Privacy Policy

15. How Hivewell learns from patterns

Hivewell learns what works best in each trade by counting patterns across companies that use ManageBee. For example, it can count which sections most Reports include or which drafts are sent without changes.

Trade learning starts no earlier than December 2, 2026, and only after all the checks below pass and at least 30 days after these terms are posted and account owners are emailed. Only companies whose owner or admin separately chooses Allow take part. Only eligible work after both the effective date and that saved choice counts. Earlier work never counts. Done or continued use does not give permission.

We count common template sections, types of fields, completed or skipped steps, draft layouts, changes to drafts, and selected feedback reasons. We keep no prompts or before-and-after text in the shared dataset.

Shared patterns hold only counts and shapes, grouped so no company or person can be picked out. They hold no Reports, customer details, prices, photos, voice, or text from any company. Messages to a company's own clients are never counted. The exclusions in section 3 of the AI & Training Data Terms still apply.

Each company counts once per pattern. A pattern needs at least 5 companies, and counts are kept only as rough bands. A finer group is used only when that group and the rest of the trade each have at least 5 companies. Rare shapes are merged or dropped. A person at ManageBee approves each pattern before use. We test whether anyone could reasonably pick out a person or household, link them to other data, or learn private facts about them. Before any pattern is used, we test on sample companies that planted names, addresses, codes and prices never appear in a pattern, and that no single company can be picked out from the patterns, from patterns combined, or from how they change over time. We check again after removals. Five companies alone is not proof that a pattern is anonymous.

We will not try to work out which company or person a pattern came from. We hold the companies that help run Hivewell to the same promise, and we never sell patterns. Hivewell's trade learning uses Google (Vertex) under our existing agreement. Before we send this data to any provider, we check that its terms bar it from training on the data, keeping it longer than needed, or passing it on. Before receiving patterns, every recipient must agree in writing to keep them safe, use them only in anonymous form, make a public promise not to try to work out who they came from, and hold anyone they pass them to to the same rules. For shared trade learning, no customer-derived data goes to TypeSafe (Jev) until a data agreement with it is signed and these checks pass.

An owner or admin can choose Allow or Don't allow in Hivewell settings > Privacy. They can also turn off each kind: How Reports and messages are written, How work is set up, and Answers that were not right. The first kind never includes messages to clients. Each person can turn off their own contributions in Settings > You > Privacy and data > Help improve Hivewell. To turn Hivewell off for the company, use Hivewell settings > General > Hivewell on.

Opting out excludes activity from the time of your saved choice; every queued contribution is checked again before counting. Work from a time when a switch was off is never added later. Choosing Don't allow for trade learning does not turn off Hivewell's help.

You can ask us to delete your company's linked shapes. Shared trade patterns are fully anonymous. Pattern records linked to your company keep the same protections and deletion limits as your other data, and are deleted with your company. Only patterns proven anonymous may remain afterward. Linked shapes are deleted with the company under the Data Terms, including their backup limits. Patterns may be kept after a company leaves only because they are fully anonymous, with no way to trace them back to that company or a person.

ManageBee decides how personal data is used for shared learning (in legal terms, we act as a controller). We rely on a legal ground called legitimate interests: improving Hivewell while weighing that benefit against people's privacy. Before reading source work, we record why this use fits the purpose the data was first collected for, why the data is needed, and how we weigh privacy risks for owners, staff and homeowners. We need limited shapes from several companies to learn what works across a trade. We weigh that benefit against people's privacy through the exclusions, anonymity checks, deletion rules, and choices above. An owner's Allow choice is not personal consent for staff or homeowners. Anyone named in source work can object or ask for deletion of linked shapes by emailing privacy@managebee.com, even without an account. See the Data Terms for the roles and lawful-basis record.