Fieldwork → Twenty CRM Sync Contract

Date: 2026-05-11 Version: 1.0 Status: Draft — pending implementation

1. Source of Truth Principle

Fieldwork/Postgres is canonical. Twenty CRM stores CRM-facing synced copies.

All data originates in Fieldwork (CT604 PostgreSQL, lifehouse.fieldwork schema).
Twenty CRM (VM700) receives read-only synced copies for organiser-facing views.
No data flows from Twenty back to Fieldwork in the initial sync direction.
Fieldwork remains the authoritative source for all doorstep, consent, demographic, and engagement data.
If a conflict arises, Fieldwork wins. Twenty records are overwritten on next sync.

2. Table-to-Object Mapping

Fieldwork Table	Twenty Object	Direction	Notes
`locations`	Custom: `fieldworkLocation`	→ Twenty	Wards/streets/addresses
`households`	Custom: `fieldworkHousehold`	→ Twenty	Household groupings
`people`	Twenty `Person` + Custom fields	→ Twenty	Core person records
`organisations`	Twenty `Company`	→ Twenty	Local organisations
`doorstep_contacts`	Custom: `fieldworkEngagement`	→ Twenty	Doorstep interactions
`concerns`	Custom: `fieldworkNeed`	→ Twenty	Extracted needs/concerns
`resources`	Custom: `fieldworkOffer`	→ Twenty	Offers/resources from residents
`skills`	Custom: `fieldworkSkill`	→ Twenty	Skills from residents
`demographics`	Custom: `fieldworkDemographic` (restricted)	→ Twenty	Sortition layer, restricted access
`transcription_jobs`	DO NOT SYNC	—	Internal pipeline only
`volunteer_voice_submissions`	DO NOT SYNC	—	Internal source log only

3. Field-Level Mapping

3.1 `fieldwork.locations` → Twenty `fieldworkLocation`

Source Column	Target Field	Transform	Privacy Note
`id`	`fieldworkExternalId`	Direct (text)	Internal UUID, not PII
`address_normalised`	`address`	Direct	Cleaned address — safe
`postcode`	`postcode`	Direct
`ward`	`ward`	Direct
`area`	`area`	Direct
`created_at`	`sourceCreatedAt`	Direct
`updated_at`	`sourceUpdatedAt`	Direct	For change detection
`address_raw`	DO NOT SYNC	—	May contain PII
`geom`	`latitude`, `longitude`	Extract from PostGIS if needed	Optional

3.2 `fieldwork.households` → Twenty `fieldworkHousehold`

Source Column	Target Field	Transform	Privacy Note
`id`	`fieldworkExternalId`	Direct (text)
`name`	`name`	Direct
`location_id`	`locationId`	Lookup Twenty location ID by fieldworkExternalId
`last_contact`	`lastContactAt`	Direct
`contact_count`	`contactCount`	Direct
`created_at`	`sourceCreatedAt`	Direct
`updated_at`	`sourceUpdatedAt`	Direct
`notes`	`notes`	Sync only if reviewed flag present	May contain raw text

3.3 `fieldwork.people` → Twenty `Person` + Custom Fields

Source Column	Target Field	Transform	Privacy Note
`id`	Custom: `fieldworkExternalId`	Direct (text)
`name_given`	`name.firstName`	Direct
`name_family`	`name.lastName`	Direct
`email`	`email`	Direct	PII — ensure consent before sync
`phone`	`phone`	Direct	PII — ensure consent before sync
`household_id`	Custom: `householdId`	Lookup Twenty household ID
`role`	Custom: `fieldworkRole`	Normalise to picklist	Free-text needs mapping
`consent_status`	Custom: `consentStatus`	Direct	Critical for GDPR
`consent_updated_at`	Custom: `consentUpdatedAt`	Direct
`created_at`	`sourceCreatedAt`	Direct
`updated_at`	`sourceUpdatedAt`	Direct
`telegram_id`	DO NOT SYNC	—	External platform ID
`notes`	Custom: `fieldworkNotes`	Sync only if reviewed	May contain raw/sensitive text

GDPR Gate: Only sync people with consent_status IN ('granted') OR where there is a legitimate interest assessment documented. People with consent_status = 'refused' or 'withdrawn' must not be synced. People with consent_status = 'not_asked' should be flagged for review.

3.4 `fieldwork.organisations` → Twenty `Company`

Source Column	Target Field	Transform	Privacy Note
`id`	Custom: `fieldworkExternalId`	Direct (text)
`name`	`name`	Direct
`type`	Custom: `organisationType`	Normalise to picklist	Free-text needs mapping
`location_id`	Custom: `locationId`	Lookup Twenty location ID
`contact_person_id`	Custom: `contactPersonId`	Lookup Twenty person ID
`created_at`	`sourceCreatedAt`	Direct
`updated_at`	`sourceUpdatedAt`	Direct
`notes`	`notes`	Sync only if reviewed

3.5 `fieldwork.doorstep_contacts` → Twenty `fieldworkEngagement`

Source Column	Target Field	Transform	Privacy Note
`id`	`fieldworkExternalId`	Direct (text)
`household_id`	`householdId`	Lookup Twenty household ID
`person_id`	`personId`	Lookup Twenty person ID
`volunteer_id`	`volunteerId`	Lookup Twenty person ID
`outcome`	`outcome`	Direct (enum match)
`input_source`	`inputSource`	Direct
`llm_summary`	`summary`	Direct	Already cleaned by LLM
`llm_extracted`	`extractedData`	Direct (jsonb→json)	Structured extraction
`follow_up`	`requiresFollowUp`	Direct
`follow_up_notes`	`followUpNotes`	Direct
`captured_at`	`engagementDate`	Direct
`created_at`	`sourceCreatedAt`	Direct
`raw_transcript`	DO NOT SYNC	—	Raw STT — privacy risk
`job_id`	`sourceJobId`	Direct (reference only)	Audit trail

3.6 `fieldwork.concerns` → Twenty `fieldworkNeed`

Source Column	Target Field	Transform	Privacy Note
`id`	`fieldworkExternalId`	Direct (text)
`person_id`	`personId`	Lookup Twenty person ID
`concern`	`description`	Direct
`category`	`category`	Normalise to picklist	Free-text needs mapping
`confidence`	`confidence`	Direct	inferred/confirmed/reviewed
`source_job`	`sourceJobId`	Direct (reference)	Audit trail
`created_at`	`sourceCreatedAt`	Direct

Confidence Gate: Concerns with confidence = 'inferred' should be synced but flagged as "Unverified" in Twenty. Only confidence = 'reviewed' or 'confirmed' should display as active needs.

3.7 `fieldwork.resources` → Twenty `fieldworkOffer`

Source Column	Target Field	Transform	Privacy Note
`id`	`fieldworkExternalId`	Direct (text)
`person_id`	`personId`	Lookup Twenty person ID
`resource`	`description`	Direct
`confidence`	`confidence`	Direct	inferred/confirmed/reviewed
`source_job`	`sourceJobId`	Direct (reference)
`created_at`	`sourceCreatedAt`	Direct

3.8 `fieldwork.skills` → Twenty `fieldworkSkill`

Source Column	Target Field	Transform	Privacy Note
`id`	`fieldworkExternalId`	Direct (text)
`person_id`	`personId`	Lookup Twenty person ID
`skill`	`description`	Direct
`confidence`	`confidence`	Direct	inferred/confirmed/reviewed
`source_job`	`sourceJobId`	Direct (reference)
`created_at`	`sourceCreatedAt`	Direct

3.9 `fieldwork.demographics` → Twenty `fieldworkDemographic` (RESTRICTED)

Source Column	Target Field	Transform	Privacy Note
`id`	`fieldworkExternalId`	Direct (text)
`person_id`	`personId`	Lookup Twenty person ID
`age_band`	`ageBand`	Direct	Special category — restricted
`gender`	`gender`	Direct	Special category — restricted
`ethnicity`	`ethnicity`	Direct	Special category — restricted
`housing_tenure`	`housingTenure`	Direct
`employment_status`	`employmentStatus`	Direct
`has_dependants`	`hasDependants`	Direct
`disability`	`disability`	Direct	Health data — restricted
`area`	`area`	Direct
`created_at`	`sourceCreatedAt`	Direct
`updated_at`	`sourceUpdatedAt`	Direct

Access Control: This object must have role-based access control in Twenty. Only users with "Demographic Data" permission can view. Standard CRM users should not see these fields.

4. Idempotency

External ID Strategy

Every Twenty custom object must include a fieldworkExternalId field (type: text, unique).

fieldworkExternalId = fieldwork_table_name + ":" + id

Examples:

locations:550e8400-e29b-41d4-a716-446655440000
people:660e8400-e29b-41d4-a716-446655440001
doorstep_contacts:770e8400-e29b-41d4-a716-446655440002

Upsert Logic

1. Query Twenty: find record where fieldworkExternalId = <external_id>
2. If found:
   a. Compare sourceUpdatedAt with Twenty's sourceUpdatedAt
   b. If source is newer: UPDATE the Twenty record
   c. If same age: SKIP (no change)
3. If not found:
   a. CREATE new Twenty record with fieldworkExternalId

Change Detection

Use updated_at from Fieldwork as sourceUpdatedAt in Twenty.
Before updating, compare: only write if source.updated_at > target.sourceUpdatedAt.
This prevents unnecessary API calls and preserves any manual enrichments in Twenty (if those fields are not overwritten).

Sync State Table

Create fieldwork.sync_state in Postgres to track sync progress:

CREATE TABLE fieldwork.sync_state (
    table_name      text PRIMARY KEY,
    last_synced_at  timestamptz,
    last_row_count  integer,
    last_error      text,
    updated_at      timestamptz NOT NULL DEFAULT now()
);

5. Sync Order

Dependencies require this exact order:

locations          (no dependencies)
households         (depends on locations)
people             (depends on households)
organisations      (depends on locations, people)
doorstep_contacts  (depends on households, people)
concerns           (depends on people)
resources          (depends on people)
skills             (depends on people)
demographics       (depends on people, RESTRICTED)

Each step completes fully before the next begins. Within each step, process in order of updated_at ascending (oldest first).

Follow-Up Task Creation (Post-Sync)

After syncing doorstep_contacts, create Twenty Tasks for records where:

follow_up = true
No existing open Task linked to this engagement

6. Privacy Rules

Data That Stays Outside Twenty

Data	Location	Reason
Raw audio (`audio_data`)	CT604 only	Binary, never expose
Raw transcripts (`transcript`, `raw_transcript`)	CT604 only	May contain sensitive/PII
Telegram IDs (`telegram_id`, `telegram_user_id`, `chat_id`, `message_id`)	CT604 only	External platform identifiers
Raw addresses (`address_raw`)	CT604 only	May contain PII not in normalised form

Data That Syncs as Cleaned/Summary

Data	Treatment
`llm_summary`	Sync as-is (LLM-cleaned)
`llm_extracted`	Sync structured jsonb fields
`notes` fields	Sync only if reviewed/approved flag present
`address_normalised`	Sync instead of raw

People with consent_status = 'granted': Sync to Twenty.
People with consent_status = 'not_asked': Do not sync. Flag for consent collection.
People with consent_status = 'refused' or 'withdrawn': Do not sync. If previously synced, delete or anonymise the Twenty record.
Legitimate interest: If documented, may override for specific limited processing. Requires case-by-case assessment.

Demographic Data

Sync to a separate restricted-access object in Twenty.
Not visible in standard Person views.
Role-based access control required.
Primary use: sortition and aggregate analysis outside CRM.

Sortition Logic

Sortition (random selection for citizens' assemblies etc.) remains entirely outside Twenty.
Twenty stores only the demographic data needed for sortition eligibility filtering.
Actual random selection algorithms run against Fieldwork/Postgres directly.

7. First Implementation Plan

7.1 Python Sync Worker

Location: /root/lifehouse-twenty-sync/sync_worker.py

Dependencies:

psycopg2 (read from CT604 Postgres)
requests (write to Twenty API)
python-dotenv (configuration)

Configuration (.env):

FIELDWORK_DB_HOST=127.0.0.1
FIELDWORK_DB_PORT=5432
FIELDWORK_DB_NAME=lifehouse
FIELDWORK_DB_USER=postgres
FIELDWORK_DB_PASSWORD=<from CT604>

TWENTY_BASE_URL=http://192.168.100.220:3000
TWENTY_API_KEY=<twenty_api_key>

DRY_RUN=true
SYNC_DEMOGRAPHICS=false
LOG_LEVEL=INFO

7.2 Modes

Mode	Flag	Behaviour
Dry-run	`DRY_RUN=true`	Read Fieldwork, compute diffs, log what would change. No writes to Twenty.
Live sync	`DRY_RUN=false`	Read Fieldwork, compute diffs, write to Twenty.
Single table	`--table people`	Sync only one table (for testing).
Full sync	`--full`	Sync all tables in dependency order.
Since	`--since 2026-05-01`	Only sync records updated after date.

7.3 Sync Worker Pseudocode

def sync_table(table_name, twenty_object_type, field_mapping, dry_run):
    # 1. Read from Postgres
    rows = query(f"SELECT * FROM fieldwork.{table_name} WHERE updated_at > %s", (last_synced_at,))
    
    # 2. For each row
    for row in rows:
        external_id = f"{table_name}:{row['id']}"
        
        # 3. Check if exists in Twenty
        existing = twenty_find_by_external_id(twenty_object_type, external_id)
        
        if existing:
            if row['updated_at'] <= existing['sourceUpdatedAt']:
                continue  # No change
            if dry_run:
                log(f"[DRY-RUN] Would UPDATE {twenty_object_type} {external_id}")
            else:
                twenty_update(twenty_object_type, existing['id'], map_fields(row, field_mapping))
        else:
            if dry_run:
                log(f"[DRY-RUN] Would CREATE {twenty_object_type} {external_id}")
            else:
                twenty_create(twenty_object_type, map_fields(row, field_mapping))
    
    # 4. Update sync state
    update_sync_state(table_name)

def main():
    sync_order = ['locations', 'households', 'people', 'organisations', 
                  'doorstep_contacts', 'concerns', 'resources', 'skills']
    
    if not DRY_RUN:
        verify_twenty_connectivity()
        verify_twenty_auth()
    
    for table in sync_order:
        log(f"Syncing {table}...")
        sync_table(table, TABLE_TO_OBJECT[table], FIELD_MAPPINGS[table], DRY_RUN)
        log(f"Done: {table}")
    
    if not DRY_RUN:
        create_follow_up_tasks()
    
    log("Sync complete.")

7.4 Sync Logs

Log to /var/log/fieldwork-twenty-sync.log with format:

2026-05-11T19:00:00Z [INFO] Starting sync: locations
2026-05-11T19:00:01Z [INFO] locations: 0 creates, 0 updates, 0 skipped
2026-05-11T19:00:01Z [INFO] Starting sync: households
2026-05-11T19:00:02Z [INFO] households: 3 creates, 1 update, 0 skipped
...
2026-05-11T19:00:15Z [INFO] Sync complete. 15 creates, 3 updates, 42 skipped.

7.5 Optional Zulip Notification

After successful sync, post to Zulip stream #fieldwork-sync:

Fieldwork → Twenty sync complete
- Locations: 0 created, 0 updated
- Households: 3 created, 1 updated
- People: 5 created, 2 updated
- Engagements: 7 created, 0 updated
- Needs: 4 created, 0 updated
- Offers: 3 created, 0 updated
- Skills: 2 created, 0 updated
Total: 24 creates, 3 updates

8. Missing Fields / Schema Gaps

8.1 Fields to Consider Adding to Fieldwork/Postgres

Table	Suggested Field	Type	Purpose
`people`	`reviewed_at`	`timestamptz`	When notes were last reviewed for CRM sync
`people`	`reviewed_by`	`text`	Who reviewed
`households`	`reviewed_at`	`timestamptz`	Same as above
`organisations`	`reviewed_at`	`timestamptz`	Same as above
`concerns`	`reviewed_at`	`timestamptz`	When concern was human-reviewed
`concerns`	`reviewed_by`	`text`	Who reviewed
`concerns`	`status`	`text`	e.g. open, in_progress, resolved
`resources`	`reviewed_at`	`timestamptz`	Same as above
`skills`	`reviewed_at`	`timestamptz`	Same as above
`doorstep_contacts`	`reviewed_at`	`timestamptz`	When contact was reviewed
—	`consent_records` (new table)	—	Audit log of consent changes
—	`sync_state` (new table)	—	Track what's been synced to Twenty

8.2 Fields to Consider Adding to Twenty

Twenty Object	Suggested Field	Type	Purpose
All custom objects	`fieldworkExternalId`	`text` (unique)	Idempotency key
All custom objects	`sourceCreatedAt`	`datetime`	Origin timestamp
All custom objects	`sourceUpdatedAt`	`datetime`	Change detection
`fieldworkEngagement`	`sourceJobId`	`text`	Audit trail to transcription job
`fieldworkNeed`	`sourceJobId`	`text`	Audit trail
`fieldworkOffer`	`sourceJobId`	`text`	Audit trail
`fieldworkSkill`	`sourceJobId`	`text`	Audit trail

8.3 Normalisation Mappings Needed

Fieldwork Column	Current Type	Needed Mapping
`people.role`	free-text	→ Controlled picklist: resident, volunteer, staff, trustee, councillor, other
`concerns.category`	free-text	→ Controlled picklist: housing, health, finance, isolation, food, employment, education, digital, transport, other
`organisations.type`	free-text	→ Controlled picklist: charity, council, business, faith_group, community_group, school, health, other
`resources.resource`	free-text	→ May benefit from category tagging
`skills.skill`	free-text	→ May benefit from category tagging

9. Twenty Custom Objects Definition

9.1 `fieldworkLocation`

Fields:
- fieldworkExternalId  (text, unique)
- address              (text)
- postcode             (text)
- ward                 (text)
- area                 (text)
- latitude             (number, optional)
- longitude            (number, optional)
- sourceCreatedAt      (datetime)
- sourceUpdatedAt      (datetime)

9.2 `fieldworkHousehold`

Fields:
- fieldworkExternalId  (text, unique)
- name                 (text)
- locationId           (relation → fieldworkLocation)
- lastContactAt        (datetime)
- contactCount         (number)
- notes                (text)
- sourceCreatedAt      (datetime)
- sourceUpdatedAt      (datetime)

9.3 `fieldworkEngagement`

Fields:
- fieldworkExternalId  (text, unique)
- householdId          (relation → fieldworkHousehold)
- personId             (relation → Person)
- volunteerId          (relation → Person)
- outcome              (select: unknown, engaged, declined, not_home, follow_up)
- inputSource          (select: telegram_voice, web_upload, ocr_image)
- summary              (text)
- extractedData        (json)
- requiresFollowUp     (boolean)
- followUpNotes        (text)
- engagementDate       (datetime)
- sourceJobId          (text)
- sourceCreatedAt      (datetime)

9.4 `fieldworkNeed`

Fields:
- fieldworkExternalId  (text, unique)
- personId             (relation → Person)
- description          (text)
- category             (select)
- confidence           (select: inferred, confirmed, reviewed)
- sourceJobId          (text)
- sourceCreatedAt      (datetime)

9.5 `fieldworkOffer`

Fields:
- fieldworkExternalId  (text, unique)
- personId             (relation → Person)
- description          (text)
- confidence           (select: inferred, confirmed, reviewed)
- sourceJobId          (text)
- sourceCreatedAt      (datetime)

9.6 `fieldworkSkill`

Fields:
- fieldworkExternalId  (text, unique)
- personId             (relation → Person)
- description          (text)
- confidence           (select: inferred, confirmed, reviewed)
- sourceJobId          (text)
- sourceCreatedAt      (datetime)

9.7 `fieldworkDemographic` (RESTRICTED ACCESS)

Fields:
- fieldworkExternalId  (text, unique)
- personId             (relation → Person)
- ageBand              (text)
- gender               (text)
- ethnicity            (text)
- housingTenure        (text)
- employmentStatus     (text)
- hasDependants        (boolean)
- disability           (boolean)
- area                 (text)
- sourceCreatedAt      (datetime)
- sourceUpdatedAt      (datetime)

Access: Only users with "Demographic Data" permission.

9.8 Person Custom Fields (added to standard Twenty Person)

Custom Fields on Person:
- fieldworkExternalId  (text, unique)
- fieldworkRole        (select)
- householdId          (relation → fieldworkHousehold)
- consentStatus        (select: not_asked, granted, refused, withdrawn)
- consentUpdatedAt     (datetime)
- fieldworkNotes       (text)
- sourceCreatedAt      (datetime)
- sourceUpdatedAt      (datetime)

9.9 Company Custom Fields (added to standard Twenty Company)

Custom Fields on Company:
- fieldworkExternalId  (text, unique)
- organisationType     (select)
- locationId           (relation → fieldworkLocation)
- contactPersonId      (relation → Person)
- sourceCreatedAt      (datetime)
- sourceUpdatedAt      (datetime)

10. Blockers

Twenty custom objects must be created first — The sync worker cannot run until all custom objects and fields are defined in Twenty. This requires Twenty admin access.
Twenty API key with write permissions — The existing API key may be read-only. Verify it has permission to create/update custom objects and standard objects.
Consent status handling — Need a clear policy on what happens when consent is withdrawn for a person already synced to Twenty. Options: delete the Twenty record, anonymise it, or flag it as "consent withdrawn — do not contact".
Demographic data access control — Twenty must support role-based permissions on custom objects. If not, demographic data should not be synced until access control is available.
Normalisation picklists — The free-text fields (role, category, type) need agreed-upon controlled vocabularies before sync can populate select fields correctly.
Review workflow — Currently no explicit review step before data reaches Twenty. Consider adding a simple review flag or status field to key tables.

11. Next Steps

Immediate (Before Coding)

Create Twenty custom objects and fields as defined in Section 9.
Verify Twenty API key has write permissions.
Agree on normalisation picklists for role, category, type.
Decide consent withdrawal policy.

Implementation (Next Prompt)

Build the Fieldwork → Twenty sync worker at /root/lifehouse-twenty-sync/sync_worker.py

Requirements:
- Python 3
- Read from CT604 Postgres (lifehouse.fieldwork)
- Write to Twenty API at http://192.168.100.220:3000
- Dry-run mode first
- Follow the sync contract in FIELDWORK_TO_TWENTY_SYNC_CONTRACT.md
- Log to /var/log/fieldwork-twenty-sync.log
- Do not sync raw transcripts, audio, or telegram IDs
- Respect consent_status gate
- Use fieldworkExternalId for idempotency
- Sync in dependency order

Appendix A: Quick Reference — Sync Order & Dependencies

locations          ← no deps
households         ← locations
people             ← households
organisations      ← locations, people
doorstep_contacts  ← households, people
concerns           ← people
resources          ← people
skills             ← people
demographics       ← people (RESTRICTED)

Appendix B: Quick Reference — Privacy Gates

Gate	Rule
Consent	Only sync people with `consent_status = 'granted'`
Raw transcripts	Never sync
Audio data	Never sync
Telegram IDs	Never sync
Demographics	Restricted access object only
Notes	Sync only if reviewed
Addresses	Sync normalised, not raw

1. Source of Truth Principle​

2. Table-to-Object Mapping​

3. Field-Level Mapping​

3.1 fieldwork.locations → Twenty fieldworkLocation​

3.2 fieldwork.households → Twenty fieldworkHousehold​

3.3 fieldwork.people → Twenty Person + Custom Fields​

3.4 fieldwork.organisations → Twenty Company​

3.5 fieldwork.doorstep_contacts → Twenty fieldworkEngagement​

3.6 fieldwork.concerns → Twenty fieldworkNeed​

3.7 fieldwork.resources → Twenty fieldworkOffer​

3.8 fieldwork.skills → Twenty fieldworkSkill​

3.9 fieldwork.demographics → Twenty fieldworkDemographic (RESTRICTED)​

4. Idempotency​

External ID Strategy​

Upsert Logic​

Change Detection​

Sync State Table​

5. Sync Order​

Follow-Up Task Creation (Post-Sync)​

6. Privacy Rules​

Data That Stays Outside Twenty​

Data That Syncs as Cleaned/Summary​

Consent Gate​

Demographic Data​

Sortition Logic​

7. First Implementation Plan​

7.1 Python Sync Worker​

7.2 Modes​

7.3 Sync Worker Pseudocode​

7.4 Sync Logs​

7.5 Optional Zulip Notification​

8. Missing Fields / Schema Gaps​

8.1 Fields to Consider Adding to Fieldwork/Postgres​

8.2 Fields to Consider Adding to Twenty​

8.3 Normalisation Mappings Needed​

9. Twenty Custom Objects Definition​

9.1 fieldworkLocation​

9.2 fieldworkHousehold​

9.3 fieldworkEngagement​

9.4 fieldworkNeed​

9.5 fieldworkOffer​

9.6 fieldworkSkill​

9.7 fieldworkDemographic (RESTRICTED ACCESS)​

9.8 Person Custom Fields (added to standard Twenty Person)​

9.9 Company Custom Fields (added to standard Twenty Company)​

10. Blockers​

11. Next Steps​

Immediate (Before Coding)​

Implementation (Next Prompt)​

Appendix A: Quick Reference — Sync Order & Dependencies​

Appendix B: Quick Reference — Privacy Gates​