Sumith Jayawickrama

The framework

The Unified Decision Engine.

Every company sits on three pools of information it never joins up — its enterprise systems, its AI tools, and the living memory of its departments. The UDE is a simple proposition for joining them up, and a test for whether AI is doing anything useful.

The Unified Decision Engine A diagram showing how enterprise data, AI models, and departmental systems converge into a central decision engine, which then feeds leadership reporting, compliance, and partner portals. Sources Engine Outputs Enterprise data ERP · SAP · Oracle · Infor Codes · Parameters Bill of materials · Master files AI layer Large language models Claude · Gemini · GPT · DeepSeek Microsoft 365 Copilot Departments Independent & interdependent Marketing Sales Operations Procurement Finance Warehousing R & D Quality HR IT … the second brain. The Unified Decision Engine APP DATA · MANUAL DATA · AI DATA A clear layer between what the organisation knows and what it decides. Leadership decisions What to do next, and why. Board reporting · Scenarios Continual improvement Authorities Local & international compliance. Statutory filings Audits · Regulated categories Partner access Outside the walls, on purpose. Client portals Vendor portals · Data exports Served Consumers · Customers · Clients

Sources on the left. Engine in the middle. Outputs on the right. The arrows are the whole point: information has to flow, and decisions have to flow back.

The idea in one paragraph.

Most AI conversations happen one layer too high — talking about tools, models, and productivity — and one layer too low — talking about prompts and integrations. The useful conversation happens in the middle. The middle is where the organisation actually decides things. If an AI project is not measurably improving the quality or the speed of a real decision, it is not doing the work. The UDE is a way of keeping that middle in view.

01 · Sources

The three kinds of data every organisation already has.

Enterprise systems. An AI layer. And departments. In most companies these three exist, but they do not speak. The UDE starts by accepting them exactly as they are — the messy ERP, the free-text emails, the half-finished spreadsheets, the supervisor's notebook. Nothing is required to be perfect. It is required to be reachable.

  • Enterprise systems — SAP, Oracle, Infor, the unloved ERP that everyone complains about and nobody replaces.
  • An AI layer — LLMs like Claude, Gemini, GPT, DeepSeek, or Copilot, selected for the work at hand.
  • Departmental memory — marketing, sales, operations, procurement, finance, warehousing, HR, quality, R&D.

02 · Engine

The decision layer.

The engine is not a product. It is a posture. It is a small, disciplined layer that takes the three data streams, connects them against the handful of decisions the organisation actually makes, and produces a reasoned view of what to do next. App data. Manual data. AI data. All three, held in one place, in service of a call.

  • App data — the structured transactions the business already records.
  • Manual data — the notes, the context, the things humans know but never typed.
  • AI data — synthesised views, summaries, scenarios, and recommendations produced by the model layer.

03 · Outputs

What the engine delivers.

When the engine works, three things change. Leaders make better decisions, faster. Compliance and reporting stop being a scramble. And the organisation's partners — clients, vendors, authorities — get clean, timely, consistent access to the information they need. None of this is glamorous. All of it is valuable.

  • Leadership decisions — board reporting, scenario planning, continual improvement.
  • Authorities — local and international compliance, audits, regulated categories.
  • Partner access — client portals, vendor portals, structured data exports.

What the UDE is not.

It is not a product. You cannot buy it. I am not selling it.

It is not a platform. Every organisation that has got close to this has built its own version, on top of whatever ERP and model layer they already have.

It is not an AI strategy deck. It is a shape. A way of looking at the company that says: if information is flowing here, and decisions are flowing here, and the loop is closing, we are doing the work. If not, we are not.

Where I apply it.

I write about the UDE on LinkedIn, use it as the spine of my keynotes, and — inside the operations I help run — use it as a test for whether a proposed AI initiative is worth the energy.

If you are thinking through AI inside your own organisation and want a second opinion grounded in operations rather than slideware, write to me . Short notes get answered first.