Legal
DPA Annex T — DM Triage
Version 1.0 · Effective date: 2026-11-17
This Annex supplements the Data Processing Agreement (version 1.3.1, "DPA") between the accepting business ("Controller" or "Customer") and productivity-boost.com Betriebs UG (haftungsbeschränkt) & Co. KG, Reichenberger Str. 2, 94036 Passau, Germany, Amtsgericht Passau HRA 12725 ("Processor"). The Processor is represented by its general partner productivity-boost.com UG (haftungsbeschränkt), represented by managing director Florian Standhartinger. The Controller accepts the DPA and this Annex electronically before any Zendesk connection or retrieval of its data, including free reports, shadow mode and dry runs. We log the accepting person, Customer/Zendesk account, versions and time; electronic acceptance satisfies Art. 28(9) GDPR. The German version prevails.
T1. Scope and instructions
T1.1 For DM Triage, the Service and Main Contract in DPA 1.1 include the DM Triage Additional Terms. This Annex applies only to processing on the Customer's behalf. Account administration, billing and operational aggregate usage data used for our own purposes are covered by the privacy addendum, not by this Annex.
T1.2 The DPA remains applicable except that T2–T6 replace Annex 1 sections 1–5 and 8 and supplement section 6; T5 replaces DPA 11.1; T7 supplements and, in case of conflict, overrides Annex 2 for this app; T8 supplements Annex 3. The API's statements about not storing request content do not describe the DM Triage database, which stores results, field values, undo records and encrypted report snapshots. They continue to apply to the inference gateway within the limits of DPA Annex 1 section 5.
T1.3 Documented instructions include connecting Zendesk, selecting fields/groups, thresholds, shadow/live mode, backfill and report requests, enabling writes, undo, deletion and disconnect. DPA 2 governs additional instructions. Where the Customer itself acts as processor for another controller, it must obtain authority for our engagement as its sub-processor.
T2. Subject, nature and purpose
T2.1 The subject is classification and routing of Zendesk support tickets. Processing consists of OAuth-authorised retrieval of ticket subject, description and channel and relevant field definitions/values and group names; transmission to Decision Models EU-tier inference; computation of classifications and probabilities; display and storage of results; writing authorised values to Zendesk; maintaining an undo change log; producing accuracy reports; and deletion. Report snapshots let the Customer inspect the original ticket inputs used in an accuracy report. They are the sole intended persistent storage of ticket subject/description in the app, encrypted under T7.2.
T2.2 Purpose is solely to supply these functions under the Customer's instructions. No training or fine-tuning on Customer Content takes place. Customer Content is not sent to development or support agents, third-party AI tools or used for published model evaluations. The requested EU-tier inference is the contracted processing in T2.1.
T3. Data and people
T3.1 Data types are determined by the Controller: ticket subject/description, channel and identifiers; field definitions/values, group names; classification labels, probabilities and before/after field values for undo; report snapshots; and agent IDs/emails where used to connect or attribute Customer actions. Ticket text and custom field values may contain names, contact data, messages, complaints, opinions, health information or other personal data. Art. 9 and Art. 10 data require the Controller's assessment and lawful basis under DPA 3.2. IDs, field values and predictions may remain personal data even without ticket text.
T3.2 Subjects typically include ticket requesters, the Customer's customers/prospects, employees and agents, connecting administrators, and people mentioned in tickets. Administrator contact and login data used for our own account purposes are separately covered by the privacy addendum; their use in a Customer-directed audit record remains within this Annex.
T4. Locations and recipients
T4.1 App and live Postgres database: Hetzner HEL1-DC6, Helsinki (Tuusula), Finland. Streaming standby replica: Hetzner HEL1-DC7, Finland (LeChonk). Local database dumps: HEL1-DC6, approximately two days. Snapshot text reaches these database copies only as application-level ciphertext; its key is held only in the app environment, not in database dumps, replicas or backup repositories.
T4.2 Inference uses the Decision Models EU tier only, through its EU gateway, and may be carried out by UpCloud Oy (EU data centres listed on the sub-processor page) and GPU Server Hub (4DataCenter Katowice, Poland). This Annex's explicit acceptance authorises these listed providers for this product from activation, including Katowice. No Peer-to-Peer, Lium, worldwide or independent network processing is used. The gateway and usage database do not store request content as described, and subject to the limits, in DPA Annex 1 section 5. That provision does not exclude operating-system swap or crash dumps.
T4.3 Customer Content is processed and stored only in the EU; no transfer to third countries takes place. Stripe and Resend receive no Customer Content from the app.
T4.4 Zendesk is the Customer's source and destination system, selected and contracted by the Customer; the Customer controls its Zendesk locations, permissions and retention. Zendesk is not appointed by us as a sub-processor under this Annex. We make no EU-only promise for the Customer's own Zendesk deployment.
T5. Retention, return and deletion
Deleting a ticket in Zendesk does not automatically delete an existing report snapshot; it remains until the seven-day expiry or earlier on-demand snapshot deletion.
T5.1 Ticket subject/description is held transiently for processing, except encrypted accuracy-report snapshots retained in the live database for a maximum of seven days after creation, or until earlier deletion on request. Snapshot deletion does not mean immediate erasure from historical backups. Field values and undo records must not be used as an alternative store of ticket text.
T5.2 Ticket IDs, classifications, probabilities, field values, configuration and undo logs remain while the connection is active to provide results and undo. On disconnect processing stops, the app's stored Zendesk credentials are deleted immediately, and all Customer Content is purged from the live database within 30 days. Snapshot expiry continues during that period. Other deletion instructions are handled under DPA 2 and 8; request support at info@productivity-boost.com for deletion or correction of an individual ticket's stored records.
T5.3 Live deletions propagate to the streaming replica at HEL1-DC7. Deleted data are removed from the local dumps on rotation, at the latest 180 days after deletion from the live database; local dumps are normally retained for approximately two days. Backups are used only for recovery. Before restored data are made accessible or processing resumes, we reapply snapshot expiry, disconnect, credential revocations and Customer deletion instructions using a content-free deletion record. Encryption alone does not mean deletion while a usable key remains.
T5.4 The Controller may export retained results before disconnect or request return before the purge deadline; the Processor then deletes its remaining copies as above, unless Union or Member State law requires retention. Ticket texts not retained by the app cannot be exported from it later. We confirm deletion in text form on request (DPA 11.2). This does not delete originals or prior changes in the Customer's Zendesk.
T6. Duration
Processing continues for the product contract and thereafter only as needed for deletion and restricted backup rotation in T5, consistent with DPA 1.4. Statutory billing retention does not authorise retaining ticket content.
T7. Technical and organisational measures
T7.1 DPA Annex 2 applies, with the app-specific measures below. Access to credentials, results, reports and undo is restricted by Customer account and authorised administrator; tenant ownership is checked on each operation. Zendesk scopes are limited to the contracted functions. Browser/app and external API connections use TLS; the gateway-to-inference path has the protection and relay limits stated in DPA Annex 2 section 2.1.
T7.2 Accuracy-report snapshots are encrypted before database storage using Fernet and an application key kept only in the app environment. Database replicas and dumps therefore contain ciphertext of snapshot text; the key is excluded from those copies and from Sandy/LeChonk job environments. Other stored fields are not asserted to have this application-level encryption. Local dumps are not asserted to have whole-dump encryption.
T7.3 OAuth credentials are protected against unauthorised access and deleted from the app immediately on disconnect. Hosted-login tokens are stored only as hashes, valid for 15 minutes. Application logs, traces, error reports, email, payment metadata and analytics must contain no ticket text, snapshot plaintext or OAuth/login secrets. No such data enter development/test fixtures. Operating-system swap and crash dumps are subject to the limits in T4.2; no broader "never on disk" claim is made.
T7.4 A named person may read ticket content only at the Customer's request for support and only as necessary, under confidentiality. No automated agents or third-party AI tools read Customer Content, including results and records in the standby replica. Development, testing and quality measurement use our own test/sample data. Database dumps containing Customer Content are not transferred to our development environment. Application encryption of snapshots alone does not protect other personal field values and undo records. Snapshot keys must never enter the workshop. Neither possession of a ciphertext copy nor a support request authorises an agent to decrypt/read it.
T7.5 Separation of live processing and backup recovery, deletion jobs, restricted key access and restoration-time expiry/revocation checks implement T5. DPA sections 4, 5 and 8–10 continue to apply, including confidentiality, security, data subject assistance, breach notification and audits. Certifications remain limited to the provider/location scope stated on the sub-processor page; no certification is asserted for DM Triage itself or extended to HEL1-DC7.
T8. Sub-processors
| Sub-processor | Function and location for DM Triage |
|---|---|
| Hetzner Online GmbH, Industriestr. 25, 91710 Gunzenhausen, Germany | App/live database and local dumps: HEL1-DC6; standby replica: HEL1-DC7; EU inference gateway: Helsinki, Finland. Snapshot text is application-encrypted in database copies. |
| UpCloud Oy, Helsinki, Finland | GPU inference in the EU data centres listed, with their applicable dates, on the sub-processor page. |
| GPU Server Hub (4DataCenter Katowice, Poland) | EU GPU inference; listed provider identity, contractual and data-centre details as on the sub-processor page. |
The current product-specific list is at /legal/subprocessors. Acceptance approves this initial list only for DM Triage; it does not waive rights for other services. Subsequent additions/replacements follow DPA 6.2 (notice by email and list update, generally at least seven days, with the stated objection rights; the DPA's legacy 30-day rule remains where applicable). DPA 6.3 requires equivalent contractual protection and keeps the Processor responsible. If an objection ends the paid product, the unused prepaid base-fee portion is refunded; valid accrued excess charges remain due. Stripe/Resend account-data roles are in the privacy addendum.
T9. Rights and contact
Direct requests from ticket subjects should go to the Customer; we assist under DPA 8. Customer instructions, support, return and deletion requests: info@productivity-boost.com. Previous versions.