Authority fixed for issue of material: data provenance in autonomous GTM
Data provenance is the record of where each data point came from, when it was obtained and verified, what changed it, and who may use it. Before an AI agent contacts a buyer, two things should be on record: the source of every field it acts on, and the rule that allowed it to take the record. GDPR Article 14 makes the source something you must tell people whose data came from elsewhere, at the latest at your first communication with them.
The 1940s waste audit we applied to the modern revenue org asks whether there is authority fixed for issue of material: who may take stock from the storeroom, and is that rule enforced or social? In a revenue team, the material is data, and the storeroom is the CRM. Most teams can say which rep owns an account. Few can say where the email address on that account came from, or which rule let an agent use it.
Why provenance becomes urgent with agents
A rep looking at a contact notices when the title is three jobs out of date or the company was acquired last year. An agent working through a queue acts on whatever the record says, at whatever volume it is set to. Bad material moves faster through an automated line, which is why the data problems in our pieces on CRM data decay and poisoned pipelines grow once agents are involved.
Provenance does not fix bad data. It makes bad data visible before an agent acts on it, and it makes every action explainable after the fact.
What GDPR Article 14 asks at first contact
Outbound usually runs on personal data the prospect did not give you: enrichment vendors, purchased lists, event scans, public web pages. GDPR Article 14 covers exactly that case. Among the information it requires is “from which source the personal data originate, and if applicable, whether it came from publicly accessible sources.”
The timing matters for outbound. If the data is used to contact the person, the information is due at the latest at the first communication. Otherwise it is due within a reasonable period, at the latest within one month. A plain-language explainer from Legiscope lists lead enrichment among the sources the article covers. There is an exception where providing the information would be impossible or involve disproportionate effort; it is narrow, and whether it applies is for counsel.
The practical point is simple. If your systems cannot say where a record came from, you cannot give the notice the first message may need to carry. The first message already carries a disclosure question under the EU AI Act, covered in our Article 50 outbound post. Both duties point at the same moment.
The provenance record, field by field
Keep these for every field an agent acts on. For fields agents never read, record-level provenance is usually enough.
| Field | Example | Why it matters |
|---|---|---|
| Source | Enrichment vendor, purchased list, event scan, inbound form, public web page | GDPR Article 14 asks you to name it; trust depends on it. |
| Publicly accessible? | Yes (company website) or no (vendor database) | Article 14 asks for this alongside the source. |
| Obtained on | Date the value first entered your systems | Starts the one-month notice clock when you are not contacting the person. |
| Last verified on | Date the value was last confirmed | Stale titles and addresses are where agents go wrong at volume. |
| Changed by | Which tool or agent last overwrote the value, and when | Enrichment overwrites fields one at a time; this shows which value was used. |
| Purpose and basis | B2B prospecting, with the lawful basis your counsel set | An agent should not repurpose data collected for something else. |
| Suppression checked | Do-not-contact, objection and unsubscribe status, with the check time | A suppressed contact must stay suppressed across every tool. |
| Authority rule | Owner, territory or queue rule that let the agent take the record | Authority fixed for issue of material, made enforceable. |
| Notice given | Where and when the Article 14 information was provided | Shows the first communication carried what it had to. |
Field level is the point. Enrichment tools overwrite one field at a time, so a record stamped “source: vendor A” may carry a job title from vendor B and a phone number a rep typed in. Your GTM stack inventory tells you which tools can write to each field; the provenance record tells you which one did.
Authority fixed: who may take a record
In most teams, the rule for who may work a lead is social. Reps know not to touch a named account; managers settle disputes. An agent has no social sense, so the rule has to be written down and enforced by the system: record ownership, territory, queue, suppression and do-not-contact status, and any objection on file. Access scoping is the other half, covered in our least-privilege CRM permission audit.
Check it at the moment of action
Provenance recorded at import and never read again does not help. The check belongs immediately before the agent acts.
- 01
Source known
Every field the action uses has a recorded source. An unknown source is a stop, not a warning.
- 02
Fresh enough
The value was verified inside the window your team set for that field. Titles and emails go stale faster than company names.
- 03
Not suppressed
The contact is checked against do-not-contact, objection and unsubscribe lists at the moment of action, not at import.
- 04
Authority satisfied
An ownership, territory or queue rule allowed this agent to take this record. If not, the record goes to its owner.
- 05
Recorded
The result of the four checks is logged with the action, so the reason it was allowed is on file next to what was done.
Where PrescientIQ fits
PrescientIQ’s agents work in your CRM, and every action they take is recorded with its rationale, its before-and-after state, and the approver or policy behind it. That is the record of what the agents did with the data. Where third-party data originally came from still has to be captured in your CRM and your data vendors’ records; together, the two show origin and use. Your team chooses how each action class is governed:
Human-in-the-loop (HITL). The action is drafted and held. It does not execute until a named person on your team approves it.
Human-on-the-loop (HOTL). The action executes under a standing policy your team sets. A named person supervises and keeps intervention, override, and revocation authority.
Your team chooses the mode for each action class, based on its risk tolerance, and can change it at any time.
Every action, in either mode, is recorded to the audit ledger with its rationale, before-and-after state, and the approver or policy behind it.
Every action class starts in human-in-the-loop until your team changes it.
See how the Revenue Accelerator Stack works in an existing CRM, or how Compliance Shield supports regulated teams.
Where it runs
PrescientIQ is hosted and operated by MatrixLabX on Google Cloud. SOC 2, ISO 27001, and PCI DSS attestations are held by Google Cloud, which operates the underlying infrastructure. They are not MatrixLabX certifications. MatrixLabX application-layer SOC 2 is in progress.
Frequently asked questions
- What is data provenance in sales and marketing?
- Data provenance in sales and marketing is the record of where each data point came from, when it was obtained and last verified, which tools changed it, and on what basis it may be used. For AI agents it answers two questions before any action: can this data be trusted, and is the agent allowed to act on it.
- Does GDPR require you to tell prospects where you got their data?
- Yes, in most cases. When personal data was not obtained from the person, GDPR Article 14 requires telling them which source it came from, and whether that source was publicly accessible. If the data is used to contact them, that information is due at the latest at the first communication. Narrow exceptions exist; counsel decides whether one applies.
- What does “authority fixed for issue of material” mean for AI agents?
- It means a written, system-enforced rule for which records an AI agent may take and act on, instead of an informal understanding. The rule covers record ownership, territory, suppression and do-not-contact lists, and objections. If the system cannot show that a rule allowed the agent to pull a record, the agent should not act on it.
- Should provenance be tracked per record or per field?
- Track provenance per field for every field an AI agent acts on, such as email, job title, company and phone. Enrichment tools overwrite fields one at a time, so a record-level source hides which vendor supplied the value the agent used. For fields agents never read, record-level provenance is usually enough.
- When should an AI agent check data provenance?
- An AI agent should check data provenance at the moment of action, not at import. Before it contacts someone or changes a record, it should confirm the source is known, the value was verified recently enough, the record is not suppressed, and an authority rule allowed it to act. A failed check routes the record to a person.
- Does PrescientIQ track where my prospect data came from?
- PrescientIQ records every agent action to its audit ledger with the rationale, the before-and-after state, and the approver or policy behind it. That covers changes the agents make. Where third-party data originally came from still has to be captured in your CRM and data vendor records. Together, the two records show origin and use.
Related reading
- The 1940s Waste Audit That Still Diagnoses Your 2026 Revenue Org
- A Record of Tools Available: The GTM Stack Inventory Most Companies Lack
- Disclosure Is an Inspection Step: EU AI Act Article 50 in the Outbound Workflow
- Why Your CRM Data Decays Faster Than Your Cleanup Cycle
- What “AI-Ready Data” Means for Revenue Teams
Sources
- Regulation (EU) 2016/679 (GDPR), Article 14 — EUR-Lex
- GDPR Article 14: indirectly collected data — Legiscope
This post is general information, not legal advice. Whether and how GDPR Article 14 applies to a given outbound program, including any exception, is a question for your counsel.
See where your own execution effort is going
The Autonomous Audit Report models where your team's execution capacity is currently spent, what your configuration is actually paying for, and what the governed alternative looks like on your own data — before any commitment.
Get your free AAR benchmark