presentsUNAIVERSE
The UN AI timeline

Learn AI /AI Agents and Automation /Workflow

Workflow

Also called: pipeline, process.

Detail

In five seconds

What it is

A workflow is an ordered set of steps that moves information or tasks from input to result, often combining software, people and checks.

In plain words

A workflow is the organised route a task follows from start to finish.

An everyday example

Preparing a parcel follows a workflow. First, you check the destination and what is being sent. The contents determine the next step: a fragile item needs protective wrapping, while an ordinary item may not. You then seal and weigh the parcel, choose a delivery option, pay, attach the label and hand it over. Each completed step supplies what the next step needs, and a missing address or an overweight package sends the process down a different path.

Where the story stops working

This analogy uses one person following a study routine, so it can look simpler than a real AI workflow. Actual workflows may run automatically across several systems, use APIs to pass data between tools, branch in many directions, and handle errors or approvals. Also, a student can reflect in ways a software step does not; the analogy only shows order and handoff, not human judgement or machine limitations.

Remember thisA workflow is the whole route from input to result, including sequence, branches, tool use, human checks and the system where the approved output is kept.

Understand it

Understand it

Why it matters

Non-specialists often hear impressive claims about AI systems, but the practical value usually depends on the workflow around the model. A well-designed workflow can reduce delays, make handoffs clearer, and add review points where mistakes would be costly. A poor workflow can spread errors faster, hide who is responsible, or automate the wrong step. If you can recognise a workflow, you can ask useful questions: What is the input? What are the steps? Which tools are called? Where does a human check happen? What comes out, and who is accountable for it? Those questions help with procurement, risk assessment, quality control and daily use.

How a workflow turns an input into an outcome

After viewing this, a reader should be able to explain that a workflow is a sequence of linked steps, sometimes with a human check, not just one model response.

  1. 01

    1. Input arrives

    A task starts with something concrete, such as an email, form, image or request.

    Draw a labelled box 'Input' with small icons for document, email and image; add an arrow to the next box.

  2. 02

    2. First processing step

    The system extracts or organises useful details, such as names, dates, categories or questions.

    Draw a box 'Extract key details' with text snippets flowing in and structured fields flowing out.

  3. 03

    3. Tool or service step

    The workflow may call another system, such as a database, calendar or map, to add information or perform an action.

    Draw a box 'Use tool/service' with a small database icon and a connector line to show outside software.

  4. 04

    4. Review or rule check

    A rule or a human can check the result before it continues, especially if confidence is low or the case is sensitive.

    Draw a decision-style box 'Check' with two small labels: 'pass' forward and 'needs review' upward to a human icon, then back.

  5. 05

    5. Output and record

    The workflow produces a result, such as a draft, alert, ticket update or saved record.

    Draw a final box 'Output' with a small archive icon to show the result can also be stored.

Show a left-to-right step diagram with five connected boxes. Start with an input, move through analysis and tool use, include a review gate, and end with an output. Read from left to right to see how each stage passes something to the next stage.

On hover or tap, each step reveals a short example of what goes in and what comes out at that stage.

Describe this without the diagram

The diagram is a five-step sequence read from left to right. Step 1 is Input arrives: a document, email or image starts the task. Step 2 is First processing step: the system extracts key details and turns unstructured material into organised fields. Step 3 is Tool or service step: the workflow calls another system, such as a database, to fetch information or do an action. Step 4 is Review or rule check: either a rule or a person checks whether the result is acceptable; if not, it goes to human review before returning. Step 5 is Output and record: the workflow produces a draft, alert or updated record and may save it. The main message is that a workflow is a chain of stages with handoffs and checks.

With reduced motion: Show the same five boxes and arrows as a static diagram, with the review path drawn visibly instead of animated.

A workflow is the path a task follows from start to finish. Something goes in, such as a request, a document or a data file. A series of steps then happens in a chosen order. A result comes out, such as a summary, a decision draft or a completed record.

In AI systems, a workflow often mixes model output with other actions. For example, a system might receive an email, extract key details, call a tool to look up a reference number, ask a person to review the draft, and then save the final version. A tool here means another piece of software the system can use, such as a calendar, database or mapping service.

The key idea is sequence and structure. Each step has a purpose, and a step's output can become an input to another step. Some workflows are simple and fixed. Others include conditions, such as "if the confidence is low, send to a human reviewer".

People often confuse workflow with an AI agent. They are related but not the same. A workflow is the organised process. An AI agent is one kind of component that may operate inside that process. A workflow can exist without an agent, and an agent can be only one step inside a larger workflow.

How “Sending a parcel” maps onto the real thing

In the storyIn the AI system
answer sheet and notebookinputs and records used at different steps
mark, sort, practise, teacher checkordered stages in a workflow
choosing weakest topicsa conditional step based on earlier results
action plan at the endfinal output of the workflow

A worked example

An information management officer in Dakar uses a workflow to process side-event concept notes before the programme is circulated.

Starting from: Concept notes containing proposed titles, organisers, preferred time slots and room requirements.

  1. The workflow extracts the required fields from each concept note and places them in a standard record.
  2. A calendar tool checks the requested slot and room availability.
  3. Rules route complete requests to a draft schedule. Requests with missing information or a clash instead produce a query for the organiser.
  4. The focal point reviews the proposed allocations and any flagged exceptions.
  5. After approval, the schedule is saved to the event portal and notification drafts are prepared.

Result: The focal point receives a proposed schedule, a separate list of unresolved requests and draft notifications. Nothing is published until the allocations have been reviewed and approved.

In a United Nations workplace

A duty officer in a regional bureau is preparing a delegation briefing for clearance before close of business while replies from capitals are still pending. An approved workflow extracts requested changes from incoming messages, compares them with the cleared briefing note and updates a status table. If a reply affects an agreed line, the draft is routed to the desk officer and chief of section. Otherwise, it proceeds to the front office for final clearance and filing.

The extraction is only one stage. The workflow controls the order of work, branches according to what the reply contains, brings in the appropriate reviewer and preserves the cleared result in the file.

Before you try thisIncoming messages and draft briefing lines may contain confidential or politically sensitive material. They must not be pasted into an external or otherwise unapproved AI tool. The workflow must preserve access controls and version history, and the responsible officer must verify and clear the briefing before circulation.

Where you may hear it

  • A vendor demo to an ICT governance board
  • A procurement unit meeting on routing approvals
  • A country office workplanning session
  • A service desk review after a ticket misses its deadline
  • An internal audit walkthrough of a digital process

Avoid confusion

Avoid confusion

Often confused with

Common misconceptions

  • Common misconception: A workflow is just another word for an AI agent.

    Correction: They overlap, but they are not the same. A workflow is the overall sequence of steps. An AI agent is one possible component that may act within those steps. Many workflows use no agent at all and simply follow fixed rules or tool calls.

  • Common misconception: Once a workflow is automated, it no longer needs people.

    Correction: Automation can remove repetitive work, but many real tasks still need human review, especially when cases are unusual, high-risk or sensitive. Good workflows usually define where a person must check, approve or correct the output.

  • Common misconception: If each step works reasonably well, the whole workflow will be reliable.

    Correction: Errors can accumulate across steps. A small extraction mistake can lead to the wrong tool call, which then produces a misleading final output. Reliability depends on the full chain, including checks, fallback rules and logging.

Quick check

Which description best matches a workflow in an AI system?

Go deeper

Go deeper

The simple version

A workflow is the organised route a task follows from start to finish.

The technical version

In AI and automation, a workflow is a structured sequence of operations that transforms an input into an output through staged processing. Steps may include prompt-based generation, data extraction, tool use, API calls, rule-based routing, human approval, memory updates and persistence to a system of record. The output of one stage becomes the input to the next. Workflows may be linear or conditional, and they are often implemented through orchestration logic that controls order, branching, retries, permissions and error handling. This is different from a single model call, which produces one response without managing the wider process around it.

Reviewed external resources

  • Primary sourceModel Context Protocol (opens in a new tab)Model Context Protocol

    The Model Context Protocol material suits readers who want a more technical view of how tools and systems can connect inside a workflow.

    Deeperlink checked 2026-07-31

  • EncyclopaediaWikipedia search: Workflow (opens in a new tab)Wikipedia

    A Wikipedia search results page helps readers find broad background on the term from multiple related entries, and suits beginners who want orientation rather than one fixed definition.

    search results, not one pagelink checked 2026-07-31

  • ArticleBuilding effective agents (opens in a new tab)Anthropic

    Anthropic's page on agents is useful for readers who want to see how workflows relate to agent-style AI systems in practice.

    Intermediatelink checked 2026-07-31

  • VideoVideo explanations: Workflow (opens in a new tab)YouTube search

    A YouTube search results page suits visual learners who want to compare different video explanations of workflows, but it is a search page rather than one vetted lesson.

    search results, not one pagelink checked 2026-07-31

  • DiscussionCommunity discussion: Workflow (opens in a new tab)Reddit search

    A Reddit search results page lets readers see common questions, confusions and practical experiences about workflows, and suits people who want informal discussion with caution about accuracy.

    search results, not one pagelink checked 2026-07-31

Where this came from

Drafted with a language model through an offline pipeline, checked against this site's schema, and every external link above fetched to see whether it answers. Treat it as a starting point and check anything that matters against the primary sources. How this site is built and what that means.

The one thing to remember

A workflow is the whole route from input to result, including sequence, branches, tool use, human checks and the system where the approved output is kept.