Architecture

Built around a model of your business — open at the data layer, composable at the capability layer.

ManagedAnalytics is a layered platform. Your data sources stay where they are. The digital twin sits above them, encoding your business logic. AI agents reason over the twin. Plug-in modules add capability without re-architecting. Every layer is open, secure, and audit-ready.

At a glance

Five layers. Designed to compose, not to lock you in.

L5

Executives, applications and embedded experiences

Native applications · Microsoft 365 · Power BI · Tableau · partner apps via API.

L4

Plug-in modules

Composable capability you turn on as you need it.

L3

AI / reasoning layer

LLM agents reason over the twin, not the warehouse.

L2

The ManagedAnalytics digital twin

A live, executable model of how your business actually works — data + structure + business logic + metadata.

L1

Your data layer

Stays where it is. We connect to it.

Security · Access control · Auditability

Layer 01 — Your data layer

Connects to where your data already lives.

ManagedAnalytics does not require you to move your data. We integrate with the systems you have already invested in — operational, financial, planning, IoT, ERP, EAM, time-series, document. Your warehouse, lakehouse and operational systems remain the source of truth. The twin reads from them; we never become a competing data store.

CLOUD DATASnowflake
CLOUD DATADatabricks
CLOUD DATAMicrosoft Fabric
CLOUD DATABigQuery
CLOUD DATARedshift
ERP & FINANCESAP S/4HANA
ERP & FINANCEOracle Fusion
ERP & FINANCEMicrosoft Dynamics
ERP & FINANCENetSuite · Workday
EAMIBM Maximo
EAMSAP PM · Oracle EAM
EAMHexagon EAM
TIME-SERIES & IoTOSI PI
TIME-SERIES & IoTAVEVA Wonderware
TIME-SERIES & IoTGE Proficy · Aspen IP.21
TIME-SERIES & IoTAWS IoT · Azure IoT Hub
PLANNINGAnaplan
PLANNINGOracle EPM
PLANNINGSAP Analytics Cloud
PRODUCTIVITYMicrosoft 365 · Slack · Teams

— Push or pull. Real-time streaming, near-real-time CDC, scheduled batch, event-driven, file-based, API-based.


Layer 02 — The digital twin

A live, executable model of your business.

The twin is what makes everything else possible. It is not a glossary, not a semantic layer, not a dashboard model. It is an executable representation of how your business actually runs — assets, processes, value chains, costs, revenue drivers, operational constraints, organisational rules and decisions.

DATA

The mirrored, conformed view.

Of your operational and financial reality, pulled from the systems in Layer 1.

STRUCTURE

Hierarchical model of your business.

Assets, sites, business units, value chains, organisation. Built as a drag-and-drop visual model.

BUSINESS LOGIC

How value is calculated.

How constraints bind. How processes interact. Encoded once, used everywhere — no re-defining KPIs every time someone builds a report.

METADATA

The descriptive layer.

Capacity ratings, working calendars, units of measure, asset criticality, lineage. Makes the model self-aware.

Because business logic is encoded in the twin — not buried in BI reports, scripts or spreadsheets — there is one place where "what gross margin means" is defined, and every output downstream uses that definition.

Layer 03 — AI & reasoning

AI that reasons over the twin — not over your warehouse.

LLMs are pattern-matchers. Drop them on raw enterprise data and they hallucinate, misattribute, and produce confident nonsense — because the data has no embedded sense of what is happening. Drop them on a digital twin and they reason. Causal explanations become possible. Scenario answers become defensible. Constraint analysis becomes computable.

Causal AI

Variance, performance and value-driver shifts come with explanations grounded in the twin — not generated from text patterns over data.

AI Strategy Analyst & Challenger

Structured strategic analysis at executive speed, with adversarial pressure-testing of assumptions.

Scenario & simulation engine

Run any strategic option, capital reallocation or operational change against the twin and quantify the outcome.

AI commentary

Board packs, exec packs, MBR packs — drafted by AI, audited by humans, grounded in the twin and traceable to source.

Natural-language interaction

Frontline managers and executives ask the platform questions in plain English. The twin makes the answers reliable.

PROVENANCE & CONFIDENCE

Every AI output carries lineage (where the data came from), provenance (which twin elements were used), and confidence (how grounded the answer is). Required for board, audit and regulator-facing use — and the reason executives trust the platform with decisions, not just summaries.


Layer 04 — Plug-in modules

Composable capability. Configure what you need; add more later.

Modules sit on top of the twin and share its data and business logic. You configure the modules you need today — and add more capability over time without re-architecting, re-modelling or re-integrating.


Layer 05 — Executives, applications & embedded experiences

Where the work happens.

Output is wherever your teams already are. Native ManagedAnalytics applications, embedded experiences in your existing tools, BI integration, and the API for everything else.

Native applications

Executive command portal · dashboards · scenario workbench · initiative management · board pack studio.

Embedded in productivity tools

Twin-aware AI inside PowerPoint, Word, Outlook, Teams. Where executives already work.

BI & data tools

Surface twin-derived data in Power BI, Tableau, or any tool that reads SQL or REST.

Custom & partner applications

Built by you or your partners using the open API.


Open API access

Open at every layer.

Every part of the platform is API-accessible. Read from the twin. Write to the twin. Trigger AI agents. Run scenarios. Generate packs. Subscribe to events. The API is how partners and customers extend ManagedAnalytics into their own applications, automations and downstream systems.

  • REST and GraphQL across data, twin, AI, modules and outputs
  • Webhook and event-stream subscriptions for real-time integration
  • SDKs in Python, JavaScript / TypeScript, and .NET
  • Versioned, documented, with sandbox environments
  • OAuth 2.0 / OIDC, scoped tokens, fine-grained permissions
Security, access control & auditability

Wraps every layer. By design, not by patch.

ManagedAnalytics is deployed in regulated, board-facing and capital-market-facing workflows. Security and audit are not bolted on — they are part of the architecture, present at every layer.

ACCESS CONTROL

Role- and attribute-based, integrated with your enterprise SSO (SAML, OIDC, Azure AD, Okta).

DATA ISOLATION

Single-tenant deployment options; private cloud and on-premises available for sensitive workloads.

ENCRYPTION

In transit (TLS 1.2+) and at rest (AES-256). Customer-managed keys supported.

AUDIT TRAIL

Every action — by humans or AI — logged, timestamped, attributable.

LINEAGE & PROVENANCE

Every dashboard, every AI output, every initiative metric traces back to source data through the twin.

COMPLIANCE

SOC 2, ISO 27001 (current certification status on request). GDPR-aware. Customer DPAs available.

Deployment options

Cloud, private cloud, on-premises.

Most customers deploy ManagedAnalytics in our managed cloud. Customers in regulated industries or with sovereignty requirements deploy in their own private cloud (AWS, Azure, GCP) or on-premises. The platform, twin and all modules are identical across deployment modes — only the operational responsibility differs.

Time to value

Weeks, not quarters.

  • 8–12 wkBoard pack wedge — first live executive output.
  • 12–16 wkMBR diagnostics wedge — first quarterly business review run from the twin.
  • 16–24 wkTransformation tracking wedge — value-tracked initiative portfolio.
  • 2–4 qtrFull platform deployment, phased — capability expands as twin scope grows.
Frequently asked technical questions

For the technical evaluator.

Do we have to move our data?
No. The twin reads from where your data already lives — Snowflake, Databricks, Microsoft Fabric, BigQuery, on-premises warehouses, operational systems. We sit above the data layer, not in it.
How is the twin built and maintained?
AI tooling builds the twin in roughly 20% of legacy modelling time. After initial build, the twin updates automatically as data changes and is enriched as your team configures additional structure, logic and metadata.
Is the twin a black box?
No. The twin is fully inspectable — your team can see every asset, process, business rule and KPI definition encoded in it, and edit them through a visual modeller. The AI does the heavy lifting; humans stay in control.
Do you replace our BI tool?
No. We sit above your data layer and beside your BI tool. Power BI, Tableau and Qlik can read twin-derived data through the API. Most customers continue to use their BI tool for self-service reporting while using ManagedAnalytics for executive intelligence.
How is this different from a semantic layer?
Semantic layers (dbt Semantic, Snowflake Cortex, Microsoft Fabric semantic models) define what KPIs mean. The digital twin goes further: it encodes how the business works — assets, processes, drivers, constraints, rules — so AI can reason causally, not just describe.
What about hyperscaler AI (Copilot, Cortex, Genie)?
Hyperscaler AI applies LLMs to data warehouses. We apply AI to a model of your business. We are interoperable — we read from those platforms — and we sit above them as the reasoning layer.
How do you handle data governance and lineage?
Every data point in the twin traces back to its source. Every AI output carries provenance. Every action is logged. Required for board, audit and regulator-facing use.
What is your security posture?
SSO, role-based access, encryption in transit and at rest, customer-managed keys, single-tenant and private-cloud deployment options. SOC 2 and ISO 27001 certifications (current status on request). Customer DPAs available on request.

See the architecture in your environment.

A technical briefing covers integrations, twin construction methodology, security posture and deployment options. Typically a 60-minute call with our solutions team.