Shared context
A governed source of business context that every AI capability reads from, with clear rules on what may be written back.
OptisourceAI · Methodology
The Optisource Enterprise Intelligence Framework
Enterprise AI works when it is designed around how an organisation actually operates: as interdependent systems that share information, depend on each other's decisions and need clear limits.
Why interdependence matters
A decision in one department is an input to another
A discount agreed in sales changes what finance invoices. A recurring fault logged by support changes what operations should prioritise. Leadership decisions depend on all of these signals being accurate and current.
If each department automates on its own, these links remain manual, inconsistent or invisible. The OEIF starts from the links: what information moves between teams, who relies on it, and what controls it needs.
About the example on this page
The workflow below is an illustrative model used to explain the framework. It is not a description of a specific client, and it does not represent measured results.
Interactive model
Enterprise systems, shared data and AI orchestration
Choose a department to see what it contributes, what it relies on, what AI does there and where people stay in control.
Sales and customer acquisition
- Contributes to shared context
- Qualified opportunities, agreed scope and customer requirements.
- Draws on
- Product availability from operations and credit position from finance.
- Role of AI
- Qualifies enquiries and drafts proposals using approved pricing rules.
- Control and human oversight
- Discounts above an agreed threshold need approval from a sales lead.
Worked example
From a new customer to an enterprise decision
Sales and customer acquisition, then customer support, finance, operations and enterprise decision-making. Each step adds to a shared context that the next step relies on.
| Department | Role of AI | Shares with others | Control and human oversight |
|---|---|---|---|
| Sales and customer acquisition | Qualifies enquiries and drafts proposals using approved pricing rules. | Qualified opportunities, agreed scope and customer requirements. | Discounts above an agreed threshold need approval from a sales lead. |
| Customer support | Triages requests, suggests responses and routes complex cases to people. | Service issues, recurring problems and satisfaction signals. | Refunds and service credits need human approval. Support sees only the financial data it needs. |
| Finance | Prepares invoices from agreed scope and flags billing anomalies for review. | Billing status, payment history and revenue information. | Changes to credit limits need finance approval, and every change is logged. |
| Operations | Forecasts demand and proposes schedules based on capacity and commitments. | Delivery capacity, schedules and operating costs. | Schedule changes beyond agreed limits need approval from an operations manager. |
| Enterprise decision-making | Summarises trends and models scenarios for leadership to consider. | Priorities, targets and policy changes that apply across the organisation. | AI provides analysis only. Decisions stay with accountable leaders. |
Governance
How information flows are governed
Connecting departments increases value and increases risk. These five mechanisms keep the connections useful, traceable and safe.
Controlled interfaces
Each exchange between systems or agents is defined in advance: what is sent, in what format, how often and with what limits.
Enterprise data architecture
Owners, definitions, quality checks and lineage for the data that AI depends on.
Authorisation policies
Role-based and purpose-based permissions for people, systems and agents, applying least privilege by default.
Human oversight
Approval points, escalation routes and audit trails for decisions that affect customers, money or compliance.
Applying the framework
How an engagement progresses
Assess
Understand current initiatives, systems, data and constraints, and agree outcomes with leadership.
Design
Map the flows between departments, then design the target architecture, controls and operating model.
Build
Implement priority workflows in stages, with testing and evaluation built in from the start.
Govern
Put monitoring, permissions and incident handling into day-to-day operation.
Scale
Extend to further departments once results and controls are proven.
Map your own enterprise flows
Start with a conversation about where AI is already in use in your organisation and where the connections between teams matter most.