AI Disclosure

1. Where AI is used in Chronox

Chronox uses artificial intelligence in a few, clearly bounded ways. Each of these is generative AI - it writes words - unless noted otherwise:

• The Chronox assistant. An optional chat and helper feature that answers questions, drafts text, explains your data, and - for users whose role permits it - can carry out routine actions you ask for (see §2). It is powered by third-party AI models (currently Anthropic and OpenAI), can read images you attach, and can look things up on the public web when a question calls for it (see §6).
• Summaries, weekly reports, and recaps. Optional plain-language write-ups of your own timekeeping, project, and task activity, including the weekly report and its "dive deeper" view. The AI writes the narrative around figures that a deterministic engine computes (see §5).
• Card analyzer - "Tap-to-Explain." On request, explains a single record you tap - a time entry, job, site delay, extra-work item, tools-used entry, cost code, or notification - in a few short sentences covering what it is, its current state, and a suggested next step.
• Info center - card info pop-ups. Short, plain-language descriptions of what a dashboard card or metric means and how to use it. These describe the feature in general; they are generated from the card's name and Chronox's own product documentation, not from your company's data, and they are cached and reused so most are pre-written and shown again rather than generated fresh.
• Notifications and nudges. Some notifications are written by AI from your own data - for example, flagging entries that need attention or summarizing what's pending. They point you to something to review; they do not take action on their own.
• Scheduled AI tasks. You can ask the assistant to run a task on a recurring schedule - for example, a Monday-morning summary. These run under the same rules as the assistant: anything that would create or change records, or touch pay or money, is prepared as a draft for a person to review and is never finalized automatically (see §2).
• Document capture and extraction. If you photograph a receipt or delivery ticket, AI reads the image and proposes a structured draft (vendor, amount, date, line items). Nothing posts to a job or to cost until a person accepts the draft, and which job it is coded to is chosen by ordinary text matching, not by the AI.
• Import field mapping (AI-assisted). When you import records from a spreadsheet, an AI model may suggest how your columns line up with Chronox fields. This is a suggestion only, it is confirmed by a person, and it is described in detail in §3.
• Behind-the-scenes construction knowledge (retrieval). To answer general "how does this work in construction" questions, the assistant can draw on a general construction knowledge base using AI text-embeddings. This knowledge base is general reference material, not your company's data, and is kept separate from it.
• Facial-recognition identity verification (a separate kind of AI). If your company turns on Face ID, facial-recognition AI confirms who is punching at clock in/out. This is a narrow verification tool with a human reviewing the resulting records - it does not rate, rank, or make employment decisions. It is covered in depth in the Privacy Policy (§12) and the Face ID Choices page, and is never connected to the generative assistant above. Chronox also offers a photo (selfie) timekeeping mode that uses no facial recognition and no AI - a person confirms identity from the photo.

These generative features are optional - you can use Chronox for timekeeping without them - and they run under Chronox's role-based data rules: an AI feature only ever works with data your role is already allowed to see, and nothing it shows you crosses those permissions. Your company can limit or disable them.

The numbers in Chronox - hours totals, dollar figures, forecasts, insights, and predictions - are produced by deterministic (non-AI) calculation engines, not written by AI. See §5.

2. What the assistant can and cannot do

For users whose role already grants the underlying permission, the assistant can do more than answer questions - on your instruction, it can help create or edit records (such as a time entry, tools-used entry, extra-work item, site delay, daily log, job, or cost code), draft and generate files, send a message by email to members of your own company, and set up the scheduled tasks described in §1. It does these things as a convenience, and always within limits:

• Only on your instruction. The assistant does not act on its own initiative. It acts when you tell it to.
• Only within your role. It can never do something your own account is not allowed to do. Foremen and field workers use the assistant in an answer-only mode - it will not create or change records for them at all.
• Never a final pay or employment decision. Steps that would set or change pay, run payroll, or move money are prepared as a draft for a person to review and approve - the AI does not commit them. The assistant does not approve or deny time, dock pay, discipline, hire, fire, or schedule workers on its own.
• Email stays inside your company. When the assistant sends email for you, recipients are limited to verified members of your own company; it is not a tool for emailing the outside world.
• You are accountable for what you ask it to do. An action you direct the assistant to take is treated as your action, and you remain responsible for reviewing the result.

Because an action the assistant proposes can be based on an AI misunderstanding, review what it did or drafted before you rely on it, just as you would review the work of a new assistant.

3. Import field mapping (AI-assisted)

When you bring records into Chronox from a spreadsheet (CSV), Chronox has to work out which of your columns corresponds to which Chronox field - for example, that a column called "Emp Rate" is the pay-rate field. This "field mapping" step is AI-assisted, and it works in a deliberately conservative way:

• Deterministic matching comes first and is authoritative. Chronox first matches your column headers against a built-in list of field names and known synonyms using ordinary (non-AI) text matching. If that succeeds, that is the mapping.
• AI only fills the gaps, as a suggestion. For columns the deterministic matcher could not confidently place, an AI model (currently OpenAI) may propose a likely mapping. If the deterministic matcher and the AI ever conflict, the deterministic match wins. If the AI is unavailable, mapping still works using deterministic matching alone.
• A person confirms every mapping. The proposed mapping is shown to you for review and confirmation before any data is imported. The AI's role ends at suggesting labels.
• Your personal data is redacted before anything is sent. To help the model tell, say, a "rate" column from a "date" column, Chronox may send your column headers and a small sample of cell shapes. Personal values in that sample are replaced with generic tokens (for example <text>, <email>, <phone>) before leaving the platform - so an employee's name, email, or phone number is not transmitted to the AI provider. Numbers and dates are kept because they help identify a column's purpose and are not personal identifiers.
• The AI never imports, reads, or writes your records. It only labels columns. The actual validation and import of your data are performed by deterministic Chronox logic, not by AI.

4. Chronox AI can make mistakes

Every generative AI feature above runs on generative AI models. Like all such systems, they can produce output that is inaccurate, incomplete, outdated, or entirely made up ("hallucinations") while still sounding confident and correct. They may misread an attached document or a record, misunderstand a question, mislabel an imported column, or miss context you did not provide. Because of how these models work, the same question can produce different answers at different times.

5. The numbers vs. the words

When the assistant, a summary, a report, or a card explanation shows a dollar figure or an hours total, that number is calculated by a deterministic engine - ordinary arithmetic run on your own punches, rates, and estimates - and traces back to a defined metric. Insights, forecasts, and predictions are likewise produced by deterministic calculation, not written by AI. The AI's job is to write the sentence around the number, not to invent the number, and Chronox runs an automated check that removes figures it cannot tie back to a computed metric. This design reduces the risk of a made-up figure, but it does not eliminate risk: the AI can still describe a number incorrectly, apply it to the wrong period, or draw a wrong conclusion from correct data.

Your time records and payroll data - not the assistant's summary of them - are the system of record. If a summary and the underlying records ever disagree, the records govern. Always confirm figures in your reports and time entries before using them for payroll, invoicing, or any decision.

6. Attachments and web search - what can leave your company

Two AI features can send content beyond your own Chronox records, so they deserve a specific note:

• Attachments. When you attach an image or document to the assistant, or photograph a receipt or delivery ticket for capture, that image is sent to an AI provider (Anthropic or OpenAI) so the model can read it. Only attach material you are authorized to share.
• Web search. When a question calls for public information, the assistant can search the public web through the AI provider's search tools and cite what it finds. This means your question - and the search terms derived from it - reach that provider's search infrastructure. Do not include confidential company information, personal data, or secrets in a request you expect the assistant to search the web to answer. Web results are third-party content and can themselves be wrong or out of date.

7. Not professional advice

Chronox and its AI features are timekeeping software. Nothing the assistant, a summary, a report, a mapping suggestion, or an in-product prompt produces is legal, tax, accounting, payroll, financial, employment, HR, safety, or compliance advice, and none of it is a substitute for a qualified professional. The AI does not act as, and must not be treated as, a lawyer, accountant, payroll specialist, or other licensed professional. For decisions in those areas - including wage-and-hour, overtime, worker classification, tax, and compliance questions - consult your own qualified advisor. This mirrors the "no legal advice" and "AS IS" terms in our Terms of Service (§7).

8. A human stays in charge - no automated decisions

Chronox is designed so that a person, not AI, makes any decision that matters to a worker. The AI does not approve or deny time, set or dock pay, run payroll, discipline, hire, fire, schedule, or make any other significant employment decision on its own. Where the assistant can act at all (see §2), it acts only on a person's instruction, only within that person's role, and any step affecting pay or money is prepared for human review rather than finalized. The AI surfaces information and drafts; your foremen, managers, and office staff review, correct, and decide, and they hold the authority to change any outcome. Because of this human-in-the-loop design, Chronox's identity verification is built to stay outside the "significant decision" definition in California's automated-decisionmaking (ADMT) rules - see the approach summarized in the Privacy Policy (§12). No solely-automated process produces legal or similarly significant effects about a worker.

9. What we send to AI providers - and what we never send

To power these features, Chronox sends the AI providers only the content each feature needs, and only data the user's role is permitted to see:

• for the assistant, summaries, reports, and scheduled tasks: the relevant timekeeping, project, and task text, computed metrics, and any documents or images the user chooses to attach;
• for the card analyzer and AI notifications: the specific record, metric, or pre-computed signal being explained;
• for info pop-ups: the card's name only - no personal or company data;
• for document capture: the image you photographed;
• for import field mapping: your column headers and a privacy-redacted sample of cell shapes, with personal values replaced by tokens (see §3);
• for web search: your question and the search terms derived from it (see §6).

10. Using Chronox AI responsibly

To keep AI use safe for everyone, when you use the assistant or any other AI feature:

• Verify before you act. Check AI output - answers, summaries, drafts, and proposed import mappings - against your real records and judgment before relying on it, sharing it, or making a decision from it.
• Don't paste in what you shouldn't. Only submit information you're authorized to share. Avoid entering other people's sensitive personal information, secrets, or confidential material you don't have the right to disclose - especially in requests the assistant may search the web to answer.
• Don't use it for emergencies or safety-critical decisions. The assistant is not a substitute for professional, medical, legal, or emergency services.
• Review anything the assistant does for you. If you direct it to create or change records, send email, or set up a scheduled task, confirm the result is what you intended.
• It's optional. You can choose not to use these AI features, and your company can limit or disable them.

11. Availability and changes

AI features are provided on an "AS IS" and "AS AVAILABLE" basis under our Terms of Service. We may add, change, limit, suspend, or discontinue AI features, models, or providers at any time, and we do not guarantee that any AI feature will be available, uninterrupted, error-free, or will produce any particular result. Because the underlying models come from third parties and are updated over time, their behavior and quality can change. AI features may not be available in every region or on every plan.

12. No warranty; limitation of liability

Except as expressly stated in the Terms of Service and the Data Processing Addendum, AI features are provided without warranties of any kind, and Chronox disclaims implied warranties of merchantability, fitness for a particular purpose, and non-infringement to the maximum extent permitted by law. Chronox is not responsible for decisions made, or actions taken, in reliance on AI output - including actions you direct the assistant to take - that has not been independently verified. Your use of Chronox, including its AI features, is governed by the warranty disclaimers (§7) and the limitation of liability (§9) in the Terms of Service, which control in the event of any conflict with this page.

13. Feedback, corrections, and contact

If any Chronox AI feature gives you a bad or wrong result, please tell us - reporting it helps us improve the safeguards, and it lets us and your employer correct the underlying records where needed. Corrections to time records are made by a person, logged to an append-only audit trail, and acknowledged in-app, as described in the Privacy Policy. Questions about how AI works in Chronox, or requests about automated processing, can go to your employer or to privacy@chronoxapp.net.

14. Changes to this disclosure

We may update this AI Disclosure as our features or the law change. Material changes are versioned, and the effective date above reflects the current version. Contact: privacy@chronoxapp.net, or Foundation Digital LLC, Attn: Privacy.