Using AI-Assisted Form Fill in the Designer

AI-assisted form fill works best when the form gives Laserfiche AI clear signals about what each field means and where the source document is. This topic walks through how to build such a form step by step, then describes best practices that improve both usability for the person filling out the form and the accuracy of extracted values.

Before you begin: gather three or four representative documents of the type your users will upload, such as real resumes or invoices with identifying information removed. You will use them to test the form in the final step.

Note: AI-assisted form fill is not available in the classic form designer.

Step-by-Step: Building the Form

Step 1: Plan the page layout

Extraction is limited to a single form page. Decide which values you want AI to fill, and plan to place all of those fields on the same page as the file upload field.

1. List the values that appear in the uploaded document, such as name, email address, and job title.
2. List the values that only the person filling out the form can provide, such as the position they are applying for or a consent checkbox.
3. Plan to place the document-sourced fields on the upload page. Fields the user must answer themselves can go on that page or a later page.

Step 2: Add the file upload field

1. Open or create your form in the form designer.
2. Drag a File Upload field onto the page.
3. Give it a label that names the expected document, such as Upload your resume rather than Attachment.
4. Add text to the tooltip, text above, or text below field attributes describing the accepted file types and what the document should contain.
5. Set the Max File Count to 1 to allow AI extraction to work.
6. For a better experience, place the field near the top of the page so users upload a file before they start manually filling in the form.

Step 3: Enable the field as an extraction source

1. Select the file upload field.
2. In the field settings, enable Extract and fill fields with AI.
3. Save the form.

Only file upload fields marked as an extraction source start AI extraction. A regular file upload field accepts the file but does not populate any values.

Step 4: Add and label the fields you want populated

1. Add the fields identified in Step 1 to the same page as the upload field, using the field type that matches the value:

   Suggested fields to use with a variety of values

   Value in the document	Recommended field type(s)
   Name, job title, address line	Single-line text
   Summary, description, notes	Multi-line text
   Email address	Email
   Years of experience, quantity	Number
   Invoice total, salary	Currency
   Date of birth, issue date	Date/time
   Appointment or arrival time	Time
   Yes/no attestation present in the document	Checkbox
   One value from a known set	Radio button or Drop-down

2. Write a specific label for each field. Candidate's most recent job title extracts more reliably than Title.
3. Add text to the text above or text below field attributes for any field whose expected value could be interpreted more than one way.
4. For radio button and checkbox fields, enter option values that match the wording used in the documents your users upload, and select Add 'Other' choice for an option that is not listed.

Step 5: Handle fields AI should not fill

1. Set fields to read-only when the value must come from a lookup, a formula, or a reviewer rather than the document. AI does not write to read-only fields, but will fill hidden fields if they are not marked read-only before the extraction.
2. Place fields that require a decision from the user, such as approvals or preferences, where they are visually separate from the document-sourced fields.

Step 6: Test with real documents

1. Save and publish the form to a test location.
2. Submit the form once for each sample document you gathered.
3. For each submission, note which fields were filled, which were left empty, and which received the wrong value.
4. Adjust field labels, the text in the text above or text below field attributes, field types, and option lists for the fields that performed poorly, then test again.
5. Confirm that the fields you expect users to complete manually were left untouched and any hidden fields were not filled by the AI. If hidden fields were filled, adjust rules to make them read only during this process step.

Step 7: Prepare users for the experience

1. Add a short instruction above or below the upload field explaining that uploading the document will fill in some fields automatically.
2. Tell users which fields will be AI-populated, and that they should review those values before submitting.
3. Publish the form.

Supported File Types

AI-assisted form fill supports the following file types for data extraction:

• Documents: doc, docm, docx, dot, dotx, odt, rtf, pdf, pub, wp, wpd, one, vsd
• Email: eml, msg
• Presentations: pps, ppt, pptm, pptx
• Images: pdf, png, jpg, jpeg, tif, tiff, gif, bmp, webp
• Spreadsheets: csv, xls, xlsb, xlsm, xlsx, xlt, ods
• Text / markup / misc: txt, htm, html, xml, xsl, css, js, asp, aspx, hta, ics, inf, ini, inx, sql, reg, bat, cmd, cls, stm, udf, lst, m3u, rct, asc, a, ext

Best Practices

Keep the upload and its target fields on one page

A file uploaded on one page never populates fields on another page. If you split a long form across pages, keep the document-sourced fields on the same page as the upload that supplies them. If two document types feed different sections, consider one upload field per page.

Write labels that would make sense to a stranger

The field label is the strongest signal AI has about what belongs in the field. Prefer labels that name the value and its owner, such as Applicant's mailing address or Invoice due date. Avoid single-word labels such as Name, Date, or Amount when the page contains several similar fields. Clear labels also reduce the chance that a user overwrites a correct value because they misread the field.

Avoid duplicate or near-duplicate labels on the same page

Two fields labeled Date on the same page are ambiguous to both AI and the person filling out the form. If a page genuinely needs several dates or several names, qualify each one: Date of birth, Date of hire, Date signed.

Use text above or text below field attributes as extra context

Text placed above or below a field is part of the context AI uses. A short hint such as Enter the total including tax, as shown on the invoice improves accuracy and answers the user's question at the same time.

Match option lists to document wording

Drop-down, radio button, and checkbox values are limited to the options you define. For checkbox and radio button fields, you may enable an "other" field for a value not listed. If your documents say Full-time but your drop-down offers FT, the field will likely be left empty. Review the wording used in your sample documents and align your options with it.

Set expectations about review

AI-populated values are a starting point, not a final answer. Tell users that they should check the filled values, and design the page so reviewing is easy: group related fields, keep the page short, and avoid burying populated fields inside collapsed sections.

Be deliberate with containers

Fields inside sections, collections, and tables are eligible when they are on the same page. Repeating structures such as collections and tables can be harder to match confidently, so test them specifically and keep the labels inside them just as descriptive as the labels elsewhere on the page.

Retest whenever the form or the documents change

If you rename a field, add options, rearrange the page, or start receiving a new document format, run your sample documents through the form again before relying on the results.

Notes on AI usage

AI-assisted form fill may not be available for all customers. If you would like to have AI-assisted form fill enabled, speak with your administrator or service provider for more information.

Per the Plan information page, each AI data extraction consumes 1 AI Unit. What this means is that each file uploaded to an AI extraction enabled upload field will consume 1 AI Unit from your plan. If used on a public form, then your AI Units may be consumed at a higher rate than expected and will need to be monitored closely if your business relies on the use of Laserfiche AI.

Per the limits page, extraction has its own size limitations, so a file that uploads successfully may fail extraction with the error "Your upload is too large for AI extraction."