7.2 Intelligent Document Automation (IDA), the Document Workspace & the IDA vs. IDA/IFR Decision

7.2 Intelligent Document Automation (IDA), the Document Workspace & the IDA vs. IDA/IFR Decision

Exam Focus: The official AP-222 outline phrases this objective with unusual precision: "Given requirements to save data from physical documents, assess the need for document processing capabilities and whether to use IDA, IDA/IFR." That wording is the whole question. Salesforce ships two separate things — Intelligent Document Automation (IDA), which receives and organizes documents, and Intelligent Form Reader (IFR), which reads them — and a large share of the exam's document questions are really asking you to notice that a scenario needs only the first one. Recommending the paid OCR add-on for a requirement that is actually about routing and tracking is the classic over-engineering trap.

---

Two Layers, Two Different Problems

Government agencies drown in documents. But "drowning in documents" splits into two genuinely different complaints, and they have different solutions:

Intelligent Document Automation (IDA) is the document handling layer. It gives agencies a single Intelligent Document Workspace where every inbound file lands, can be previewed, rotated, zoomed, split, attached to the right constituent record, and tracked to completion against a checklist.

Intelligent Form Reader (IFR) — the engine Public Sector Solutions surfaces under the name Intelligent Document Reader — is the reading layer. It calls Amazon Textract to pull key-value pairs, identity-document fields, and tables off the page and map them into Salesforce fields. Salesforce's own product messaging is explicit that IDA "also includes the Intelligent Form Reader powered by Amazon Textract" as an added capability, and IFR carries its own add-on license and its own AWS prerequisite chain (covered in detail in section 7.3).

---

Inside the Intelligent Document Workspace

The workspace is enabled from Setup (search for Intelligent Document Workspace Settings and turn on the toggle) and is built from four objects that also appear as tabs:

1. Received Document

Every inbound file creates a Received Document record. This is the unit of work in the workspace.

• Multi-channel arrival: portal uploads from an Experience Cloud site, an agency fax partner, inbound email, hospital or partner-agency transfers, and manual bulk upload. Admins can upload several files at once.
• In-workspace manipulation: documents can be rotated, zoomed, and — critically for government intake — split. A constituent who scans a 40-page benefits packet as one PDF produces one Received Document; the intake officer selects page ranges (either by clicking pages or typing page numbers into the custom page selector) and splits it into separate documents, so pages 1–4 become the application form, pages 5–12 become the proof of income, and pages 13–40 become the lease agreement.
• Attachment to business records: once split, each document is attached to the object it belongs to — IndividualApplication, BusinessLicenseApplication, Case, Contact, Account, FundingAward, and so on.

2. Document Checklist / Document Checklist Item

The Document Checklist is the agency's list of what a constituent still owes. It shows which required documents have been received and which are outstanding, and it preserves traceability back to the original uploaded file after a split. This is the object that drives the "your application is incomplete" notice.

3. Document Scan Result

When a document is processed, IDA writes a Document Scan Result (surfaced on the Received Document page layout through the OCR Scan Result related list). This is the record that holds extracted content and the evidence of what the system read — the audit artifact a records officer will ask for.

4. Action Plan Template

This is the automation hook, and it is the most commonly missed part of IDA on the exam. An Action Plan Template attached to the document workspace can automatically:

• attach a scanned document to the correct Document Checklist Item;
• assign it to a queue for a specific review team; or
• trigger a flow — for example, advancing the application status, notifying the assigned caseworker, or starting a fee calculation.

Because Action Plan Templates already do queue assignment and flow invocation declaratively, a scenario that says "route incoming documents to the licensing review queue and notify the reviewer" needs no Apex and no OCR — it needs an Action Plan Template.

---

The Document Lifecycle End to End

1. Receipt — a file arrives through any channel and becomes a Received Document.
2. Classification — the intake officer (or automation) identifies the document type.
3. Splitting — multi-document scans are separated into logically distinct documents.
4. Attachment — each document is linked to the correct application, case, or constituent record and to its Document Checklist Item.
5. Processing — the document is scanned; a Document Scan Result is written. If IFR is licensed and a template exists, this is where OCR extraction populates mapped fields.
6. Post-processing review — staff verify the captured data against what was entered, correct discrepancies, and mark the checklist item complete.
7. Automation — the Action Plan Template fires the queue assignment or flow that moves the case forward.

---

The Decision Framework the Exam Actually Tests

Work the scenario against this table. The deciding question is never "do they have a lot of paper?" — it is "does a field value need to leave the page and land in a record?"

Three traps in the wording

1. "Digitize our paper files" does not automatically mean OCR. An agency that wants paper replaced by a tracked digital workspace is describing IDA. Read on for a sentence about field values.
2. Cost and procurement are legitimate architecture inputs. IFR is a paid add-on that also requires the agency's own AWS account. On a scenario where the requirement is satisfied by IDA alone, recommending IFR is wrong on both design and stewardship grounds — and public money makes that judgment sharper, not softer.
3. Business Rules Engine is not a document tool. A distractor that offers BRE for a document-processing requirement is testing whether you know BRE decides and calculates; it does not read pages.

---

💡 Real-World AP-222 Exam Scenarios & Case Analysis

Scenario 1: County Building Department Intake Backlog

A county building department receives permit supporting documents by fax from contractors, by portal upload from homeowners, and by email from architects. Plan sets frequently arrive as one enormous PDF containing the site plan, the structural calculations, and the contractor's insurance certificate. Reviewers cannot tell which permits are complete, and 30% of applications stall because a single missing document is discovered weeks later.

What should the architect recommend?

• Intelligent Document Automation, without Intelligent Form Reader. Every requirement in the scenario is a handling requirement.
• Configure the Intelligent Document Workspace so faxed, emailed, and portal-uploaded files all land as Received Documents in one queue.
• Use page splitting to break the combined plan set into three documents, each attached to its own Document Checklist Item on the BusinessLicenseApplication.
• Configure a Document Checklist per permit record type so "complete versus incomplete" is visible on day one instead of week three.
• Use an Action Plan Template to assign structural calculations to the engineering review queue and insurance certificates to the risk queue automatically.
• No field values need to be lifted off the page, so the IFR add-on is not justified.

Scenario 2: State Human Services Income Verification

A state human services agency already runs IDA. Its problem is different: caseworkers open each uploaded pay stub and W-2 and manually retype the employer name, employer identification number, and gross wages into the benefits application. Average handling time is 18 minutes per document, and transcription errors trigger benefit overpayment audits.

What changes?

• Add Intelligent Form Reader on top of the existing IDA foundation — this is the "IDA/IFR" answer.
• Scope the prerequisites honestly into the project plan: the IFR add-on license, the agency's own AWS account, an IAM user with Amazon Textract permissions, and an external plus named credential.
• Build a document type and Form Template for the W-2, map Box 1 and Box 3 to the income fields, and set a straight-through processing confidence threshold.
• Keep every IDA capability already in place: the workspace, checklist, splitting, and Action Plan Templates continue to do the handling; IFR only adds the reading.

Scenario 3: Records Officer Audit Request

Six months after go-live, a records officer must produce, for a contested license denial, the original scanned document, what the system read from it, and who touched it.

Where does that evidence live?

• The original file is the Received Document, preserved even after splitting, with traceability from each split document back to the source.
• What the system read is the Document Scan Result (exposed via the OCR Scan Result related list on the Received Document page layout).
• What was owed and when it was satisfied is the Document Checklist Item history.
• This is why the exam favours the standard document objects over a custom "Uploaded_File__c" object: the audit trail is already modelled.