Glossary

Software, ERP and AI terms, defined plainly.

37 definitions of the terms that come up in enterprise software work. Each one is written to stand on its own, without marketing attached.

Trusted by 500+ Clients

Enterprise systems

ERP: Enterprise resource planning (ERP) is a single system of record that runs an organisation's core operations, such as finance, inventory, purchasing, manufacturing and HR, on one shared database.The defining characteristic is the shared database, not the feature list. Separate best-of-breed tools can cover the same functions, but each keeps its own copy of the data, so the reconciliation work moves to the people using them.See also: Odoo, ERP implementation, System of record
Odoo: Odoo is an open-source ERP and business application suite built from modular apps that share one database and one data model.Odoo is distributed in a free Community edition and a paid Enterprise edition. Because the source is available, an Odoo implementation can be extended in Python rather than only configured, which is the main practical difference from closed SaaS ERPs.See also: ERP, ERP implementation
ERP implementation: An ERP implementation is the project of configuring an ERP system to an organisation's processes, migrating its existing data into it, and moving users onto it.Most of the effort is usually process and data work rather than software work. The two failure modes are customising the software to preserve processes that should have changed, and changing processes the business genuinely needed to keep.See also: ERP, Odoo, Data migration
System of record: A system of record is the single authoritative source for a given piece of data, which every other system copies from rather than competes with.See also: ERP

Delivery

Data migration: Data migration is the process of moving data from one system into another, including mapping it to the target system's structure and correcting what does not fit.Migration exposes data quality problems that were invisible while the data sat in the old system, because the new system enforces constraints the old one did not. Budget for cleaning, not just transfer.See also: ERP implementation
Minimum viable product (MVP): A minimum viable product is the smallest version of a product that can be released to real users to test whether the core assumption behind it holds.The purpose is to learn, which means an MVP is defined by the question it answers, not by a reduced feature count.See also: Discovery, Product engineering
Discovery: Discovery is the phase before build in which the problem, users, constraints and success measures are established well enough to scope the work.See also: Minimum viable product (MVP)
Product engineering: Product engineering is software development organised around a product's outcomes and ongoing evolution, rather than around delivering a fixed scope and handing it over.See also: Minimum viable product (MVP), Staff augmentation
Staff augmentation: Staff augmentation is an engagement model in which external engineers join and are directed by the client's own team, rather than delivering a separately managed project.The client keeps technical direction and accountability for the outcome. That is the main difference from managed delivery, and the main reason it fails when the client has no capacity to direct the work.See also: Product engineering, Dedicated team
Dedicated team: A dedicated team is a group of external engineers assigned exclusively to one client over an extended period, usually with their own delivery management.See also: Staff augmentation
Service level agreement (SLA): A service level agreement is a contractual commitment to specific, measurable service levels, such as uptime or response time, usually with defined consequences for missing them.An SLA without a stated measurement method and a remedy is a statement of intent, not an agreement.
Uptime: Uptime is the percentage of a period during which a system was available and functioning.The figure is meaningless without its window and its definition of available: 99.9% measured monthly permits about 43 minutes of downtime, and 99.9% measured annually permits nearly nine hours.See also: Service level agreement (SLA)

AI

Large language model (LLM): A large language model is a neural network trained on large volumes of text to predict likely continuations, which lets it generate and transform language across many tasks without task-specific training.An LLM has no built-in access to your data and no inherent notion of truth. Both are supplied by the system built around it, which is why retrieval and evaluation matter more than model choice for most business applications.See also: Retrieval-augmented generation (RAG), AI agent, Hallucination, Fine-tuning
Retrieval-augmented generation (RAG): Retrieval-augmented generation is a technique that retrieves relevant documents from a knowledge source and supplies them to a language model as context, so answers are grounded in specific material rather than in the model's training data alone.RAG is the usual answer to 'the model does not know our internal information'. It also makes answers auditable, because the retrieved passages can be cited. Retrieval quality, not model size, is normally the limiting factor.See also: Large language model (LLM), Vector database, Hallucination, Embedding
Embedding: An embedding is a numeric vector representing a piece of text, image or other content, positioned so that items with similar meaning sit close together in the vector space.See also: Vector database, Retrieval-augmented generation (RAG)
Vector database: A vector database stores embeddings and retrieves the entries closest to a query vector, which makes it possible to search by meaning rather than by exact keyword.See also: Embedding, Retrieval-augmented generation (RAG)
AI agent: An AI agent is a system that uses a language model to decide which actions to take, calls tools or APIs to carry them out, and repeats until a goal is met or it stops.The distinction from a chatbot is the ability to act, not the ability to converse. That is also what raises the stakes: an agent with write access can cause real effects, so permissions and audit trails matter more than prompt quality.See also: Large language model (LLM), Agentic AI, Human in the loop
Agentic AI: Agentic AI describes systems that pursue multi-step goals with a degree of autonomy, choosing their own sequence of actions rather than following a fixed script.See also: AI agent, Human in the loop
Hallucination: A hallucination is output from a language model that is fluent and confident but factually wrong or unsupported by any source.Hallucination is a property of how these models generate text, not a bug that gets patched. It is managed by grounding answers in retrieved sources, constraining what the system is allowed to assert, and evaluating outputs, rather than eliminated.See also: Retrieval-augmented generation (RAG), Large language model (LLM), Evaluation (evals)
Fine-tuning: Fine-tuning is further training of an existing model on a specific dataset so it adapts to a particular style, format or task.Fine-tuning changes behaviour, not knowledge. If the problem is that the model does not know your facts, retrieval is the fix; if the problem is that it does not respond in the form you need, fine-tuning may be.See also: Large language model (LLM), Retrieval-augmented generation (RAG)
Evaluation (evals): Evaluation is the practice of measuring an AI system's output against a fixed set of test cases so that changes can be compared objectively rather than by impression.Without evals, every prompt or model change is a guess, because the failures are probabilistic and will not reliably show up in manual spot checks.See also: Large language model (LLM), Hallucination
Human in the loop: Human in the loop is a design in which a person reviews or approves an automated system's output before it takes effect.See also: AI agent, Agentic AI

Engineering

Technical debt: Technical debt is the future cost created by choosing an implementation that is quicker now but harder to change later.Debt taken deliberately to hit a date is a legitimate trade. Debt accumulated by accident is just damage. The distinction is whether anyone recorded the decision and the cost.See also: Legacy modernisation, Refactoring
Legacy modernisation: Legacy modernisation is the work of bringing an older system onto current technology and practices without losing the business rules it encodes.The risk is rarely the technology. It is that the old system's behaviour was never fully documented, so a rewrite silently drops rules the business still depends on.See also: Technical debt, Strangler pattern, Refactoring
Strangler pattern: The strangler pattern is a modernisation approach in which new functionality is built alongside an existing system and traffic is moved across piece by piece until the old system can be retired.It is preferred over a single cut-over because each step is small enough to reverse, which matters when the old system's behaviour is only partly understood.See also: Legacy modernisation
Refactoring: Refactoring is changing the internal structure of code without changing its external behaviour.See also: Technical debt
API: An application programming interface is a defined contract through which one piece of software requests data or actions from another.See also: REST API, Webhook, Middleware
REST API: A REST API is an API that exposes resources at URLs and operates on them using standard HTTP methods such as GET, POST, PUT and DELETE.See also: API, Webhook
Webhook: A webhook is an HTTP callback that one system sends to another when an event occurs, so the receiver is notified instead of having to poll for changes.See also: API, REST API
Middleware: Middleware is software that sits between systems and handles the translation, routing or queueing needed for them to work together.See also: API, System integration
System integration: System integration is the work of making separate applications exchange data reliably and agree on what that data means.The hard part is usually semantic rather than technical: two systems that both have a 'customer' record rarely mean exactly the same thing by it.See also: Middleware, API, System of record

Mobile

Cross-platform development: Cross-platform development is building an application from one shared codebase that runs on multiple operating systems, such as both iOS and Android.See also: Native app, Flutter, React Native
Native app: A native app is built with a platform's own languages and tools, such as Swift for iOS or Kotlin for Android, and runs directly against that platform's APIs.See also: Cross-platform development
Flutter: Flutter is Google's open-source framework for building cross-platform applications from a single Dart codebase, rendering its own UI rather than using the platform's native controls.See also: Cross-platform development, React Native
React Native: React Native is an open-source framework for building mobile applications in JavaScript or TypeScript that render using the platform's own native UI components.See also: Cross-platform development, Flutter
Progressive web app (PWA): A progressive web app is a website built to behave like an installed application, with offline support, an app icon and background capabilities.See also: Cross-platform development

Commerce

Headless commerce: Headless commerce separates an ecommerce platform's storefront from its backend, so the customer-facing experience can be built independently and served through APIs.See also: API