The coordination challenge
A margin or inventory question can require channel reports, stock positions, shared metric definitions and an authorized decision-maker. The work crosses several systems.
Back to the case study overview
Deployment story · isA × SN
One platform.Two ways to run it.
Hosted on isagent.io. Deployed for SN as aiX. See how the same platform connects agent work, enterprise data and governed decisions in two different environments.
The isA platform01 / 02
Shared product foundation
App
Console
AEP
DXP
Hosted SaaSisAisagent.io
Customer environmentaiXSN · Profile Edition
The business problem
Bring the evidence and the work together.
SN (SweetNight), the sleep and home brand behind the aiX customer edition, works across commerce, marketing, supply chain, product development, manufacturing and operations. Decisions depend on information owned by different business systems and teams.
The working loop
Bring authorized evidence into a business workspace, investigate it with agents, prepare a decision for review and use approved interfaces for any subsequent action.
Choose the operating environment
The platform is shared. The boundary is yours.
Explore how identity, data, models and operations are arranged in hosted SaaS and in SN’s customer-specific deployment.
Customer-controlled environment
aiX
Configured for SN
aiX · customer Profile Edition
AppConsoleAEPDXP
What this enables
Example: investigate margin and inventory through customer data products, send a decision draft to the authorized reviewer and follow up on any approved business-system action.
- People & workspace
- Give business teams, builders and reviewers a customer-branded entry into platform and domain applications.
- Identity & authority
- Connect enterprise identity and role-based access to governed capability requests, approvals and audit evidence.
- Data & business systems
- Connect customer-owned business systems and the data platform. Domain applications consume governed data products through agreed interfaces.
- Model connections
- Configure customer model infrastructure and permitted external connections. Local inference and network egress follow deployment policy.
- Operations & releases
- Pin platform releases and customer extensions to the target environment. Customer configuration and rollout ownership stay with the deployment.
Product access and executable integrations follow the selected plan or deployment agreement. The two environments have their own configuration, capacity and release lifecycle.
One business question, end to end
A margin question becomes a reviewable decision.
Follow an illustrated commerce and inventory investigation. Each step shows the work, its output and the authority needed to continue.
“Why did margin change, and should we adjust pricing or replenish stock?”
DXP · Data applications01 / 06
Find the business signal
A business user inspects channel performance and inventory context, then checks metric definitions, freshness and quality before drawing a conclusion.
Work produced
A business question linked to its source data and quality context.
Control point
Only data products available to the user’s role enter the investigation.
Interactive illustration · no live customer data
Console · Agent platform02 / 06
Investigate with agents
A configured agent uses the permitted models and tools to compare pricing, channel costs and stock information. Builders can inspect the run and its tool activity.
Work produced
A traceable analysis with supporting evidence and unresolved questions.
Control point
Tool scope and runtime permissions bound the agent’s work.
Interactive illustration · no live customer data
App + DXP · Business workspace03 / 06
Prepare the decision
The agent organizes findings into a reviewable brief: possible causes, a proposed action and the information still needed from the business owner.
Work produced
A decision draft, such as a pricing review or replenishment proposal.
Control point
The reviewer can inspect, edit or reject the recommendation.
Interactive illustration · no live customer data
AEP · Agent governance04 / 06
Check authority and approval
The requested capability, action scope and applicable policy determine whether the next step is allowed or needs approval from an authorized owner.
Work produced
An approval or refusal associated with the requested capability.
Control point
A recommendation alone does not authorize a business-system change.
Interactive illustration · no live customer data
Connected business systems05 / 06
Act through an approved interface
When the delivery includes an executable connector, an authorized action can be dispatched to the owning system. Otherwise, the proposal is handed to the business operator.
Work produced
A system response or an explicit handoff for manual execution.
Control point
The provider system retains its transaction rules and operational ownership.
Interactive illustration · no live customer data
DXP + audit evidence06 / 06
Review the result
Bring execution responses and subsequent business data back into the review. Distinguish a submitted request, an accepted action and an observed business result.
Work produced
Evidence for follow-up, business review and the next decision.
Control point
A successful tool call is checked against the actual downstream result.
Interactive illustration · no live customer data
This is a workflow illustration, not a recording of an executed customer transaction. External write-back requires an enabled connector and the appropriate authorization; recommendations can remain drafts for human review.
The SN business context
A shared foundation across six business domains.
The customer solution organizes data and agent-assisted work around the questions each team needs to answer.
Commerce
How do channel performance, pricing and inventory affect the decision?
Marketing
What do campaign performance and customer feedback suggest we change?
Supply chain
Which demand, inventory or logistics exception needs attention?
Product development
Which research and product feedback should shape the next priority?
Manufacturing
What do production and quality signals tell the responsible team?
Operations
Who owns an exception, what evidence supports it and what happened next?
Domain coverage describes the solution structure. Data readiness, connectors and authorized actions are established for each delivery.
What changes between the deployments?
| Decision | Hosted isA SaaS | Customer Profile Edition · SN example |
|---|---|---|
| Hosting | isA-managed cloud workspace | Customer cloud or on-premise environment |
| Identity & brand | isA account and organization | Customer identity, roles, brand and domain |
| Data connections | Workspace-authorized files and tools | Customer data products and business-system interfaces |
| Models & egress | Available hosted model connections | Customer model infrastructure and permitted external providers |
| Business applications | Capabilities available to the account | Customer-specific applications and integrations |
| Releases & operations | Hosted service releases managed by isA | Target-specific releases, configuration and operating responsibilities |
| Commercial scope | Published plans and included usage | Agreed modules, capacity, support and acceptance criteria |
From platform to customer edition
Reusable core. Customer-owned context.
The delivery model separates platform products from customer business applications and integration work. A Profile configures how those parts fit the enterprise.
- 01
Define the business outcome
Choose a workflow, its business owner, source systems and an observable acceptance result.
- 02
Connect the customer context
Configure identity, branding, model connections, governed data products and the required integrations.
- 03
Verify the whole path
Check data access, analysis, human review, authorization and the behavior of each connected interface.
- 04
Release and operate
Agree the deployment, release ownership, monitoring, support and rollback process before production handover.
isA × SN · aiX
Bring one of your workflows.
Start with a concrete business question. We can map the data, agent work, approval points and deployment that it requires.