Super Intelligence · Agents · Data · Integration · Cloud

Architecture for systems that cannot go down.

Arch SI designs and modernizes the integration, API, and data platforms that regulated enterprises run on — banks, healthcare payers, and government agencies — and extends them to Fabric, Databricks, Snowflake, Palantir Foundry, and super intelligence (SI).

Start a conversation How we engage
20+years in enterprise integration
2013Arizona corporation since
3regulated sectors served
MainframeCICS · AS/400 Core bankbatch · DB2 SaaS + partnersSalesforce · SAP · EDI Gateway + MQ DataPower · API Connect OAuth 2.0 · AAA · SLM Kafka · orchestration Cloud servicesAzure · AWS · GCP · gRPC Data platformsFabric · Databricks · Snowflake SI + agentsClaude · OpenAI · Gemini 40 TPS → 1,000+ TPS real-time payments · one reference architecture

Practices

Four practices, one architecture discipline.

Every engagement starts from the same question: what has to keep running while we change it? The answer shapes the integration layer, the cloud target, the data model, and where super intelligence is safe to use.

Super intelligence (SI), agents & agent skills

Putting models and agents — what federal guidance now calls super intelligence — behind the same gateway, identity, and audit controls your APIs already have. Model-agnostic: Claude, OpenAI, Gemini, and Copilot, chosen per workload.

  • SI-readiness assessment: data access, governance, and integration gaps
  • Agent architecture: tool use, MCP servers, and agent skills that give models governed access to enterprise APIs and data
  • Gateway patterns for agent traffic: authentication, policy, logging, cost control
  • Copilot and retrieval architectures over M365, Fabric, and operational data
  • Proof-of-concept builds that leadership can evaluate before committing budget
ClaudeOpenAIGeminiMicrosoft CopilotAzure AI FoundryMCPagent skillsRAGagent gatewaysPython

Data platforms: Fabric, Databricks, Snowflake & Palantir Foundry

Lakehouse, analytics, and operational data platforms wired to the systems that produce the numbers — on whichever platform your enterprise has standardized.

  • Microsoft Fabric lakehouse and pipeline design, through Power BI semantic models and governed reporting
  • Databricks lakehouse, Delta Lake, and ML pipelines on Azure, fed from operational systems
  • Snowflake warehouse design, data sharing, and Snowpark workloads with governed access
  • Palantir Foundry ontology, pipelines, and integration with core systems for operational decisioning
  • Microsoft 365 tenant, SharePoint, and Teams information architecture
Microsoft FabricPower BIDatabricksDelta LakeSnowflakePalantir FoundryData FactorySharePointEntra ID

Enterprise integration: from Salesforce to agentic SI and beyond

Twenty years of connecting the systems enterprises actually run on — Salesforce, SAP, mainframes, payment rails — now extended to agents that call those same systems. One integration layer, from CRM to agentic SI.

  • SaaS and platform integration: Salesforce, SAP, ServiceNow, Workday, CPQ, billing, and e-signature flows
  • API-first design with OpenAPI; gateway security policies (OAuth 2.0, AAA, rate limiting, SLM)
  • ESB, message-oriented middleware, and event-driven saga/orchestration for real-time processing
  • Agentic integration: exposing the same APIs and events to agents through MCP and governed tool access
  • Center for Enablement (C4E) setup so your teams ship reusable APIs
SalesforceMuleSoftSAPIBM DataPowerAPI ConnectIBM MQACE / IIBApigeeKafkaTemporalMCPgRPC

Cloud on Azure, AWS & GCP: build to run

Moving mainframe- and AS/400-dependent workloads to cloud-native services on Azure, AWS, or Google Cloud — and staying on to run them. From landing zone to steady-state operations, without breaking the business processes on top.

  • Build: landing zones, identity, networking, and current-to-target-state architecture (TOGAF / ArchiMate, BIAN for banking)
  • Migrate: legacy-to-microservices decomposition; .NET/Java services to API gateway and managed cloud runtimes
  • Run: infrastructure as code, CI/CD, observability, cost management, and patching on a managed-services retainer
  • Multi-cloud and hybrid designs where regulators or vendors dictate where data lives
AzureAWSGoogle CloudTerraformKubernetesGitHub ActionsSpring BootJavaPythonTOGAFBIAN

Reference architectures

The four patterns most engagements start from.

Each is a starting point we have delivered in production, not a vendor slide. Boxes with a teal border are where the governance lives — the part that makes the rest safe to change.

Agentic SI behind the gateway

Agents get tools, not credentials. Every call an agent makes crosses the same gateway as every API call, so identity, policy, and audit are enforced once.

People + appschat · Teams · web Agent runtime Claude · OpenAI · Gemini Copilot · agent skills planning · memory MCP tool server tools = your APIs SI gateway OAuth 2.0 · AAA policy · rate · cost audit log Core APIsbank · SAP SalesforceCRM · CPQ Datalakehouse human approval above a risk threshold · every tool call logged with the prompt that caused it

Governed lakehouse: Fabric, Databricks, Snowflake

One source of truth feeds reports, agents, and models. The platform is interchangeable; the ingestion contracts and governance are what we design.

Core systemsDB2 · AS/400 SaaSSalesforce · SAP EventsKafka · MQ Ingest Data Factory Kafka Connect Lakehouse Fabric OneLake Databricks Delta Snowflake Foundry ontology governance · lineage RLS · PII masking Power BIsemantic models Agentsgoverned access MLfeatures medallion layers (bronze → silver → gold) · one semantic model shared by dashboards and agents

API-led integration: Salesforce to agents

System, process, and experience APIs built once and reused. Agents are just another experience consumer, so nothing is re-integrated for them.

SalesforceCRM · CPQ SAP / ERPorders · billing ServiceNowcases · tickets MainframeMQ · CICS Experience APIsweb · mobile · agents Process APIsquote-to-cash · claims System APIsMuleSoft · API Connect Web + mobilecustomers · staff PartnersEDI · B2B APIs AgentsMCP · agent skills Kafka · MQ events alongside the APIs · a Center for Enablement owns the reusable assets

Cloud build-to-run on Azure, AWS, or GCP

The same pipeline and controls on any of the three clouds. Build the landing zone, migrate the workloads, then stay on to run them.

1 · BUILD Landing zone identity · network · policy Terraform · IaC target-state architecture TOGAF · BIAN 2 · MIGRATE Decompose + move legacy → services AKS · EKS · GKE managed databases CI/CD · GitHub Actions 3 · RUN Operate observability · SLOs cost reviews · patching on-call escalation monthly retainer Azure AWS Google Cloud multi-cloud and hybrid where regulators or vendors dictate where data lives

Selected engagements

Workflows we have built, and the agents they are ready for.

Each engagement is a production workflow our architects designed or ran, described by sector. The agentic layer shows how the same integration becomes an SI workflow — the pattern we deliver now. References available on request.

Global bank · treasury

Real-time receivables workflow

Payment received → matched to invoice → exception routed → ledger posted, re-architected from batch to event-driven orchestration with BIAN-aligned services, scaling from ~40 TPS toward 1,000+.

Agentic layer: an agent works the exception queue — reads remittance, proposes the match, posts with human approval above a threshold

Regional bank · core banking

Coreless banking process

Account servicing decoupled from core processing: each customer action (balance, hold, transfer, dispute) exposed as a governed API on API Connect behind a DataPower gateway, with the AS/400 core untouched.

Agentic layer: a servicing agent calls those same APIs through MCP, so a banker or chatbot resolves a request without a screen-scrape

Security & smart-home provider

Quote-to-cash and device onboarding

Lead in Salesforce → CPQ quote → contract e-signed → billing account created → devices provisioned, run as API-led flows on MuleSoft with a Center for Enablement owning the reusable assets.

Agentic layer: a sales agent drafts the quote and provisions the order end to end, with the C4E’s APIs as its only tools

State Medicaid agencies (2)

Claims eligibility workflow

Provider inquiry (EDI 270) → eligibility lookup → response (271) → claim status (276), with ICD-9-to-ICD-10 mapping and AAA, SLM, and XML security policies enforced on the DataPower ESB.

Agentic layer: an eligibility agent answers provider questions over the same 270/271 flow, with every call logged and policy-checked at the gateway

City government · benefits

Benefits case processing

Caseworker request → mainframe screen transaction → eligibility decision → notice issued, through a gateway that turned Unisys screens into services with HL7 support and role-based access.

Agentic layer: a casework agent drives the screen transactions through the gateway, so caseworkers review decisions instead of keying them

State health agency · vital records

Birth, death, and immunization reporting

Hospital event → HL7 v2.6 message → BPEL process → registry updated → Medicaid and partner systems notified, for programs serving 800,000+ children.

Agentic layer: a records agent reconciles mismatched events across partners and raises only the true conflicts for a human

How we engage

Small, senior, and hands-on.

You work directly with the architects who design the solution. No handoff to a bench. Engagements are fixed-scope where the problem is known, and retained where it isn't yet.

Start here

Architecture assessment

Current-state inventory of integrations, gateways, and data flows. You get a written findings report, risk register, and a prioritized roadmap.

2–3 weeks · fixed fee

Then

Design & proof of concept

Target-state architecture with the hard parts built as working code so leadership evaluates something real, not slides.

4–8 weeks

Delivery

Build & enablement

Lead the build alongside your engineers, set design standards, and stand up a Center for Enablement so the practice outlives the engagement.

Program length · milestone-based

Ongoing

Run & fractional architect

A standing seat for teams that need senior judgment on platform direction, plus managed operations for what we built: monitoring, patching, cost reviews, and on-call escalation.

Monthly retainer

About

Architects who have run the systems they design.

Arch SI is the consulting practice of Cloud Services and Consulting Inc, an Arizona corporation founded in 2013. We are a group of enterprise architects with more than twenty years of experience across banking, healthcare, insurance, retail, state and federal government agencies, public health departments, and connected-home platforms. Our architects have held architecture and delivery-lead roles at a global bank, a national security provider, and multiple state and city agencies, and have authored RFP responses and modernization roadmaps that won and shaped engagements.

Contact

Tell us what has to keep running.

Send a few lines about the platform, the constraint, and the timeline. We reply within two business days with whether we can help and what a first step would look like.

Email [email protected] Arch SI · Cloud Services and Consulting Inc · Arizona, USA