Map an AI-Enabled Child Welfare Intake

Design AI-assisted intake workflows and BA artifacts for a CCWIS modernization project.

Introduction

30 Second Summary

When someone calls a hotline to report concerns about a child, the intake worker on the other end has to simultaneously interview the caller, document details, search years of case history, and make a high-stakes screening decision. That is an enormous amount of cognitive load concentrated on a single person under time pressure.

In this project, you will play the role of a Business Analyst on a child welfare system modernization team. You will map the current intake process, identify where it breaks down, and design an AI-assisted workflow that supports workers without replacing their judgment on safety-critical decisions.

What You'll Build

Picture yourself presenting a complete artifact set to stakeholders at "Greenfield County DCS" and walking them through the full story: here is how intake works today, here is where workers struggle, here is how AI improves it, and here are the requirements that make it buildable.

By the end of this project, you'll have:

A swimlane process map of a child welfare hotline intake with annotated pain points pinpointing where workers lose time, introduce inconsistency, or face cognitive overload.
A redesigned To-Be workflow where AI handles narrative structuring, prior-case matching, and screening decision support while keeping humans in control of every safety-critical determination.
A set of user stories with testable acceptance criteria in Given/When/Then format, ready to hand off to a development team.
Secret Mission: Build a requirements traceability matrix connecting pain points to user stories to UAT scenarios, proving end-to-end coverage.

Are there any prerequisites?

None. This project uses draw.io (free, browser-based, no account required) and Google Docs for written artifacts. You do not need prior experience with process mapping or CCWIS systems. The guide covers all notation and domain context as you go.

Before We Start

Before We Start...

Before we dive in, let's lock in what you're about to build and why it matters to you.

Set Up draw.io and Learn Process Mapping Notation

Before you can map any process, you need two things: a diagramming tool and a shared visual vocabulary.

Without understanding what each shape means, your diagrams will confuse stakeholders rather than clarify the workflow. In this step, you will set up draw.io and learn the four shapes that make up process mapping notation.

In this step, get ready to:

Open draw.io and create a new swimlane diagram file.
Learn the four core shapes used in process mapping.
Configure four labeled swimlanes and verify your tool is working.

Open draw.io and create a new swimlane diagram

draw.io is a free, browser-based diagramming tool. It requires no account and saves files directly to your device.

Open a new browser tab and navigate to https://app.diagrams.net.
When prompted for a storage location, select Device. This saves your diagram locally with no account needed.
Click Create New Diagram.
Name your diagram CW-Intake-AsIs.
In the template selector, find the Basic template collection.
Select Swimlane Diagram, then click Create.

You should now see a starter swimlane diagram on your canvas with pre-labeled lanes and sample shapes inside them.

What is a swimlane diagram?

A swimlane diagram divides a process flow into horizontal (or vertical) lanes. Each lane represents a different actor or system responsible for the activities inside it.

When a connector crosses from one lane to another, it shows a handoff between actors. This makes it easy for stakeholders to see exactly who does what and where delays or miscommunication can happen.

Don't see the template selector?

If draw.io opened a blank canvas instead of showing templates, you may have dismissed the dialog too quickly. Close the current tab and navigate to https://app.diagrams.net again. The storage and template prompts appear on first load.

If you already have a blank diagram open, you can insert a template manually: click Arrange then Insert then Template and select the Swimlane Diagram from the Basic collection.

Still stuck? Help me get a swimlane template open in draw.io.

Learn the four shapes you need

Process mapping uses a small set of standardized shapes. Each shape signals a different type of step in the workflow. Stakeholders reading your diagram will expect these conventions.

Rectangle (rounded corners) represents an activity or task someone performs (e.g., "Document allegation narrative").
Diamond (rhombus) represents a decision point where the flow branches (e.g., "Does report meet statutory definition?").
Circle or oval represents the start and end points of the process.
Parallelogram represents data input or output (e.g., "Intake report entered in case management system").

Connectors (arrows) show the flow direction between shapes. When a connector crosses from one swimlane into another, it indicates a handoff between actors.

Where to find these shapes in draw.io

The General, Misc, and Advanced shape libraries are enabled by default in draw.io. You will find rectangles, diamonds, circles, and parallelograms in the left-hand shapes panel without enabling any additional libraries.

Set up your four swimlanes

The starter template has sample content you do not need. You will clear it, rename the lanes for this project's actors, and place test shapes to confirm everything works.

Select all the shapes inside the existing lanes (click and drag to select all, or press Cmd+A on macOS or Ctrl+A on Windows).
Press Delete or Backspace to remove the sample shapes. Leave the lane structure itself intact.

Accidentally deleted a lane?

Press Cmd+Z (macOS) or Ctrl+Z (Windows) to undo. You can undo multiple times to get back to the original template state.

Double-click the first lane's header text to edit it. Rename it to Reporter.
Double-click the second lane's header and rename it to Intake Specialist.
Double-click the third lane's header and rename it to Supervisor.
Double-click the fourth lane's header and rename it to Case Management System.

What if the template has fewer than four lanes?

If your template only shows two or three lanes, you can add more. Select the outer pool border (the rectangle surrounding all lanes), then in the Arrange tab on the right panel, click Insert Row After. Repeat until you have four lanes.

Now confirm the tool is working by placing test shapes and a connector.

From the shapes panel on the left, drag a rectangle into the Intake Specialist lane.
Drag a diamond shape into the same lane, to the right of the rectangle.
Hover over the rectangle until you see small blue direction arrows appear on its edges.
Click and drag from one of those blue arrows toward the diamond, then release on the diamond shape. A connector arrow should appear linking the two shapes.
Save your diagram by pressing Cmd+S (macOS) or Ctrl+S (Windows). Choose a location on your device when prompted.

Before you check your work, take a moment to predict: what should your canvas look like right now?

You should see four horizontal lanes labeled Reporter, Intake Specialist, Supervisor, and Case Management System. Inside the Intake Specialist lane, a rectangle and a diamond are connected by an arrow.

Connector not attaching to the diamond?

Make sure you start the drag from one of the blue direction arrows on the rectangle's edge, not from the center of the shape. When you hover over the target diamond, you should see it highlight in blue before you release.

If the arrow appears but floats disconnected, click the arrow's endpoint and drag it onto the diamond until the diamond highlights, then release.

Still having trouble? Help me connect two shapes with an arrow in draw.io.

Your diagramming tool is ready and you know the visual vocabulary. Next up, you will use these shapes to map a complete child welfare intake process from start to finish.

Map the As-Is Intake Process

Your draw.io canvas is ready with four labeled lanes and you have confirmed the tool works. Now you need to fill those lanes with the actual intake process as it happens today.

Before you can propose AI improvements, you need to document how the process actually works. This As-Is diagram becomes the foundation every stakeholder validates against. It also gives you concrete evidence of pain points you will address with AI later.

In this step, get ready to:

Map the reporter-to-intake specialist interaction on the left side of the diagram.
Map the screening decision and prior history check in the middle of the diagram.
Map the supervisory review and case assignment on the right side of the diagram.

Map the reporter and intake specialist interaction (left side of diagram)

Every intake starts the same way: someone calls a hotline to report suspected abuse or neglect, and an Intake Specialist takes over from there. You will map this initial handoff first.

Select and delete the test rectangle, test diamond, and connector you created in the previous step. Your canvas should now show only the empty labeled lanes.

Now place the following shapes in the Reporter lane, working left to right:

Drag a circle (oval) into the Reporter lane at the far left. Double-click it and label it Abuse/neglect report initiated. This is your process start point.
Drag a rectangle to the right of the start circle. Label it Provide incident details to hotline.
Draw a connector (arrow) from the start circle to this rectangle.

Next, place these activities in the Intake Specialist lane. These represent the tasks triggered by the reporter's call:

Add a rectangle labeled Interview reporter: gather incident, victim, perpetrator, and family details.
Add a rectangle labeled Screen for domestic violence using DV screening questions.
Add a rectangle labeled Clarify vague or incomplete statements.
Add a rectangle labeled Document allegation narrative in case management system.

Connect all shapes in sequence with arrows (left to right).
Draw a cross-lane connector from the Reporter's Provide incident details to hotline rectangle down into the Intake Specialist's Interview reporter rectangle. This arrow crossing lanes represents a handoff between actors.

You should now see 6 shapes across two lanes with arrows connecting them in a clear left-to-right flow. The connector crossing from Reporter down into Intake Specialist is your first cross-lane handoff.

Why cross-lane connectors matter

Every time an arrow crosses from one swimlane into another, it signals a handoff. Handoffs are where delays, miscommunication, and dropped information happen. Marking them explicitly makes your diagram useful for identifying process friction, not just documenting sequence.

Shapes jumping to the wrong lane?

If a shape snaps into a different lane when you drop it, drag it back into the correct lane. Make sure the shape's center is fully inside the target lane's boundaries before releasing.

If connectors are not attaching to shapes, hover over the edge of the shape until you see a green circle (connection point) appear, then drag from that point to the target shape.

Map the screening decision and history check (middle of diagram)

The Intake Specialist now needs to check whether this report has any prior history and whether the allegations actually meet the legal definition of abuse or neglect. This is where your first decision diamond comes in.

Continue adding shapes in the Intake Specialist lane, to the right of your last rectangle:

Add a rectangle labeled Search case management system for prior history on all participants.
Add a diamond shape labeled Do allegations meet statutory definition of child abuse/neglect?. This is your first decision point.

The decision diamond creates two paths. Map both outcomes:

From the diamond's No path: add a rectangle labeled Screen out report, then an end circle. This path terminates the process.
From the diamond's Yes path: add a rectangle labeled Determine response priority timeframe (Immediate/24hr/72hr).
After the priority rectangle, add another rectangle labeled Determine assessment track (Investigative or Family).

Label each connector out of the diamond with Yes or No by double-clicking the connector line and typing the label.

Now add shapes in the Case Management System lane to show the system's role:

Add a rectangle labeled Return prior case history and open intakes for participants.
Add a parallelogram (data shape) labeled Intake report created with allegation narrative. To find the parallelogram, search parallelogram in the shape search panel on the left.
Draw a connector from the Case Management System's Return prior case history rectangle back up to the Intake Specialist's Search case management system rectangle. This is your second cross-lane handoff, showing data flowing back.
Draw a connector from the Intake Specialist's Document allegation narrative rectangle down to the parallelogram in the Case Management System lane. This is your third cross-lane handoff.

Check your canvas. You should now see shapes in three lanes (Reporter, Intake Specialist, Case Management System) with arrows crossing between lanes in at least three places.

What does the parallelogram mean?

In process mapping notation, a parallelogram represents data input or output. It signals that information is being created, stored, or retrieved from a system. Using it here distinguishes the intake report (a data artifact) from the activities that produce it.

Can't find the parallelogram shape?

Type parallelogram in the search bar at the top of the shapes panel on the left side. If search returns nothing, click More Shapes at the bottom of the panel and enable the General library.

Help me find the parallelogram shape in draw.io.

Map the supervisory review and assignment (right side of diagram)

After the Intake Specialist finishes their work, the report goes to a Supervisor for quality review. This is where your second decision diamond and potential correction loop live.

Place these shapes in the Supervisor lane:

Add a rectangle labeled Review intake report for accuracy and completeness.
Add a diamond labeled Approve routing to local office?. This is your second decision point.

Map both outcomes from this decision:

From the diamond's No path: add a rectangle labeled Return to Intake Specialist with corrections. Draw a connector from this rectangle back into the Intake Specialist lane (pointing at the Document allegation narrative activity). This creates a correction loop.
From the diamond's Yes path: add a rectangle labeled Forward to local county office for final review.
After the forward rectangle, add an end circle labeled Case assigned for assessment.

Draw a cross-lane connector from the Intake Specialist's Determine assessment track rectangle up into the Supervisor's Review intake report rectangle. This shows the completed report moving to supervisory review.
Label the connectors out of the second diamond with Yes and No.

Why does the correction loop matter?

The correction loop (Supervisor sends back to Intake Specialist) is one of the most common sources of delay in real intake processes. Every round trip adds time before a child receives services. When you annotate pain points in the next step, this loop will be important context.

Save your diagram by pressing Cmd+S (macOS) or Ctrl+S (Windows).

Before you check your work, take a moment to think. How many shapes do you expect to count across all four lanes? How many connectors cross between lanes?

Count the shapes on your canvas. Your completed As-Is diagram should have at least 10 shapes across four lanes, with a clear start (report initiated) and end (case assigned or screened out). You should see two decision diamonds and at least three connectors that cross between lanes.

What should I see?

Your diagram tells a story from left to right: a report comes in through the Reporter lane, gets processed through a series of Intake Specialist activities, checks against the Case Management System, passes through two decision gates, and ends either screened out or assigned for assessment through the Supervisor.

If your shapes are scattered or hard to follow, drag them to align left-to-right within each lane. The flow direction should be obvious to anyone reading the diagram.

Missing shapes or connectors?

If your count is below 10 shapes, check that you added both the Screen Out end circle (on the No path) and the Case Management System lane shapes. These are easy to miss.

If you have fewer than three cross-lane connectors, check that you drew: (1) Reporter to Intake Specialist, (2) Intake Specialist to Case Management System (the narrative), (3) Case Management System back to Intake Specialist (the history return), and (4) Intake Specialist to Supervisor.

Help me check my As-Is diagram is complete.

You now have a complete As-Is process map that documents exactly how a child welfare intake works today. Next up, you will annotate the pain points on this diagram to build the case for AI intervention.

Annotate Pain Points on the As-Is Diagram

Your As-Is diagram captures the full intake workflow from report initiation through case assignment. That's solid documentation.

Documentation alone doesn't justify change or investment. In this step, you'll mark specific activities where workers face delays, inconsistency, and cognitive overload. These pain point annotations transform your process map into a diagnostic tool that makes the business case for AI intervention.

In this step, get ready to:

Add colored callout shapes as pain point markers on your diagram.
Annotate four specific activities where workers experience friction, inconsistency, or cognitive overload.
Add a legend and export your annotated diagram as a PNG for your portfolio.

Add pain point callout shapes

Pain points need to stand out visually from your process shapes. You'll create a distinctly colored callout shape and connect it to the activity it describes with a dashed line.

Switch back to your CW-Intake-AsIs diagram in draw.io.
Drag a rectangle from the shape library onto a blank area near the activity labeled "Document allegation narrative" in the Intake Specialist lane.
Select the rectangle you just placed.
In the Style tab of the format panel (right side), click the color button next to Fill to open the color palette.
Choose a light red or orange color (e.g. enter f8cecc in the hex field without the leading #), then click Apply.
Add a border by confirming the Line checkbox is checked. Set the line color to a darker red or orange (e.g. b85450).

You should see a small red/orange rectangle on your canvas. This is your pain point callout shape.

Now draw a connector from your callout shape to the "Document allegation narrative" activity by hovering over the callout until you see a direction arrow, then dragging to the target activity.
Select the connector you just drew.
In the Style tab, change Pattern from solid to dashed.
Set the source and target arrow heads to None so the connector is a plain dashed line.

You now have a dashed line connecting your red/orange callout to the activity it annotates. This is the visual pattern you'll repeat for each pain point.

Speed up with Copy Style

After styling your first callout, select it and press Option+C (macOS) or Alt+C (Windows) to copy its style. Then select each new callout rectangle and press Option+V (macOS) or Alt+V (Windows) to paste the style instantly.

Annotate these four pain points

Each pain point targets a specific activity where workers lose time, make inconsistent decisions, or face cognitive overload. You'll place a callout near each activity and type the annotation text directly into the shape.

Double-click your first callout shape (the one connected to "Document allegation narrative") to enter edit mode.
Type this annotation text: "PP-001: Narratives vary by worker. No structured format. Critical details often missing. Later readers must re-interpret."

You should see the text appear inside your red/orange callout. Resize the shape if the text overflows.

Create a second callout rectangle near the "Search case management system for prior history" activity in the Intake Specialist lane.
Apply the same red/orange fill style (use Paste Style or set it manually).
Connect it with a dashed line to the activity.
Double-click the callout and type: "PP-002: Manual search across multiple name spellings, aliases, and addresses. Takes 8-12 minutes per participant. Workers may miss linked cases."

Two pain points annotated. Two more to go.

Create a third callout rectangle near the decision diamond "Do allegations meet statutory definition of child abuse/neglect?"
Apply the red/orange fill style and connect with a dashed line.
Double-click and type: "PP-003: Screening criteria applied from memory. Inconsistent decisions across workers. No decision-support tool. High cognitive load under time pressure."

Create a fourth callout rectangle near the activity "Determine response priority timeframe" in the Intake Specialist lane.
Apply the red/orange fill style and connect with a dashed line.
Double-click and type: "PP-004: Priority assignment relies on worker judgment without structured risk scoring. Similar cases get different response times."

Your diagram now has four colored callout annotations, each connected by a dashed line to the specific activity it describes.

Why these four pain points?

Each one targets a place where AI can realistically help. PP-001 is a structuring problem (AI can extract structured data from free text). PP-002 is a search problem (AI can match across aliases and spelling variants). PP-003 and PP-004 are consistency problems (AI can apply criteria systematically as decision-support).

Notice that none of them suggest AI should make the final decision. They all point to support tasks, not safety-critical determinations. This distinction becomes central when you design the To-Be workflow.

Add a legend and save

A legend tells anyone reading your diagram what the colored callouts mean. It ensures your artifact is self-explanatory without verbal explanation.

Find a blank area below or beside your swimlanes.
Drag a rectangle onto the canvas and type "LEGEND" inside it.
Next to the LEGEND label, place a small rectangle with the same red/orange fill you used for your callouts.
Type "Pain Point / AI Opportunity" inside this colored rectangle.

Your legend should now show a small sample callout labeled "Pain Point / AI Opportunity" next to the word LEGEND.

Save your diagram by pressing Cmd+S (macOS) or Ctrl+S (Windows).
Export a PNG copy by clicking File, then Export as, then PNG.
Save the PNG file to your device for your portfolio.

Before you review your final result, take a moment to predict: how many distinct pain point callouts should you count on your canvas?

You should count four red/orange callout shapes (PP-001 through PP-004), each connected by a dashed line to a specific activity, plus a legend at the bottom or side. Your diagram now tells a story, not just documents a process.

Callout shapes not standing out?

If your callouts blend in with the process shapes, check that the Fill color is set to a red or orange. White callouts disappear against white activity rectangles.

If the dashed lines appear solid, select the connector and confirm Pattern is set to dashed in the Style tab. The default pattern is a solid line.

Why aren't my pain point callouts visible on the diagram?

Your As-Is diagram is now a diagnostic artifact. It shows not just what happens, but where the process breaks down. Next up, you'll design the To-Be workflow that uses AI to fix each of these pain points while keeping humans in the loop for safety-critical decisions.

Design the To-Be AI-Enabled Intake Workflow

Your annotated As-Is diagram shows exactly where the intake process breaks down. The next question is: how do you fix it?

In child welfare, the fix comes with a critical constraint. Final safety decisions must remain with human workers. AI can support those decisions, but positioning it as the decision-maker fails regulatory and ethical review. This step teaches you to design AI as decision-support in a high-stakes government workflow.

In this step, get ready to:

Create a new To-Be diagram page with a fifth swimlane for AI Services.
Insert AI-powered activities that directly address each pain point from your As-Is analysis.
Add human-in-the-loop verification points ensuring AI never makes final safety decisions.

Create a new diagram page for the To-Be

Your As-Is diagram is a finished artifact you do not want to modify. The To-Be design lives on a separate page in the same file so stakeholders can flip between the two for comparison.

In draw.io, click the + tab at the bottom of the canvas to add a new page.
Double-click the new page tab label and rename it to To-Be AI-Enabled.
In the template selector, choose the Swimlane Diagram template from the Basic collection (the same template you used in Step 1).

You now have a fresh swimlane canvas on a new page. Next, set up the five lanes your To-Be process needs.

Delete any default shapes inside the lanes (select them and press Delete).
Double-click each lane header to rename. Set the lanes in this order from top to bottom: Reporter, Intake Specialist, AI Services, Supervisor, Case Management System.
If you need to add extra lanes, select the outer pool border, then in the Arrange tab of the format panel click Insert Row After.

Why a fifth lane for AI Services?

Giving AI its own swimlane makes it visually clear where automation sits versus where humans act. Stakeholders can instantly see that every AI output feeds into a human lane before any decision is final.

Copy the basic flow structure from your As-Is page into this new page: a start circle in the Reporter lane, an end circle in the Supervisor lane, and the general left-to-right sequence of activities in the Intake Specialist lane.
Leave space between activities in the Intake Specialist lane. You will insert AI connections into those gaps in the next substep.

Your canvas should now show five clearly labeled swimlanes with a basic skeleton of the intake flow copied in.

Only seeing the original As-Is page?

Look at the bottom of the draw.io canvas for page tabs. If you only see one tab, you may have renamed the existing page instead of adding a new one. Click the + icon to add a fresh page.

Help me add a new page in draw.io

Insert AI-powered activities addressing each pain point

Each pain point you annotated on your As-Is diagram maps to a specific AI capability. You will now add four AI activities in the AI Services lane, each one directly solving a documented problem.

Work through these one pain point at a time, placing shapes in the AI Services lane and drawing connectors to show how data flows.

For Pain Point 1 (inconsistent narratives): drag a rounded rectangle into the AI Services lane and label it: Auto-structure narrative: extract victim, perpetrator, incident type, location, and timeline into structured fields from free-text dictation.
Draw a connector from the Intake Specialist's Record reporter information activity into this AI shape.
Draw a second connector from this AI shape back out to the Case Management System lane, showing structured output flowing into the system.

You should now see your first AI activity connected between the Intake Specialist and the Case Management System. Three more to go.

For Pain Point 2 (slow manual history search): add a rounded rectangle in the AI Services lane and label it: Automated prior-case matching: search all participant names, aliases, addresses, and linked households; surface matching cases with confidence scores.
Connect: when the Intake Specialist enters participant details, a connector goes into this AI shape.
Draw the output connector from this AI shape back to the Intake Specialist lane, representing matched cases returned for human review. This replaces the manual search from the As-Is process.

For Pain Point 3 (inconsistent screening): add a rounded rectangle in the AI Services lane and label it: Screening decision support: apply statutory criteria checklist to structured intake data; flag met/unmet criteria; generate recommendation with reasoning.
Connect the structured intake data flowing into this AI shape.
Draw the output connector to the Intake Specialist lane where the worker reviews the recommendation and makes the final screening decision.
Add a text annotation near this connector: Human makes final screening decision. AI provides recommendation only..

For Pain Point 4 (inconsistent priority assignment): add a rounded rectangle in the AI Services lane and label it: Risk-scored priority recommendation: calculate response priority based on structured risk factors; present recommended timeframe with rationale.
Connect this AI shape so it receives data from the structured intake.
Draw the output connector to the Intake Specialist lane where the worker confirms or overrides the priority.

What pattern do all four AI activities share?

Every AI activity follows the same shape: data flows in from a human action, AI processes it, and the output flows back to a human for review. AI never connects directly to a final decision diamond or an end circle. This is the human-in-the-loop pattern. It ensures accountability stays with credentialed workers.

Your AI Services lane should now contain four rounded rectangles with connectors flowing in and out. Each one maps to a pain point from your As-Is analysis.

Connectors not attaching to shapes?

Hover over the source shape until you see blue direction arrows or small blue connection points appear. Drag from one of those points to the target shape. If you drag from empty canvas space, the connector will float unattached.

Help me fix floating connectors in draw.io

Add human-in-the-loop verification points and save

Your AI activities are placed and connected. Now you need to make the human-in-the-loop constraint visually obvious to any stakeholder who reviews this diagram.

Check every connector leaving an AI Services shape. Confirm that it lands in the Intake Specialist or Supervisor lane on a human review activity (never directly on a decision diamond or end circle).
If any AI output connects directly to a decision or endpoint, insert a human review activity between them. Label it something like Review AI recommendation or Confirm or override priority.

Next, add visual markers that make the AI-assisted steps instantly recognizable.

For each AI activity in the AI Services lane, add a small callout shape with a green fill color.
Label each green callout: AI-Assisted.
Position each green callout beside its corresponding AI activity shape.

Your four AI activities should now each have a green callout beside them. The final piece is a design principle notation that frames the entire diagram.

Add a text box or rectangle at the top of your diagram, above the swimlanes.
Type this text into the box: Design Principle: AI provides decision-support only. All safety-critical determinations remain with credentialed human workers per 45 CFR 1355.52 and agency policy..

Why reference 45 CFR 1355.52?

This is the federal regulation governing CCWIS systems. It requires efficient and effective administration of child welfare programs. Citing it signals to stakeholders that your design respects the regulatory framework.

The human-in-the-loop constraint is a best-practice design principle grounded in both this regulation and agency policy.

Save your diagram by pressing Cmd+S (macOS) or Ctrl+S (Windows).
Export a PNG copy of this To-Be page: click File, then Export as, then PNG. Save it to your device as your portfolio artifact.

Before you check your work, predict: how many AI-powered activities do you expect to count in your AI Services lane?

Count the rounded rectangles in your AI Services lane. You should see four activities, one addressing each pain point. Verify that every one of them has a connector leading to a human review activity in the Intake Specialist or Supervisor lane. None should connect directly to a final screening decision or end circle.

An AI activity connects directly to a decision diamond?

Insert a human review activity between the AI output and the decision. Select the connector, delete it, place a new rounded rectangle in the Intake Specialist lane labeled with the review action, then reconnect: AI shape to human review, human review to decision diamond.

Help me insert a human review step between my AI activity and a decision diamond

Your To-Be diagram is complete. You now have two diagram pages telling a before-and-after story: here is how intake works today, and here is how AI improves it while keeping humans in control. Next up, you will translate these AI touchpoints into buildable requirements that a development team can act on.

Write User Stories with Acceptance Criteria

Your To-Be diagram now shows where AI fits into the intake process and who interacts with it. But a diagram alone does not tell a development team what to build.

User stories translate your process design into buildable requirements. Each story captures one feature from a specific user's perspective, and its acceptance criteria define exactly how QA and UAT will verify the feature works.

In this step, get ready to:

Set up a structured user story document with project metadata.
Write four user stories in As a / I want / So that format with Given/When/Then acceptance criteria.
Assign MoSCoW priorities to each story based on business impact.

Set up your user story document

Before writing stories, you need a structured document that a development team can reference during sprint planning. A clear header section identifies the project, module, and version at a glance.

Open Google Docs in your browser (or your preferred word processor).
Create a new blank document.
Title the document CCWIS AI-Enabled Intake - User Stories.

Add the following header section at the top of your document:

Project: Greenfield County CCWIS Modernization
Module: Intake Screening
BA Author: [[YOUR_NAME="your name"]]
Date: [[TODAYS_DATE="today's date"]]
Version: 1.0

Why include a document header?

Professional BA artifacts always identify the project, module scope, author, and version. This lets anyone picking up the document months later understand its context immediately.

Below the header, add a section heading: User Stories.
Under that heading, create a structure for each story with these fields: Story ID, Title, User Story, Acceptance Criteria, Priority (MoSCoW), Story Points (leave blank).

Your document now has the metadata header and a template structure ready for stories.

Write user stories for each AI touchpoint

Each user story maps directly to one of the four AI activities in your To-Be diagram. The standard format is: "As a [role], I want [feature], so that [benefit]." The acceptance criteria use Given/When/Then format to make each requirement testable.

Write the first story in your document by copying the following content. This addresses Pain Point 1 (inconsistent narrative documentation):

Story ID: US-INT-001
Title: Automated Narrative Structuring
Priority:

User Story: As an Intake Specialist, I want the system to automatically extract structured data fields (victim name, perpetrator name, incident type, location, date/time, and relationship) from my free-text dictation of the reporter's statement, so that I can ensure consistent documentation without manually reformatting every narrative.

Acceptance Criteria:
AC-1: Given a free-text allegation narrative of at least 50 words, when the narrative is submitted, then the system extracts and displays victim name, perpetrator name, incident type, location, and date/time in separate structured fields within 10 seconds.
AC-2: Given extracted fields contain low-confidence values (below 80% confidence), when displayed to the Intake Specialist, then low-confidence fields are highlighted in yellow for manual review.
AC-3: Given the Intake Specialist edits any extracted field, when saved, then the corrected value overwrites the AI-extracted value and is logged as a manual correction.

What makes these acceptance criteria testable?

Each criterion follows Given/When/Then format. "Given" sets the precondition, "When" names the trigger action, and "Then" states the observable outcome. A QA tester can execute each one as a test case without guessing what "working correctly" means.

Notice the specific thresholds: 50 words minimum, 80% confidence, 10-second response time. Vague criteria like "the system should be fast" are untestable.

Write the second story below the first. This addresses Pain Point 2 (slow manual history search):

Story ID: US-INT-002
Title: Automated Prior-Case Matching
Priority:

User Story: As an Intake Specialist, I want the system to automatically search for prior cases matching any participant's name, alias, date of birth, or address and present results ranked by confidence score, so that I can identify relevant case history in under 30 seconds instead of 8-12 minutes of manual searching.

Acceptance Criteria:
AC-1: Given participant details are entered (minimum: first name, last name), when the intake is initiated, then the system returns matching prior cases within 30 seconds, ranked by confidence score.
AC-2: Given a match is found with greater than 90% confidence, when displayed, then the matched case summary (case ID, status, allegation type, date) is shown with a "High Confidence" badge.
AC-3: Given no matches are found, when the search completes, then the system displays "No prior history found" and logs the search as complete.
AC-4: Given the Intake Specialist dismisses a suggested match, when dismissed, then the dismissal is logged with reason for audit trail.

Why does this story have four acceptance criteria?

US-INT-002 covers multiple outcome paths: matches found, high-confidence matches, no matches found, and dismissed matches. Each outcome path needs its own testable criterion so QA can verify every branch.

Review your document. You should see two complete stories, each with a clear user story statement and numbered acceptance criteria.

Write the third story below the second. This addresses Pain Point 3 (inconsistent screening decisions):

Story ID: US-INT-003
Title: Screening Decision Support
Priority:

User Story: As an Intake Specialist, I want the system to evaluate the structured intake data against the statutory definition of child abuse/neglect and present a recommendation (screen in / screen out) with the specific criteria met or unmet, so that I can make more consistent screening decisions under time pressure.

Acceptance Criteria:
AC-1: Given a completed structured intake with all required fields populated, when the Intake Specialist requests screening analysis, then the system displays each statutory criterion as met/unmet with a citation to the relevant statute within 15 seconds.
AC-2: Given the system recommends "Screen In," when the Intake Specialist selects "Screen Out" (override), then the system requires a written justification and flags the override for supervisory review.
AC-3: Given the system recommendation, when displayed, then the display includes "AI Recommendation - Final decision rests with credentialed worker" disclaimer.

Why does AC-2 require justification for overrides?

In child welfare, screening decisions are safety-critical. When a worker overrides the AI recommendation, logging the justification creates an audit trail. This protects both the child (supervisors can review override patterns) and the worker (their reasoning is documented).

AC-3 reinforces the human-in-the-loop principle from your To-Be diagram. The disclaimer ensures no one mistakes AI output for a final decision.

Write the fourth and final story below the third. This addresses Pain Point 4 (inconsistent priority assignment):

Story ID: US-INT-004
Title: Risk-Scored Priority Recommendation
Priority:

User Story: As an Intake Specialist, I want the system to calculate a recommended response priority (Immediate/24hr/72hr) based on structured risk factors from the intake, so that similar cases receive consistent response timeframes regardless of which worker handles the intake.

Acceptance Criteria:
AC-1: Given a screened-in intake with structured risk factors, when priority calculation is requested, then the system displays a recommended priority level with the top 3 contributing risk factors.
AC-2: Given the Intake Specialist overrides the recommended priority, when a different priority is selected, then the override is logged with justification and flagged for supervisor review.
AC-3: Given the priority recommendation, when displayed, then the rationale includes specific data points from the intake that drove the recommendation.

What pattern do you notice across all four stories?

Every story follows the same human-in-the-loop pattern: AI produces a recommendation, the human reviews it, and any override is logged. This consistency reflects the design principle from your To-Be diagram that AI provides decision-support only.

Scroll through your document to confirm you have all four stories (US-INT-001 through US-INT-004) with their acceptance criteria written out.

Missing a story or criteria?

Check that each story has a unique ID (US-INT-001, US-INT-002, US-INT-003, US-INT-004). Verify each story has the "As a... I want... So that..." structure with no parts missing.

Confirm US-INT-001 has 3 acceptance criteria, US-INT-002 has 4, US-INT-003 has 3, and US-INT-004 has 3. Each criterion must have all three parts: Given, When, Then.

Help me check my acceptance criteria

Assign MoSCoW priorities

MoSCoW prioritization categorizes requirements as Must Have, Should Have, Could Have, or Won't Have. This helps the development team decide what to build first during sprint planning.

Go back to each story in your document and fill in the empty Priority: field with the appropriate MoSCoW value.
For US-INT-001 (Narrative Structuring), assign: Should Have. This improves consistency but workers can function without it.
For US-INT-002 (Prior-Case Matching), assign: Must Have. Missing prior history is a direct child safety risk.
For US-INT-003 (Screening Decision Support), assign: Must Have. This is core to CCWIS data quality requirements.
For US-INT-004 (Priority Recommendation), assign: Should Have. Workers already have response priority tools, though they are applied inconsistently.

Why are only two stories marked Must Have?

Must Have means the system cannot launch without this feature. US-INT-002 and US-INT-003 directly address child safety risks (missed history, inconsistent screening). If the system launched without them, the core problem remains unsolved.

US-INT-001 and US-INT-004 improve efficiency and consistency but are not safety-critical blockers. Marking everything as Must Have defeats the purpose of prioritization.

Save your document.

Before you review your finished document, think about what it should contain. How many stories do you expect? How many total acceptance criteria across all four?

Scroll through your completed document. You should have 4 user stories in standard format, 13 total acceptance criteria (3 + 4 + 3 + 3), and a MoSCoW priority assigned to every story. Two stories are marked Must Have and two are marked Should Have.

✔️ I see all four stories with priorities assigned

Your user story document is complete. You now have a buildable specification that a development team can use to plan sprints and a QA team can use to write test cases.

ⓧ I'm missing stories or criteria

Compare your document against this checklist:

US-INT-001: Automated Narrative Structuring, Should Have, 3 acceptance criteria.
US-INT-002: Automated Prior-Case Matching, Must Have, 4 acceptance criteria.
US-INT-003: Screening Decision Support, Must Have, 3 acceptance criteria.
US-INT-004: Risk-Scored Priority Recommendation, Should Have, 3 acceptance criteria.

Scroll up to the relevant story section in this step and copy any missing content into your document.

Your requirements document is ready. Next up, you will build a requirements traceability matrix that connects every pain point to a user story and a testable scenario.

Secret mission

Build a Requirements Traceability Matrix

You have pain points, a To-Be design, and user stories. But can you prove every gap has a solution, every solution has a testable requirement, and nothing slipped through the cracks? Build the artifact that closes the loop between analysis and delivery.

Clean Up Your Resources

Decide whether to keep your resources running, pause them to come back later, or delete them entirely. This project runs entirely locally and uses free browser-based tools, so there are no ongoing costs.

Resources you used:

draw.io diagram file (CW-Intake-AsIs) with As-Is and To-Be pages, saved locally on your device.
Exported PNG files (CW-Intake-AsIs.png and CW-Intake-ToBe.png), saved locally on your device.
Google Doc containing your user stories (CCWIS AI-Enabled Intake - User Stories).
Requirements Traceability Matrix (in Google Sheets or within your user story document).

Keep everything running

No action needed. Choose this if you plan to continue refining your artifacts, adding more detail to your diagrams, or using these as portfolio pieces.

Your draw.io file stays on your device wherever you saved it. You can reopen it anytime at https://app.diagrams.net by selecting Open Existing Diagram.
Your PNG exports remain in the folder where you saved them.
Your Google Doc and Google Sheets files remain in your Google Drive.

Pause - I'll come back to this later

Since all files are saved locally or in Google Drive, there is nothing to pause. Your work persists without any running processes or active services.

Bookmark https://app.diagrams.net so you can return to edit your diagrams later.
Consider moving your PNG exports and draw.io file into a dedicated portfolio folder so you can find them easily.

Delete - I don't want to use this again

Remove all project files from your device and Google Drive.

Delete local files. On macOS, open Finder, locate your CW-Intake-AsIs draw.io file and both PNG exports, then move them to the Trash. On Windows, open File Explorer, locate the same files, then press Delete.
Delete your Google Doc. Open Google Drive, right-click the CCWIS AI-Enabled Intake - User Stories document, and select Move to trash.
Delete your RTM. If you created a separate Google Sheet for the Requirements Traceability Matrix, right-click it in Google Drive and select Move to trash. If you added the RTM as a section within your user story document, it was already removed in the previous step.
Empty your trash. In Google Drive, click Trash in the left sidebar, then click Empty trash to permanently delete the files.

Nice Work!

You just produced a complete business analysis artifact set for a CCWIS modernization project. From a blank canvas to a portfolio-ready deliverable suite, you worked through the full BA lifecycle: documenting the current state, diagnosing problems, designing an AI-enabled solution, and specifying testable requirements.

You've learned how to:

Map a real-world process as a swimlane diagram with multiple actors, decision points, and cross-lane handoffs, then annotate pain points that transform documentation into diagnostic insight.
Design an AI-enabled To-Be workflow that positions AI as decision-support rather than decision-maker, respecting the human-in-the-loop requirements of child welfare practice and federal CCWIS regulations.
Write user stories with testable acceptance criteria in Given/When/Then format, with MoSCoW prioritization, ready for a development team to pick up.
Secret Mission: Build a requirements traceability matrix linking pain points to user stories to UAT scenarios, proving end-to-end coverage with zero orphaned entries.

Ready to quiz yourself?