Glossary
AI & Digital Transformation Glossary
Short, practical definitions for the terms that come up most often in our advisory work and across this site.
- AI Agent
- Unlike a chatbot that answers a single question, a software component with a defined role and permissions that can connect to systems like CRM/ERP/files and complete multi-step tasks, with or without human approval.
- API (Application Programming Interface)
- A standard interface that lets two different software systems send and receive data from each other. The basic building block of system integration.
- Audit Trail
- A record chain that lets you trace, after the fact, who changed which data or made which decision, and when. Often a regulatory requirement in public-sector and regulated industries.
- Digital Maturity
- An assessment of how effectively an organization uses digital tools across process, data, systems, people and governance. Produces separate findings per dimension, not one blended score.
- EBYS (Electronic Document Management System)
- The system used in Turkish public institutions to run official correspondence, documents and approval flows digitally. In municipalities it is expected to integrate with e-municipality applications.
- Integration Layer
- Middleware/architecture that builds a secure, traceable data flow between existing systems without replacing them. Usually faster and lower-risk than a system replacement.
- ETL (Extract, Transform, Load)
- Shorthand for pulling data from a source (extract), reshaping it as needed (transform) and loading it into a target system (load). Common in reporting and integration projects.
- Guardrail
- A control mechanism that limits what an AI model or agent can output using predefined rules — for example, blocking sensitive-data disclosure, unauthorized actions or inappropriate content.
- Human-in-the-loop
- A working model where an AI-generated output — especially for critical steps like financial transactions or customer communication — is reviewed and approved by a person before it takes effect.
- KVKK (Turkish Personal Data Protection Law)
- Turkey's Law No. 6698 on the Protection of Personal Data. Regulates how personal and special-category data may be processed, stored and transferred — a requirement to design around from the start when building AI systems.
- Low-code / No-code
- A category of tools that let you build business apps and automations with little or no traditional code. Good for a fast start, but usually needs to evolve into custom software as scale grows.
- Prompt
- The written instruction given to an AI model describing what it should do. In enterprise use, prompts should be version-controlled and auditable.
- RAG (Retrieval-Augmented Generation)
- An architecture where an AI model retrieves relevant information from an organization's own documents and data sources before generating a response, so answers are grounded in current, accurate company data rather than general model knowledge.
- RPA (Robotic Process Automation)
- Software robots performing rule-based, repetitive digital tasks — like copying data or filling forms — in place of a person. Can break when inputs change, and works best alongside AI automation rather than instead of it.
- System of Record
- The design decision that determines which system is treated as authoritative when the same information exists in multiple places. Usually the first step in an integration project.
- Local / On-Premise Model
- An AI model run on an organization's own infrastructure (using tools like Ollama) instead of a cloud API. Prompts and documents never leave the organization — preferred for sensitive data, predictable cost, or offline scenarios.
Does one of these terms relate to what you need?
Let's talk briefly about which service is the right starting point for your organization.