Business automation glossary
Search for an automation glossary and most of what you find is about factories: robots, sensors, and programmable controllers on a production line.
This one is the other kind, the software automation that runs a business, moving information between systems, handling routine steps, and increasingly making the judgement calls that once needed a person.
It is a plain English business automation glossary written for the manager or owner weighing up what to automate, not for an industrial engineer. Every term has a short, self contained definition and a note on what it means in practice.
The terms are grouped so related ideas sit together: the core building blocks first, then the confusing family of automation types compared side by side, then the newer AI native vocabulary that older glossaries miss. You do not need to read it in order.
Where a term has a fuller guide behind it, the entry links through, so the glossary doubles as a map of where to go next.
How to use this glossary
Keep this to hand when you are comparing tools or listening to a supplier, and look up any term before you let it shape a decision.
The automation types section is the one most worth reading in full, because the difference between RPA, BPA, BPM, and workflow automation is exactly where buyers get talked into the wrong thing. Knowing which is which lets you tell whether a pitch matches your actual problem.
One term that always deserves attention is governance. Where automation touches personal data, especially any process that makes decisions about people, UK data protection rules apply, and the Information Commissioner's Office guidance on automated decision making and profiling is the authoritative UK reference.
Building automation without regard for it is a risk, not a shortcut.
These are the pieces almost every automation is assembled from. Understanding them makes the rest of the vocabulary, and most tools' marketing, far easier to read.
This is the cluster buyers trip over most: a soup of acronyms that overlap and are often used interchangeably in marketing. Here is what each actually means, and how they differ. For the fuller picture, our guides to business process automation and workflow automation go deeper.
This vocabulary is newer and moves fastest. It is largely absent from older automation glossaries, and it is where much of the value now sits. Our AI agent glossary covers the agent side in full.
A
- Agentic automationAI native automation terms
- Agentic automation is automation driven by AI agents that can decide how to reach a goal across multiple steps, adapting to what they find, rather than following a fixed script. It is the leap from "do exactly these steps" to "achieve this outcome". It handles varied, judgement heavy work that rigid automation cannot, though it needs careful guardrails and oversight to run safely.
- APICore automation building blocks
- An API, or application programming interface, is a defined way for one piece of software to request data or actions from another. It is the clean, supported route for systems to connect, as opposed to imitating a human clicking around. When a tool "has an API", it can usually be integrated into an automated workflow directly and reliably.
- Approval stepRunning automation in practice
- An approval step is a deliberate pause where a person authorises the work before it continues. Placed well it costs almost nothing and removes most of the risk of automating a consequential process. Placed badly, on every case rather than the ones that matter, it recreates the bottleneck you were trying to remove.
- Audit trailRunning automation in practice
- An audit trail is the record of what the automation did, when, on what input, and what it produced. It is what turns an automated process from a black box into something you can inspect after the fact. For anything regulated or financial, treat it as a requirement rather than a feature.
- AutomationCore automation building blocks
- Automation is using software to carry out a task or process that a person would otherwise do by hand, so it runs faster, more consistently, and without constant attention. It ranges from a single automated step to an entire process running end to end. The business question is rarely whether something can be automated, but whether automating it is worth the effort.
B
- Batch processingCore automation building blocks
- Batch processing collects work and handles it in groups on a schedule, rather than one item at a time as it arrives. It is cheaper and simpler where immediacy is not required, such as overnight reconciliation. The trade off is delay, so it suits back office work and suits customer facing work poorly.
- BotCore automation building blocks
- A bot is a software program that performs automated tasks, often ones a person would otherwise do through a screen, such as copying data between systems or responding to a routine request. The word is used loosely, covering everything from a simple script to a robotic process automation robot. In business automation it usually means the latter.
- Business process automation (BPA)Automation types compared
- Business process automation is the automation of a whole business process end to end, rather than a single task, often coordinating several systems and steps towards a business outcome. It is broader than RPA, which is one possible tool within it. BPA is about the process as a whole, from trigger to result, not just the mechanical steps in the middle.
- Business process management (BPM)Automation types compared
- Business process management is the wider discipline of designing, documenting, measuring, and improving how a business's processes run, whether or not they are automated. Automation is one thing BPM might lead to, but the practice itself is about understanding and refining the process first. Good automation usually rests on the mapping and discipline that BPM provides.
C
- Condition and actionCore automation building blocks
- A condition is a rule the automation checks before deciding what to do next, and an action is the step it then takes. Together with a trigger they form the common trigger, condition, action pattern: when this happens, if this is true, do that. Most rules based automation is built from chains of these three simple pieces.
- ConnectorSystems and integration
- A connector is a prebuilt piece of integration for one specific system, saving you from writing that plumbing yourself. Coverage of the popular systems is excellent, which is exactly why the awkward parts of a business are the ones with no connector. Check for the systems you actually run rather than the ones on the marketing page.
- Cycle timeRunning automation in practice
- Cycle time is how long one item takes to travel the whole process from start to finish. It is usually the number that matters to a customer, and it is usually far longer than the sum of the actual work, because most of it is waiting. Automation earns its keep by removing waiting, not by speeding up typing.
E
- EscalationRunning automation in practice
- Escalation is routing a case to a person, or to a more senior one, when the automation cannot or should not decide. A clear escalation path is what makes an automated process safe to run at scale. It needs an owner, because work escalated to nobody in particular is work that stops.
- Event driven automationAI native automation terms
- Event driven automation runs in response to things happening in real time, an order placed, a file uploaded, a threshold crossed, rather than on a fixed schedule or a manual start. It makes a business more responsive, acting the moment a trigger occurs. Webhooks are a common mechanism for it, pushing events between systems as they happen.
- ExceptionCore automation building blocks
- An exception is any case that does not follow the normal path: a missing document, an unusual amount, a customer who replies with something nobody anticipated. Exceptions are where automation projects succeed or fail, because the happy path is the easy part. A build that has not been designed around its exceptions is not finished.
- Exception handlingCore automation building blocks
- Exception handling is what the system does when it meets a case it cannot complete. Good handling stops, preserves everything it knows, and hands a person the context to finish the job. Bad handling either guesses or fails silently, and silent failure is considerably more expensive because nobody finds out for weeks.
H
- HandoffRunning automation in practice
- A handoff is the moment work passes from one person, team or system to another. Handoffs are where things get dropped, duplicated or delayed, which is why mapping them is the first useful thing a swimlane diagram does. Removing a handoff is often worth more than automating a step.
- Human in the loopAI native automation terms
- Human in the loop describes an automated process that pauses for a person to review, approve, or correct something at a key point before continuing. It is a deliberate design choice, not a failure of automation, and it is essential wherever the stakes, the ambiguity, or the regulation demand human judgement. Most trustworthy business automation keeps a human in the loop where it matters.
- HyperautomationAutomation types compared
- Hyperautomation is the coordinated use of several automation technologies together, such as RPA, workflow automation, AI, and process mining, to automate as much of a business as is worthwhile, not just isolated tasks. It is more a strategy than a single tool. In practice it describes an organisation treating automation as a joined up programme rather than a series of one off fixes.
I
- IdempotencyCore automation building blocks
- An operation is idempotent when running it twice has the same effect as running it once. It matters because retries are normal in automation, and a payment or an invoice created twice is a real problem. Any step that writes to a system of record should be built this way.
- IntegrationCore automation building blocks
- An integration is a connection that lets two systems share data or trigger actions in each other automatically, so information flows between them without someone rekeying it. Integrations are what let a whole process run across several tools rather than stalling at each handoff. The quality of a business's integrations often decides how much it can realistically automate.
- Integration platform as a service (iPaaS)Systems and integration
- An integration platform as a service is a hosted product for building integrations without running the plumbing yourself. It suits common connections between mainstream systems and gets expensive or restrictive at the edges. It is a sensible floor to start on and a poor ceiling to aim for.
- Intelligent document processingSystems and integration
- Intelligent document processing combines OCR with models that understand layout and meaning, so a document can be read without being told in advance where each field sits. It is what allows a system to handle invoices from two hundred suppliers in two hundred formats. It is the practical answer to the limits of template based extraction.
L
- Legacy systemSystems and integration
- A legacy system is older software a business still depends on, typically with poor or no integration options. It is usually the real constraint on an automation project and the reason bespoke work beats a packaged tool. Replacing it is rarely on the table, so the automation has to reach it as it is.
- Low code and no codeAI native automation terms
- Low code and no code are platforms that let people build automations through visual, drag and drop interfaces instead of writing software, with low code allowing custom code where needed. They make automation accessible to non developers and are well suited to standard tasks. Genuinely bespoke, high volume, or high stakes processes often outgrow them and warrant a custom build.
M
- MiddlewareSystems and integration
- Middleware is software that sits between systems and passes work between them. It is where the logic of an integration usually lives: the mapping, the transformation, the retries. Treating it as a deliberate layer, rather than as scripts scattered across several machines, is what keeps an estate maintainable.
O
- Optical character recognition (OCR)Systems and integration
- Optical character recognition converts an image of text, such as a scan or a photograph, into text a computer can process. It is mature and reliable on clean documents and struggles with poor scans and handwriting. It is the first step in most document automation, and its accuracy sets the ceiling for everything after it.
- OrchestrationAI native automation terms
- In automation, orchestration is the layer that coordinates the pieces, deciding which system, step, or agent runs when, and passing information between them so a multi step process holds together as one. Individual automated steps are easy; making them cooperate reliably across a whole process is the hard part, and orchestration is where that reliability is engineered.
P
- PollingCore automation building blocks
- Polling is checking another system on a schedule to see whether anything has changed. It is the fallback when a system offers no way of announcing changes itself. It is less efficient than being notified and introduces delay equal to the gap between checks, which is why a webhook is preferable where one exists.
- ProcessCore automation building blocks
- A process is the whole sequence a piece of work travels through from trigger to completion, across every person and system it touches. It is the unit that matters, because delay and error collect at the joins rather than inside the steps. Map the process before automating any task within it.
Q
- QueueCore automation building blocks
- A queue holds work waiting to be processed, so that a burst of incoming items does not overwhelm the system handling them. It is what lets an automation absorb a Monday morning spike gracefully. It also gives you somewhere to look when you want to know what is currently backed up and why.
R
- Rate limitCore automation building blocks
- A rate limit is the cap a service places on how many requests you may make in a period. Exceed it and requests are refused, which is a common cause of automations that work in testing and break at volume. Designing for it means queuing and pacing work rather than firing everything at once.
- ReconciliationRunning automation in practice
- Reconciliation is comparing two records of the same thing and identifying where they disagree, such as a bank statement against a ledger. It is one of the most commonly automated finance tasks because it is high volume, rule shaped and tedious. The automation finds the differences; a person still decides what to do about them.
- Retry and backoffCore automation building blocks
- Retry is trying a failed step again; backoff is waiting progressively longer between attempts. Together they absorb the ordinary flakiness of networks and third party services without human involvement. Retrying immediately and endlessly turns a brief outage into a self inflicted flood of traffic.
- Robotic process automation (RPA)Automation types compared
- Robotic process automation uses software robots to imitate the exact clicks and keystrokes a person makes across applications, following fixed rules. It suits stable, repetitive, screen based tasks and needs no changes to the underlying systems. Its weakness is fragility: when a screen or format changes, the robot breaks, which is why rigid RPA is increasingly paired with or replaced by more adaptable methods.
S
- Screen scrapingSystems and integration
- Screen scraping is driving software through its user interface, reading what is on screen, because no proper interface for machines exists. It is how automation reaches older systems that were never designed to be connected. It works, and it is brittle: a layout change can break it, so it is a last resort rather than a first choice.
- Service level agreement (SLA)Running automation in practice
- A service level agreement is a commitment about how quickly work will be handled. Automation is often justified by SLA pressure, because machines do not have Friday afternoons. Measuring against it before and after a build is the most honest way to show whether the build did anything.
- Single source of truthSystems and integration
- A single source of truth is the principle that each fact has one authoritative home, with everything else referring to it rather than keeping a private copy. It is the antidote to the spreadsheet that quietly diverges from the CRM. Most data problems in a growing business are a missing version of this.
- Standard operating procedure (SOP)Running automation in practice
- A standard operating procedure is the written description of how a process should be carried out. It is a useful starting point for automation and a dangerous one to trust, because the documented process and the real one drift apart over time. Always verify it against how the work is actually done.
- Straight through processingCore automation building blocks
- Straight through processing is when a transaction or request runs from start to finish entirely automatically, with no manual intervention at any step. It is the goal for high volume, well understood processes such as standard invoices or routine approvals. Reaching it usually requires clean data and clear rules, which is why it is a target to design towards rather than assume.
- SwimlaneRunning automation in practice
- A swimlane is a way of drawing a process where each lane belongs to one actor, so work is shown moving between them. It is the right format the moment a process crosses more than one team, because it makes accountability and handoffs impossible to hide. For a single team process it is more structure than you need.
- System of recordSystems and integration
- A system of record is the place a particular kind of data officially lives, and against which every other copy should be reconciled. Naming it, per data type, prevents the common failure where three systems disagree and nobody can say which is right. It is a governance decision more than a technical one.
T
- TaskCore automation building blocks
- A task is a single unit of work inside a process: read the email, extract the reference, update the record. Automation is usually sold at the level of the task and only pays off at the level of the process. Automating one task in a chain of eight rarely changes how long the chain takes.
- ThroughputRunning automation in practice
- Throughput is how many items a process completes in a period. It is the capacity question rather than the speed question, and the two come apart: a process can be fast per item and still have low throughput because everything queues behind one person. Automation typically improves throughput first.
- TriggerCore automation building blocks
- A trigger is the event that starts an automation running, such as an email arriving, a form being submitted, or a set time being reached. Every automation begins with one. Choosing the right trigger is what makes an automation fire exactly when it should, rather than too often, too late, or on the wrong thing entirely.
V
- ValidationRunning automation in practice
- Validation is checking that data meets the rules before it is accepted: the date is a date, the total matches the lines, the reference exists. It is the cheapest possible defence against bad data spreading through connected systems. Most reconciliation work exists because validation was skipped upstream.
W
- WebhookCore automation building blocks
- A webhook is a way for one system to notify another the instant something happens, by automatically sending a message to a set address. Where a traditional integration might check for updates on a schedule, a webhook pushes them in real time. It is the mechanism behind many instant, event driven automations, such as a sale immediately updating your accounts.
- WorkflowCore automation building blocks
- A workflow is the defined sequence of steps a piece of work moves through from start to finish, such as an order going from placed to picked to shipped to invoiced. Mapping a workflow clearly is the groundwork for automating it, because you cannot reliably automate a process you have not first described step by step.
- Workflow automationAutomation types compared
- Workflow automation is software that runs a defined sequence of steps across systems automatically, moving work along and taking actions with little manual effort. It sits between a single automated task and full process automation, and it is often where businesses start. The addition of AI now lets workflow automation handle the judgement steps that once forced a person into the loop.
Nothing matches that. Clear the search to see the full list.
Common questions about automation jargon
A business automation glossary is a plain language reference to the terms used in automating business processes, from RPA and workflow automation to AI native concepts like agentic automation. This one groups terms by theme, defines each in a short self contained entry, and links to fuller guides, so a non specialist can understand and compare what different tools actually do.
Because automation vocabulary is scattered, overlapping, and often used loosely in marketing, which makes it easy to buy the wrong thing. A clear glossary lets you tell RPA from BPA from workflow automation, understand what a supplier is really offering, and judge whether a tool fits your problem rather than trusting the label on it.
Robotic process automation imitates a person's clicks to automate a task. Business process automation automates a whole process end to end, often using several tools. Business process management is the wider discipline of designing and improving processes, automated or not. RPA is a tool, BPA is the scope, and BPM is the practice around both.
Hyperautomation is the coordinated use of multiple automation technologies together, such as RPA, workflow automation, AI, and process mining, to automate as much of a business as is worthwhile rather than isolated tasks. It describes an organisation treating automation as a joined up programme, not a strategy delivered by any single product despite how some are marketed.
A webhook updating your accounts the moment a sale completes; a workflow automation routing an incoming invoice for approval; an AI agent reconciling orders across systems and flagging only the mismatches. Each maps a glossary term to a concrete outcome, which is why knowing the vocabulary helps you picture what a tool will actually do for you.
You can compare tools and suppliers on their merits, spot when a pitch overstates what a product does, and brief your team with a shared language. It also clarifies the choices that matter, such as where a human should stay in the loop, which is often the difference between automation you can trust and automation you cannot.
Match the term to your actual problem. If the task is stable and screen based, RPA vocabulary applies; if it spans systems, look at workflow automation and integration; if it needs judgement, agentic automation. Understanding the distinctions stops you buying a rigid tool for a varied job, or an elaborate one for a simple task.
The main types are robotic process automation for imitating manual clicks, workflow automation for running defined sequences across systems, business process automation for whole processes end to end, and agentic automation for AI driven, judgement heavy work. Hyperautomation is the strategy of combining these, and business process management is the discipline of designing the processes underneath.
Weighing an automation proposal against these definitions
The point of a shared automation vocabulary is that it changes automation from a set of confusing acronyms into a set of choices you can weigh.
Once the terms are clear, the real question is which of your processes is actually worth automating, and by which method, and that answer depends on how your work runs, not on any definition.
It is easy to buy the wrong layer: rigid automation for a varied task, or an elaborate build for something simple rules would handle.
An automation audit is where we untangle that with you: we walk your real processes, name which are worth automating and which are not, and match each to the simplest method that will do it reliably.
Where a build is the right call, our workflow agents service is how we deliver it and keep it running as your work changes.
To go deeper, our guide to business process automation covers the process side, the complete workflow automation guide covers building workflows, and the AI glossary and AI agent glossary cover the AI vocabulary these terms increasingly overlap with.
The other glossaries