and spec designNo governanceLayered SDD · Mentor
Agentic architectureNo governanceRegistry · Runtime · Agent Workbench
Component reuseNo governanceAtlas · capability catalog
Legacy integrationNo governanceAtlas · connectors and MCP
Development costNo governanceControl plane · FinOps
Pipeline metricsNo governanceBuilder pipeline · ODC DevOps
Quality and curationNo governanceGate · seal · continuous evaluation
02 · 3 End to end
One governed pipeline,from scope to operation.
Onni intelligence OutSystems 100% of the infrastructure
ScopeSpecifyDevelopRegisterPromoteExecuteOperate and measure
BusinessOnniFabric: each person’s workROI against the baseline
EngineeringOnniCodebase reading and redesignSDD + project specSandbox: refactor and createAtlas absorbs every pushGate, seal and evaluationGateway · Registry · RuntimeAtlas: lineage, drift and cost
ApplicationsOutSystemsODC + MentorPromotion between environmentsWeb, mobile and chat appsLogs, seals and traces
Agents, data and integrationsOutSystemsDatabase (ODC entities)Connectors, APIs and librariesAgent Workbench + AI modelsRegistry and Runtime replica
LegacySystems via OutSystems connectors
Continuous governance: identity · policy · hash trail · cost
02 · 4 In execution
One door, one catalog,one executor, one trail.
Onni where the intelligence lives
Fabric · the individual’s layerPolicies, activities, skills and rules of each role
Builder · engineering intelligenceSandboxes, layered SDD, Brain memory, control plane, preview and tests, Gateway · Registry · Runtime, gate and evaluation
Atlas · inside BuilderRegistry of engineering intelligence: reuse, lineage, drift and cost
OutSystems 100% of the infrastructure
ApplicationsWeb, mobile and chat on ODC; Mentor generates and evolves; DevOps promotes between environments
IntegrationsConnectors, APIs and external libraries: legacy comes in here
AI agentsAgent Workbench + models; Gateway, Registry and Runtime replicated
DatabaseODC entities and data, from test to production
Governance is defined in Onni (policy, gate, cost and hash trail) and executed on OutSystems (corporate identity and permissions).
02 · 5 Dimension by dimension
From how it works todayto governed.
01 · Scope, redesign and business knowledge
AI starts with the right work.
Today, without governanceEach team picks its own use case and automates the process as it is. Business knowledge stays in people’s heads.
With the architecture
- Fabric
- AI Blueprint
- Redesigned process
- Atlas: business layers
Who governs
FabricBlueprintBrainAtlas
02 · SDD and spec design
The company’s standard enters the agent before the first line of code.
Today, without governanceEach developer uses Claude, ChatGPT or Copilot with a prompt of their own. Review becomes handcraft.
With the architecture
- Company SDD
- Layered stack
- Spec before code
- Revalidation
Who governs
BuilderAtlasMentorApp model
03 · Agentic architecture
Ask for the capability: the route optimizes without changing who consumes it.
Today, without governanceEach agent is wired to an endpoint and a model. Optimizing cost means rewriting who consumes it.
With the architecture
- Capability
- Router
- Budget
- Orchestration
Who governs
RegistryRuntimeGatewayAgent Workbench
04 · Component reuse
Before generating, the agent asks what already exists.
Today, without governanceThe same triage agent is rebuilt by three teams. Nobody knows what exists, who built it or its quality.
With the architecture
- Ask Atlas
- Ranking
- Directive
- Justify the new
Who governs
AtlasRegistryBrainReusable ODC components
05 · Legacy integration
Legacy becomes a map before the first new line.
Today, without governanceEach initiative builds its own point-to-point integration. Nobody knows what breaks in the ERP or the mainframe.
With the architecture
- Import
- Impact map
- Capability
- Rebuild on ODC
Who governs
AtlasRegistryBuilderConnectors, APIs and MCP
06 · Cost and pipeline metrics
Every execution has a route, a cost and a trail.
Today, without governanceTokens and hours with no owner and no budget; nobody proves gain or regression.
With the architecture
- Preview
- Integrated tests
- Promotion
- Runtime
Who governs
Control planeRuntimeBuilderODC DevOps
07 · Quality and curation
Only what passes the same checks reaches production, and keeps passing.
Today, without governanceEach agent goes to production the way its team decided. Nobody tests the same way or knows if it got worse.
With the architecture
- Contract
- Promotion gate
- Golden set
- Continuous evaluation
Who governs
RegistryAtlasRuntimeVersioned app model
08 · Agentic security
Security imposed from outside, without depending on the agent’s code.
Today, without governanceCredentials in code, sensitive data without classification, agents with broad access. Security depends on each team’s good will.
With the architecture
- Isolated sandbox
- Safe absorption
- Classification
- Governed edge
Who governs
BuilderAtlasGatewayIdentity and permissions
02 · 6 Limits and the ask
What doesn’t exist yet,and the decision we ask for.
Limits today
- The Registry gate runs in warning mode by default; enforcing the organization’s policy is a configuration step.
- The attack suite measures, but does not yet block promotion.
- Semantic search depends on the embeddings configuration; reuse by terms works without it.
- Onni ↔ OutSystems integration: patterns defined, validated in the pilot.
The ask
A pilot with one legacy system
- 1OnniFabric + AtlasMap the work and import the legacy model.
- 2OnniRegistryPublish the legacy capabilities, with gate and trail.
- 3OutSystemsODCRebuild the application consuming those capabilities.
The pilot starts with your diagnostic.
The diagnostic shows which process and which legacy system to start with. The full technical architecture, component by component, is in the architecture deck.
03 Where both meet
Both paths startin the same place.
01 · The method02 · The architecture
AI Transformation BlueprintYour diagnosticA conversation with an Onni agent →
The diagnostic maps your work, your decisions and your systems. The method starts from what it finds, and the architecture knows where to plug in.
04 Materials
Take the detailto your team.
The two complete decks, for the CIO and for the architects. We ask for your company, a corporate e-mail and a phone number.
Co-creation offer
Methodology
From methodology to governed production: the 10-20-70 principle, the governed pipeline, the three fronts and how we co-create.
Technical deck
Architecture
From scattered AI to governed AI engineering: the ten dimensions, the pipeline in lanes, the Onni and OutSystems components and the pilot.
Don’t buy another AI project.Build your agentic transformation platform.
Models change. Vendors change. Agents change. Your knowledge, your processes and your applications stay with the company.