Prolego glossary

These definitions explain the terms Prolego uses across the library. Each describes a role or capability; the exact implementation depends on the business and task.

AI Abundance

A future in which intelligence no longer constrains the economic growth of the business, with AI at the core of every economically consequential decision. Prolego uses this vision to guide investment in a shared capability that can expand across thousands of decisions. See AI Abundance.

AI native architecture

An evolving architecture for creating value quickly and scaling toward AI Abundance. It builds shared Operations Intelligence around major workflows, uses existing authoritative source systems and brings AI into work through interfaces such as Workbenches and Coaches. The architecture keeps adapting as AI improves. See architecture.

AI native product team

A team that uses AI for communication, shared knowledge and decisions as well as engineering. Meetings, status updates, decisions and actual work feed a shared repository of project knowledge that anyone on the team can query. This reduces repeated meetings, manual reporting and the delay between learning something and acting on it. The tools and working methods are still evolving. See team practices.

AI Systems Engineer

An engineer with systems-level thinking, strong communication, fast learning and the initiative to find better ways to work with AI. Actively building and using AI every day demonstrates that disposition. The role builds reliable systems across the whole workflow; an NLP, machine-learning or AI-training specialty is not the defining qualification. See the AI Systems Engineer you need.

AI Workbench

A flexible environment where knowledgeable people investigate questions, prepare documents and use maintained company information and tools. It can combine shared guidance with reviewed local experimentation. See design patterns.

Coach

An assistance pattern that brings company guidance into an employee's existing workflow. A browser-side-panel Coach can use permitted screen context to explain a situation or prepare guidance. The assistance pattern does not itself grant authority to write records or execute transactions. See interfaces and authority.

Compounding value

The approach of choosing a first workflow that creates value and builds knowledge, connections, tools and reliability capabilities the next workflow can reuse. Operations Intelligence and the AI Workbench let the same team keep operating earlier work while adding more. See invest for compounding value.

Context window

The information a model can consider in one invocation, including instructions, source material, tool information and conversation history. Storage outside the window can hold more information, which tools retrieve as needed.

Evaluation framework

Representative tasks and expected behavior, the application being evaluated, a process that runs and assesses cases, and a report people can inspect. It makes engineering performance visible. See PDD.

Fine tuning

Additional training that changes the weights of an existing model for a task or domain. It differs from supplying instructions, retrieved sources or tools at run time. Prolego advises against fine-tuning unless it is necessary, given its cost, complexity and the pace of model improvement. See model choice.

Forward Deployed Engineer (FDE)

An engineer who works directly with the client's business experts and technical teams to understand the operation, help choose what to build, and build and operate the AI system in the client's environment. Prolego's AI systems engineers work in this model, connecting strategic judgment with delivery, evaluation and continuing improvement. See our working model and engagement responsibilities.

Hallucination

Generated content that is false or unsupported. Business failure also includes omissions, obsolete sources, misapplied rules and unauthorized actions. See failure handling.

Operations Intelligence

Operations Intelligence is the collection of systems that enables AI to run reliably in the business. It makes company information, guidance, tools, permissions, evaluation and monitoring available across applications. It sits above existing source systems and is developed through actual workflows. See the architecture.

Operational agreement

Agreement among the company's experts on which sources govern and what the AI should do in real situations. Capturing institutional knowledge and resolving different expectations often set the pace of deployment. See what sets the pace.

Performance-Driven Development (PDD)

Prolego's evolving methodology for building reliability into AI systems from the start: getting organizational consensus on what the AI should do during development, then ensuring it does it correctly in production. Performance measures and expert feedback guide the engineering and pull the biggest risk forward. See the method.

Plugin

A package of related skills and connections installed in a selected Workbench. It exposes maintained tasks through a shared interface while Operations Intelligence supplies domain sources and reliability capabilities. See architecture.

LLM auditor or SME agent

An AI reviewer trained and calibrated through expert review to flag problems in live outputs. Experts inspect samples, exceptions and scoring and improve its guidance. See production reliability.

Developer agent

An AI tool that groups production issues and proposes fixes for a developer to review, evaluate and release. See the improvement loop.

Retrieval augmented generation or RAG

A pattern that supplies relevant source material to a model so it can answer or perform a task with that context. Retrieval is one component of the application; source meaning, evaluation and authority still need engineering. See reliability engineering.

Skill

A defined task with inputs, an intended output, checks and review responsibilities. A skill may use instructions, sources and tools. Shared skills need maintained scope and ownership. See Operations Intelligence.

The 10x within two years screen

Prolego's screen for the ambition of opportunities it pursues. It challenges what the business could become. It is not an achieved result, a financial forecast or a promise. See opportunity selection.