A browser-based survey instrument for studying how managers perceive, adopt, and experience AI tools in their daily work. No build step, no framework, no package manager. Static HTML, CSS, and vanilla JavaScript, with responses piped directly into Google Sheets through a Google Apps Script deployment.
The form collects structured data across three dimensions: a five-item tech-anxiety scale, an extended AI-in-management battery, and a demographic profile. Together, these capture the attitudinal, behavioral, and contextual variables needed to analyze technology adoption patterns, managerial identity responses to AI, and AI literacy across different organizational settings.
The instrument was developed to complement qualitative research into how managers make sense of working alongside algorithmic systems, sitting at the intersection of organizational behavior, technology adoption theory, and human-computer interaction.
Tech-Anxiety Scale
Five Likert items targeting discomfort with new technology, perceived risk of AI-assisted errors, and avoidance of digital tools. Scored 1 to 5 (Strongly Disagree to Strongly Agree).
AI in Management
A composite section covering current AI tool usage across six managerial task categories, usage frequency, open-ended reflection on concrete use cases and unmet needs, attitudinal items on decision enhancement, routine task reduction, authority concerns, and required skill development, team dynamics under AI, organizational AI adoption stage and leadership support, future outlook on managerial roles, and openness to AI adoption.
Demographics
Standard organizational variables: age, gender, years in role and management, industry, organization size, education, managerial layer, team size, language proficiency, digital literacy, AI literacy, and AI exposure history.
Conditional field logic
The form uses vanilla JS event listeners to handle dynamic field visibility. When a respondent selects any frequency other than "Never" for AI tool use, a follow-up text field requesting specific tool names appears in place. "Other" fields for both select dropdowns and checkbox groups are managed through two separate handler functions (setupOtherField for selects, setupCheckboxOther for checkboxes) that toggle visibility and the required attribute simultaneously, so validation stays consistent regardless of path.
Checkbox constraint enforcement
One section asks respondents to select up to two AI applications they would find most valuable. A setupLimitedCheckboxes() function iterates over the named group and reverses any selection that would exceed the maximum, with an immediate alert. It is a light but effective client-side guard before anything hits the backend.
Google Sheets integration
On submission, form data is serialized to URL-encoded key-value pairs and dispatched via fetch() with mode: 'no-cors'. Because no-cors prevents reading the response body, the form resolves to a success state optimistically after a short delay. Before transmission, each submission receives a pseudo-random participant ID in the format P0000 (generated via Math.floor(Math.random() * 10000).toString().padStart(4, '0')) and a UTC ISO timestamp. Checkbox groups are serialized as JSON strings to preserve their array structure within the flat schema Google Sheets expects.
The Google Script endpoint in script.js is the single external dependency. Replacing it with your own Apps Script Web App URL is all that is needed to redirect submissions to a different sheet.
The participant ID generator is intentionally minimal. Swapping Math.floor(Math.random() * 10000) for a crypto.randomUUID() call adds collision-resistance if you are expecting a larger sample or running multiple concurrent deployments. That also changes the feel of the data downstream, since UUIDs carry no implicit sequence, which might matter if you are using participant IDs for anything beyond anonymous linking.
Survey items and checkbox options live directly in the HTML, so restructuring the instrument does not require touching the JavaScript. Adding a new conditional field means calling either setupOtherField() or setupCheckboxOther() in the initialization block, depending on whether the trigger is a dropdown or a checkbox.
The whole structure transfers cleanly to different research populations. Adapting it for, say, teachers assessing AI in education or clinicians reflecting on diagnostic support tools would mainly be a matter of rewriting label text and option values. The section scaffolding, scale presentation, and conditional logic carry over without modification.
No dependencies to install. Open index.html directly in a browser to preview the full form. Submissions will fail without a live Apps Script endpoint, but the browser console will log the serialized payload so you can verify data shape before wiring it up. To test the full pipeline, deploy a Google Apps Script Web App set to execute as yourself with anonymous access, then replace the URL at the bottom of script.js.
Clone the repo with: git clone https://github.com/yourusername/manager-ai-assessment.git