Main site

PR AI agent · Journalist pitch matching

Journalist Pitch Matching AI Agent

A file on a reporter who never agreed to it and does not know it exists: record where each line came from, fix the jurisdiction before the send, then let a named partner sign.

Typical delivery
4–6 weeks
Deployment
Your stack
Named partner
File first
After launch
Agent Care

What this agent does

Drafts the approach, never releases it

In

  1. 01

    A reporter enters the file, and the source of each line in it is written down beside it.

  2. 02

    A record is enriched from a third party, and the notice that record owes is scheduled, not deferred.

Reason

  1. 03

    A story is matched to a reporter, and the lines in the file that selected them are named in the match.

  2. 04

    A draft is written, and its subject line is tested before a news offer turns into an advertisement.

  3. 05

    A recipient resolves to a country and a subscriber type, because one market says opt-in and the next does not.

Decide

  1. 06

    Something of value is offered, and the offer and who was told of it are logged before the pitch goes.

  2. 07

    An item goes out for a foreign principal, and the forty-eight-hour filing clock starts with it.

Out

  1. 08

    A reporter goes quiet, and the silence is recorded as unreadable rather than scored as a result.

  2. 09

    Execute send or write actions only inside the approval boundaries agreed at implementation.

Product statement

Matching, drafting and logging belong to the agent. Release belongs to a named partner, who signs for what went out and to whom.

Example workflow

One pitch, file to release

  1. 1Reporter file assembledBylines, beat, outlet, stated preferences, prior contact and where each line came from
  2. 2Recipient fixed and classifiedThe country, the subscriber type, the corporate form and the day that classification was set
  3. 3Match and draft raised togetherThe lines that selected the reporter, the draft, the subject line and any offer inside it
  4. 4Controls appliedNotice checks, jurisdiction checks, subject-line classification and disclosure checks

No human action required

Stages 1 to 4 run unaided, and nothing leaves the building at any of them — the agent is drafting, and the partner lane opens at the release gate.

5DecisionSplits at the release gate
Ordinary trade pitch

Goes to the named partner to release.

Anything for a foreign principal

Adds a counsel read first.

Partner review

The pitch is held with the file lines that selected the reporter, the classification and the offer log.

Send · Append source · Hold for partner review

Released — by the named partner

  1. 6Outbound and filing records updatedOnly where send access and the records policy permit it
  2. 7Outcome evaluatedNotice currency, classification accuracy, partner corrections and what review found

Corrections

Each partner correction is counted in the evaluation.

What should not run autonomously

Human approval stays in control

Outside the boundary — human approval required

8 items

  • Releasing a pitch to a named reporter.
    • Deciding an item is not informational materials.
    • Judging whether a mandate sits outside FARA.
    • Signing the six-monthly statement under oath.

  • Automation boundary

    Agent acts unaided

    • ✓Write down the origin of each line in the file.
    • ✓Schedule the notice a new record owes before it is enriched again.
    • ✓Fix the country and the subscriber type before the draft is raised.
    • ✓Log the item, the channel and the named recipient.

    Nothing reaches a reporter except by a named partner, inside the agreed boundaries.

    • Judging whether an offer needs disclosing.
    • Telling counsel the outbound log is complete.
    • Choosing which reporters a mandate may approach.
  • Changes to the file, the send list or the classification.

Example output

One reporter file, annotated

This serves a press office that may have to account for a named contact years after the pitch went out; below is one record exactly as the agent leaves it.

Reporter file · single contact

Illustrative example
Selection
Technology desk, named correspondent
Recorded as
Matched on three filed lines
Outlet
Provenance, per line
Evidence of record
Published byline, 3 August 2026
Confidence
Held unsent
Held for
The named partner, by name

As received

Built from published bylines and one vendor record, and it claims nothing the reporter has not printed.

What the file holds

  • Published bylines
  • Vendor record
  • Prior contact log

Why no send here

Approaching a named reporter is a judgement a named partner makes.

What the score decides

Where the client is a foreign principal the draft gets a counsel read before release.

Action

ReleaseAppend sourceSend to counsel review

Value

Where AI adds value

The same four claims, placed at the point in the workflow where each one applies.

Where the value lands

Value 01 – 04

Each pitch drafted

From the file lines that selected the reporter

03Evidence

Where each line is put to work

The agent does not vouch for a reporter file, only for where each line came from, when it was taken, and which of those lines selected the recipient.

01Approved path

A file on a named reporter

Move the country, the subscriber type or the question of whether a partnership counts as corporate, and the same message changes footing, with nobody having lied.

02Human review

What was searched for, and not found

No data-protection authority in the sources searched was found to have ruled on a PR or media contact file as such, while a control query on those same sources returned live data-broker enforcement, so the surface works and what is missing is a carve-out rather than the risk.

04

Build an evidence trail

The pitch, the file that selected the reporter and the person who sent it stay together.

Integrations

Typical integrations

Five system groups connect to the same agent. Which of them are in scope is decided in discovery.

  • Published journalism

    • Bylines
    • outlet mastheads
    • Beat, desk and coverage filed lately

  • Contact vendors

    • Media CRM
    • list vendors
    • Names, addresses and provenance

  • Your own contact history

    • Sent mail
    • embargo records
    • Approaches made and any replies received

Agent

Pitch matching and outbound records

  • Reads the file
  • Raises the match and the draft
  • Holds for the partner

  • Filing and registry sources

    • Registry filings
    • public notices
    • Mandates carried for a principal

  • Observability & evaluation

    • OpenTelemetry
    • Langfuse
    • Supported monitoring/evaluation sources

Integration availability depends on the client's existing systems and API access.

Agent controls

Six sieves between the file and the reporter

Six sieves down one column, the last the finest. What settles is set out in the map below.

L6 · Outermost — last line of defence

Inward → L1 · closest to the model

L6Rollback / safe modeNarrow the agent to file listing when evaluation or production signals degrade.Roll back
L5Version trackingTrack model, prompt and classification rules, and note the version each pitch was drafted under.Track
L4ProvenanceRecord each file line, where it came from, the notice it owes and each read of the record.Record
L3Partner releaseHold the pitch for a named partner; the hold governs release, not whether the pitch is any good.Gate
L2Recipient guardrailsTest each recipient against its fixed country and subscriber type, and refuse a send whose footing moved.Restrict
L1Disclosure thresholdsRoute an item carrying an offer to a counsel read before the pitch reaches a reporter.Require review
Model corePitch drafted — the file lines, the classification, the offer log and the channel
L1 – L2
Test whether a pitch may go
L3
Leaves the release to a named partner
L4 – L5
Keep the pitch and the file behind it
L6
Sends nothing outbound when signals degrade

How Nestack evaluates it

Evaluate the whole approach — not only the pitch that comes out.

Coverage runs the whole depth of the workflow, and every layer is cut by slice.

Surface — the pitch a reporter opens

  1. E1Final-output evaluationDid the pitch record the file lines it was actually built on?
  2. E2Step-level evaluationDid the agent read the right reporter, the right beat and the live classification?
  3. E3Tool evaluationDid it read and write the correct contact and the correct outbound entry?
  4. E4Confidence calibrationDo low-confidence matches actually attract more partner corrections?
  5. E5Slice evaluationHow does performance change across specific outlet classes?
  6. E6Business outcomeHow many pitches needed a correction before the partner released?

Floor — the file a pitch rests on

Failure modes

Where each failure originates in the agent

Seven failure modes, each set at the stage where it first shows.

Agent lifecycle

Direction of processing →

  1. 01 · Retrieval

    1 mode

    • TW-01

      Notice never issued

      A record is enriched and the notice it owes is not sent.

    Stage gathers

    Bylines, vendor lines, beats and prior contact

  2. 02 · Reasoning

    2 modes

    • TW-03

      Subject line turns commercial

      A news offer is rewritten into an advertisement.

    • TW-05

      Paid promotion left undisclosed

      The draft omits the consideration behind it.

    Stage proposes

    The file lines, the draft and the offer log

  3. 03 · Tool / write

    2 modes

    • TW-02

      Recipient jurisdiction unfixed

      The send goes out on the wrong national footing.

    • TW-04

      Offer made with no disclosure chain

      Value is offered and nobody downstream is told.

    Stage writes

    Only where write access and approval policy allow it

  4. 04 · Output

    1 mode

    • TW-07

      Practitioner blocked, silently

      A named person is blocked and no signal returns.

    Stage returns

    The pitch a reporter opens, and what it asked

  5. 05 · Change / Version

    1 mode

    • TW-06

      Outbound log cannot be rebuilt

      Six months of sends sit in systems that do not agree.

    Stage tracks

    Model, prompt, matching rules and file dates

  • Sev-1 · a pitch sent on an unfixed footing
  • Sev-2 · a sent item never reaches the log
  • Sev-3 · file degrades, pitch held back
See the complete Public Relations failure-mode catalogue

Affected slices

Foreign-principal work absorbs the corrections

An outlet-level correction figure can read clean while foreign-principal mandates carry most of the disclosure rework. Nestack reports the correction rate by mandate class, not only in total.

Slice performance — reported separately, not only in aggregate

Illustrative example
SliceFailure rateLiftLift vs. thresholdStatus
Foreign-principal mandates10.8%3.7×Review
Multi-market send lists7.7%2.6×Review
Vendor-enriched records4.8%1.6×Watch
Established named contacts2.4%0.8×Normal

Bar: correction-rate lift vs. established-contact baseline · scale 0–4.0× · tick marks the 2.0× review threshold

2 of 4 slices over threshold

Evidence-linked improvement

What an unsourced file costs

The loop shuts when the reporter contacted from a file never disclosed is a standing case. That suite is what the next send list is measured against.

Improvement cycle · five stages

Switchback — the path turns at Improve and returns at Learn

  1. 01

    Detect

    Correction rate rises on foreign-principal mandates.

  2. 02

    Diagnose

    The pitch that reached a reporter whose record nobody could source is worked backwards until one cause is left standing.

  3. 03

    Improve

    Every send goes out numbered, with the files that selected it attached.

  4. 04

    Verify

    Nothing sends while one touched disclosure case is still red.

  5. 05

    Learn

    The case stays on, and the matching rules are redrawn beside it.

  6. Learn → Detect

    The return edge. The next send list is measured against a suite one case longer.

Typical build scope

Twelve workstreams across six weeks

The build scope read against the delivery timeline. Week structure follows the six-week plan — discovery, sources, matching and drafting, evaluation, integration, then production validation and handover.

Workstream

Week 1Week 2Week 3Week 4Week 5Week 6

  1. 01Journalist-file and automation-boundary definition.Week 1
  2. 02Published, vendor and internal sources.Week 1, Week 2
  3. 03Contact-to-recipient and disclosure-coverage mapping.Week 1, Week 2
  4. 04Contact-file ingestion.Week 2
  5. 05File, recipient and outlet binding.Week 2, Week 3
  6. 06Disclosure scoring and review routing.Week 3
  7. 07Partner release workflow.Week 3, Week 4
  8. 08Media-CRM integration.Week 5
  9. 09Disclosure and selection cases.Week 4, Week 5
  10. 10Guardrails and outbound controls.Week 4
  11. 11Send-trail instrumentation.Week 5, Week 6
  12. 12Deployment, documentation and Agent Care handover.Week 6

12 workstreams · 6 weeks · bar shows the weeks a workstream is active — several run in parallel

Final scope and sequence confirmed in discovery

Engagement tiers

What each tier includes

Rows are the capabilities named in each tier's scope. Higher tiers include everything below them.

Capability✓ in scope · — not at this tierPilotOne mandate, one cycleProductionProduction outbound workflowAdvancedMultiple mandates / markets
Introduced at Pilot
Pitch drafting to your house rules✓✓✓
Named partner release✓✓✓
Contact-file baseline✓✓✓
Introduced at Production
Reporting by outlet class—✓✓
Partner review workflow in your systems—✓✓
Approved outbound sends—✓✓
Database-and-sender integration—✓✓
Introduced at Advanced
Multi-market classification——✓
Cross-mandate pitch packs——✓
Large contact files——✓
Multi-market outbound controls——✓
Build priceFrom $5,000From $8,000Custom quote
Final build price
Confirmed after discovery based on integrations, workflow complexity, contact volume, approval controls and deployment requirements.
Separate from build
Build pricing is separate from recurring Agent Care, which covers managed monitoring, evaluations, incidents and verified improvements after launch.

What we need from you

What you bring, and what we build with it

Each input maps to a piece of build scope and a week in the delivery timeline.

  1. 01Your live mandates and the reporters each one names

    File capture and provenance versioning

    Week 1

  2. 02Representative published, vendor and internal sources

    Source binding, match logic and the contact-file baseline

    Week 2

  3. 03Your outbound calendar and the partners it names

    File mapping, provenance binding and the automation boundary

    Week 1

  4. 04Access to the APIs, feeds or exports behind them

    Published, vendor and internal source assessment, then integration setup

    Week 2

  5. 05Files you would not want disclosed

    Selection cases and the evaluation round

    Week 4

  6. 06What no pitch may offer

    Disclosure scoring, review routing, guardrails and release controls

    Week 3

  7. 07A named partner who releases the pitch

    Release to the named partner, then pilot and production validation

    Weeks 5–6

Nothing else is required

Deployment, documentation and Agent Care handover are ours.

Delivery timeline

Four phases across six weeks

Widths here follow cost rather than the grid, which is why the fifth week shows one band under another.

Phase

W1

W2

W3

W4

W5

W6

  1. DiscoveryW1
  2. BuildW2 – W3
  3. EvaluateW4 – W5
  4. Pilot & LaunchW5 – W6

Week focus

W1File discovery, provenance versioning and the automation boundary

W2Source integration and the contact-file baseline

W3Match logic, drafting rules and release controls

W4Evaluation suite, selection cases and failure-mode testing

W5Media-CRM integration, pilot sends and targeted corrections

W6One outbound cycle run under the media-relations lead, then Agent Care handover

Reading the band
Each bar spans only the weeks its own work is named for, and week five is shared by design.
At the end of W6
Once the send record validates, Agent Care assumes the agent.
Duration
Six-week plan shown · typical delivery 4–6 weeks depending on scope confirmed in discovery.

Next step · PR AI agent

Build a pitch matching agent around the reporter file your last send never wrote down.

Show us one send list you reuse each quarter and the file behind the last name on it. If nobody can say where those lines came from, or which of them chose the recipient, then the pitch is being trusted rather than defended. A file nobody can source comes back as a finding.

Typical delivery
4–6 weeks
Deployment
Your live mandates
Named partner
File first
After launch
Agent Care available

Nestack Agents · Media relations

AGT-COM-06 · Agent Care available after launch

Build a journalist pitch matching agent

Four questions so the call starts from your actual setup — then pick a time that suits you.

About you
Your beats

No obligation. We’ll map the workflow, name the automation boundary and tell you if an agent is the wrong answer.