Process Automation

What is process automation?

Process automation is the use of software to carry out business processes — whether entire end-to-end workflows or individual steps within them — with little or no human intervention. Instead of a person manually performing each task, the process is described once and then executed by a system, reliably and repeatably, whenever its conditions are met.

Process automation sits alongside related disciplines: business process modeling captures how a process should work, process mining discovers how it actually works, and process automation is concerned with making the process run by itself. Robotic process automation (RPA) is one well-known approach, but it is only a part of the wider picture.

Automating whole processes vs. individual steps

Automation can be applied at different granularities:

• →

  Whole processes: an entire workflow is automated from a starting condition through to completion — for example, taking an incoming order all the way through validation, fulfilment, invoicing, and notification.

• →

  Individual steps: a single repetitive task within an otherwise human-driven process is automated, such as extracting fields from a document, enriching a record, or posting an update to another system.

Most real-world automations are a blend of the two: humans handle judgement-heavy steps while software handles the deterministic, high-volume, or error-prone ones.

Kinds of automated step

The steps that make up an automated process can take several forms:

• →

  Deterministic actions: rules-based operations such as transforming data, performing a calculation, or moving a record between states.

• →

  Integration actions: reading from or writing to external systems and tools — see integrations.

• →

  RPA-style actions: software “bots” that mimic the clicks and keystrokes a human would make in applications that lack a proper API.

• →

  Agentic / AI steps: steps powered by generative AI — and sometimes by multi-agent systems — that handle tasks requiring interpretation, summarization, or decision-making rather than fixed rules.

Process automation in HASH

In HASH, processes are automated using Flows. A Flow is composed of building blocks that together describe when and how work happens:

• →

  Triggers start a Flow — for example on a schedule, in response to an external event, or when an entity of a particular type is created or changed.

• →

  Actions are the steps a Flow performs, which may be deterministic operations, integrations, or AI-driven steps.

• →

  Filters decide whether a Flow (or a branch of it) should proceed, based on conditions evaluated against the data.

• →

  Waits pause a Flow until a condition is met, a delay elapses, or human input is provided.

Because Flows read from and write to the HASH knowledge graph, every automated step operates against a shared, typed representation of the world rather than against disconnected data. This means an automation can react to the current state of any entity, update it, and create new entities and links — all while the graph keeps a complete, traceable history of what changed and why.

From observed processes to automation

A particularly powerful pattern is to derive automations from processes that already exist. Activity captured through process mining can reveal the real steps people take, which can then be formalized as a business process model and, where appropriate, automated as a Flow. Before being rolled out, candidate automations can be tested against a digital twin of the operation to estimate their impact and surface edge cases. Grounding automation in observed reality — and running it against the knowledge graph — helps ensure that what gets automated reflects how the organization actually works, not just how it is assumed to.