Legal

Privacy Policy

How Crucible Group handles personal data across the marketing site and the platform's three lines of work: what is collected, who processes it, how long it is kept, and the rights available to the people it describes.

Version 1.1.0 · Effective October 8, 2026

Overview

We build advisory software and deliver advisory services. Both run on records about people, so this policy says plainly what we collect, why, who touches it, how long it lives, and what you can do about it. Where our practices have gaps, we say so — the same candor we apply everywhere else. A longer, more technical account of how the platform handles data lives at gocruciblegroup.com/trust; this policy is the binding summary.

1. Who we are

Crucible Group, LLC (“Crucible,” “we”) operates the marketing site at gocruciblegroup.com and the Crucible platform at app.gocruciblegroup.com, serving three lines of work: Talent Readiness, Strategic Initiatives, and Retained Advisory. For privacy questions, write to privacy@gocruciblegroup.com. You’ll get the founder, not a form.

2. One distinction that matters

For most records in the platform, our client is the decision-maker and we process on their instructions. If your employer runs its talent program on Crucible, your employer chose what to collect and controls the records; we hold and process them under our agreement with your employer. Requests about those records (access, correction, deletion) go first to your employer — and we help them honor those requests. For everything we collect directly — visitors to our site, people who contact us, prospective clients — we are the decision-maker, and this policy is the whole story.

3. What we collect

If you visit the marketing site: standard server logs (IP address, browser type, pages requested, timestamps), kept briefly for security and operations. We do not run advertising trackers.

If you contact us or submit the interest form: what you give us — name, email, company, role, approximate organization size, and your message. We use free-text fields on purpose; nothing is hidden in them.

If you use the platform as an employee of a client organization:

– Identity and employment context: name, work email, role, reporting line, optional start date.

– Assessment responses, with a consent record captured at the time you take an assessment.

– Review and development content: ratings, written rationales, aspirations, development actions and check-ins, recognition given and received.

– Credentials and proof documents you choose to upload.

If you interact with a Strategic Initiatives engagement as a committee member: your name, work email, and the responses you give through signed one-tap links. Committee members act through those links — we do not create accounts for them or collect anything beyond what the engagement requires.

If your organization is a client: business contact details, engagement and program records, and billing information. Payments run through Stripe; we never see or store card numbers.

We do not collect data from anyone under 18. The platform is a workplace tool.

4. What we use it for

To deliver the services: run assessments and reviews, maintain engagement and program records, send the notifications the work requires, and support the advisory relationship. To operate the business: respond to inquiries, evaluate prospective clients, invoice, and keep the books. To keep the system honest: audit logs, security monitoring, abuse prevention.

We do not sell personal information. We do not share it for advertising. We do not use client records to market to anyone.

5. What AI touches — and what it never does

Parts of the platform use AI (Anthropic’s models) to draft text for human review. The boundaries are structural, not aspirational:

– No model scores, ranks, bands, or decides anything about a person. Assessment scoring is deterministic arithmetic. AI output is a draft a human accepts, edits, or discards.

– For assessment and engagement records, AI processing receives the minimum the task needs — first names, strength names, and narrative text people wrote for the record. It never receives surnames from our structured records, email addresses, scores, or proof documents.

– Workshop capture is one place a document reaches AI as written: a photographed worksheet is read by AI to extract its content, so whatever is on the sheet — including names — reaches the model exactly as written. The upload screen says so at the moment it matters.

– Business development is different, and we say so rather than pretend: to research prospective clients we process publicly published professional information — names, titles, published business addresses — and AI helps draft the outreach. Every message is approved by a human, one at a time, before it is sent. See §7.

– Audit documents under the Audit Readiness Review are the other place documents reach AI — and only after we tell the client in writing that AI-assisted review is available for their engagement. Before any document text reaches the model we strip direct identifiers, replace license numbers with engagement-only tokens, and replace drivers’ names with pseudonyms the model never sees behind; a client may separately elect to pass through four narrow identity classes; and a set of never-classes — Social Security numbers, home addresses, financial data, photographs, drug and alcohol results, Clearinghouse content, investigation-file content, medical examination reports, immigration documents — is never sent under any setting. §5A describes this in full.

– We do not permit our AI processors to train their models on this data.

5A. The Audit Readiness Review

What it is. When a motor carrier engages our Audit Readiness Review, it uploads its own safety records — driver qualification files, hours-of-service records, maintenance records, accident and inspection records, and program-level drug and alcohol records — to a private document room so we can audit them against FMCSA requirements. The people in those records are the carrier’s drivers. They did not choose to be audited; their employer did. We treat their records accordingly.

Who is responsible for what. The carrier is the employer and custodian of the records and decides what to submit; we process the records only to perform the review. The carrier warrants it has the right to disclose the records to us and gives its drivers the notice the law or its own policies require; we supply a driver notice to make that easy.

What we will not accept. Individual drug or alcohol test results, medical review officer or substance abuse professional records, Clearinghouse query results, the content of prior-employer investigation files, medical examination reports and exemption paperwork, Social Security cards, immigration documents, and financial records. If one is uploaded, we remove it and record the removal. The drug and alcohol program is reviewed through its policy, selection records, summaries, query log, and training records — never through anyone’s result.

Motor vehicle records. Carriers obtain drivers’ motor vehicle records under the Driver’s Privacy Protection Act for employment and safety purposes. When a carrier discloses them to us for the review, that is a redisclosure for the same purpose. We use them only for the review and never redisclose them except to the service providers listed in our Data Processing Agreement, under the same restrictions.

AI assistance. Until we tell a client in writing that AI-assisted review is available for its engagement, no audit document is processed by an AI model; a named advisor reads the documents, and the platform’s tooling is arithmetic — sampling, completeness, expiry dates the client typed. Once available: before any document text reaches the model we strip Social Security numbers, dates of birth, home addresses, telephone numbers, and financial account numbers; replace license numbers with engagement-only tokens; and replace each driver’s name with a pseudonym matched on our platform, never in the prompt. Checks that depend on age are computed on our side before anything is sent. Automated redaction is pattern-based and not infallible, and we say so. A client may elect — by a separate, revocable election inside the platform — to pass through four narrow classes: driver name as it appears on qualification documents; license number, class, endorsements, and expiration where the token is not enough; medical certificate dates, restrictions, and examiner identity; and MVR violation entries and dates. Nothing else can be elected. Documents we cannot extract text from are read by the advisor, not sent to a model. Every AI output is a draft; a named advisor reviews it before anything enters the record, and the platform has no path from a model to a client.

Regulation citations. Findings cite our stored, versioned copy of 49 CFR, refreshed quarterly — never a model’s memory. The official eCFR controls if they differ.

Retention. Document files stay in the client’s room for the engagement and 24 months after the report, or until the client asks us to delete them. The engagement record — findings, sampling method, report, and document metadata with driver identities pseudonymized once files are deleted — is kept for at least six years as the record of our professional work, and the client can export it.

6. Who else touches it

We use a small set of processors, each for one job:

– Neon — database (United States)

– Vercel — hosting, file storage, and site analytics (United States)

– Clerk — authentication (United States)

– Resend — transactional email (United States)

– Stripe — billing and payments (United States)

– Anthropic — AI drafting, bounded as in §5 (United States)

For the Audit Readiness Review, Anthropic receives only redacted, tokenized, and pseudonymized document text, plus any identity classes the client has elected. The current list is in our Data Processing Agreement.

Each receives only what its job requires. Beyond processors: we disclose personal information if the law compels it, and if Crucible is ever acquired or reorganized, records transfer with the obligations of this policy attached. There is no other sharing.

7. Email

Our emails are pointers, not content: notifications tell you something needs your attention and link you to it. Scores, ratings, and findings do not ride in email bodies. On the marketing side: we do not buy lists. Outbound introductions are individually researched, written for one person, and approved by a human before each send — never an automated sequence left running. Three touches without an answer and we close the loop. An ask to stop — a reply saying so is enough — is honored permanently: the address is suppressed in software, checked before any message is proposed and again before any send, and never auto-cleared. Every outbound message carries our postal address and the reply-to-stop instruction.

8. Cookies

The platform uses essential cookies only: authentication and session security (via Clerk) and the cookies Stripe requires during billing. The marketing site sets no advertising or cross-site tracking cookies. We measure traffic with Vercel Web Analytics, which uses no cookies at all: visitors are counted by an anonymous hash discarded within 24 hours, and we see only aggregate statistics — pages, referrers, country, device type — never an identified person and never activity on other sites.

9. How long records live

– While your organization is a client and you are active: records are retained to deliver the service.

– After you leave your employer: your access ends immediately. Your records are retained for a window your employer configures — default 24 months, configurable from 6 to 84 — to serve the record-keeping purpose your employer selected.

– Purge: deletion of a departed person’s content (assessments, reviews, recognition received, credentials, proof files) is a deliberate human act, never an automatic sweep. What survives a purge is deliberately minimal: audit trails, anonymized job-analysis contributions, recognition the person gave to others, and a tombstone record noting that a purge occurred.

– Audit Readiness Review records follow §5A’s retention terms.

– Leads and inquiries: kept while the conversation is live or plausibly revivable; we close the loop rather than drip forever.

– Server logs: kept briefly, then discarded.

10. Your rights

Regardless of where you live, you can ask us to tell you what we hold about you, correct what’s wrong, delete what we can delete, or export what’s yours. Write to privacy@gocruciblegroup.com. Where your employer is the decision-maker for a record (§2), we’ll route the request to them and help them honor it, and we’ll tell you that’s what we’re doing. We answer within 30 days. We don’t charge for requests and we don’t retaliate for making them.

Deletion of a prospect’s record is permanent, and the address is suppressed in software afterward so our systems cannot re-create it.

Self-service export is not yet built; today these requests are handled manually, by the founder. That’s an honest limitation, not a policy.

11. Security

Access control is enforced server-side, not in the interface. Every query is scoped to your organization. Managers see their direct reports, not peers and not raw scores. Credential proof files stream only to the employee who uploaded them and their HR. Administrative actions are audit-logged. We hold no SOC 2 or ISO 27001 certification today and say so rather than imply otherwise; security questions go to security@gocruciblegroup.com.

12. Where this policy lives legally

Crucible Group, LLC is a United States company; data is processed in the United States. This policy is governed by the laws of Georgia. If you are outside the US, understand that using the services means your data is processed in the US.

13. Changes

When this policy changes materially, the version and effective date above change with it, and platform users get a notification — a pointer, per §7. We don’t change the rules quietly.

Change log: v1.1.0 — corrected §5 to describe workshop capture and business development accurately; corrected §7 to describe outbound introductions accurately; strengthened §10 deletion; added §5A on the Audit Readiness Review; updated §14 contact. v1.0.0 (effective August 12, 2026) — first effective version.

14. Contact

Privacy: privacy@gocruciblegroup.com · Security: security@gocruciblegroup.com · Legal: legal@gocruciblegroup.com · Everything else: hello@gocruciblegroup.com

Crucible Group, LLC · 8735 Dunwoody Place, Atlanta, GA 30350

Questions about this document: privacy@gocruciblegroup.com