Accept, reject, sync-back

ACCEPT     Apply the tag in ZopNight; sync to cloud if configured
REJECT     Discard suggestion; don't show again until manual re-trigger
SNOOZE     Re-evaluate the prediction later (defer decision)

EFFICIENT BULK ACCEPT:

  Filter: confidence ≥ 90% AND no existing env tag
  Select all matching → Accept all

  This pattern processes 50-200 high-confidence predictions in seconds
  Most teams' fastest path to coverage

WHEN TO BULK-ACCEPT:
  Initial onboarding (large untagged backlog)
  Periodic discovery sweeps (new resources predicted)
  Post-reorg cleanup

WITHOUT SYNC-BACK:
  ZopNight: tag environment=dev applied locally
  Cloud-side: untagged still
  Reports in ZopNight: see the tag
  Cloud-native tools: don't see it

WITH SYNC-BACK:
  ZopNight: tag applied locally
  Cloud API call: ec2:CreateTags
    Resources=[i-0abc123]
    Tags=[{Key=environment, Value=dev}]
  Cloud-side: now tagged
  Both ZopNight and cloud-native tools see the tag

SYNC ON                            SCENARIO
─────────────────────────────────────────────────────────────
Tag-based IAM policies              Customer's IAM uses tag conditions
                                    Tag-driven access control

Cloud billing exports                Customer relies on cloud tags for billing
                                    AWS Cost Explorer, Azure Cost Management

Tag-based budget enforcement         Customer uses cloud-native budgets
                                    Per-tag budget thresholds

Compliance / audit reporting         Customer's audit trail expects cloud tags
                                    Auditor compares to cloud-side state

Multi-tool ecosystem                 Other FinOps/governance tools
                                    Expect cloud tags

SYNC OFF                          SCENARIO
─────────────────────────────────────────────────────────────
ZopNight-only attribution           Tags only used in ZopNight reports
                                    Cloud-side tagging managed elsewhere

Read-only credentials               No write permission to tag
                                    Sync-back fails

Risk-averse                          Want explicit control via cloud console
                                    Manual sync workflow preferred

Cloud IAM constraints                ZopNight credentials don't allow tagging
                                    Permissions limit

THE CLOUD now has the tag (indistinguishable from manually-applied)
ZopNight records that the tag came from auto-tagger:
  dimension_source = auto

IN Reports → Tags:
  Filter by dimension_source = auto
  See auto-tagged contributions separately from cloud-native tags

THIS MATTERS FOR:
  Tag-coverage audits ("how much of our coverage comes from auto-tag?")
  Reconciliation against cloud-side tag standards
  Understanding the auto-tagger's effectiveness
  Compliance review (where did this tag originate?)

1. Customer clicks Reject on a prediction
2. ZopNight records: prediction_status = rejected_by_user_X
3. Prediction hides from the active queue
4. Resource remains untagged (or with cloud-native tags only)
5. Next discovery sync: prediction is NOT re-suggested
   (unless user clicks "Re-evaluate" on the resource)

REJECT IS DURABLE
  The prediction won't haunt the queue forever
  Won't re-appear and waste reviewer time
  Customer's decision is respected

TO RE-EVALUATE:
  Explicit action by user
  Manual trigger; not automatic

1. Open the resource → Actions → Re-evaluate prediction
2. Auto-tagger re-runs the inference with current state
3. New prediction (which may differ from previous)
4. Active in the queue for action

USE CASES for re-evaluation:
  Resource changed environment (dev → staging)
  Resource was renamed
  Resource was moved to different group
  Tags were manually adjusted by another tool
  After major reorganization

SNOOZE = "decide later"
  Tag suggestion still in the queue
  But: removed from default view
  Reappears in 30 days (default) or custom timing

WHEN TO SNOOZE:
  Uncertain right now; want to check usage first
  Resource changing state (mid-migration)
  Waiting for team owner to confirm
  Not in this triage cycle's scope

SOME ORGS PREFER cloud tags as source of truth:

PATTERN:
  ZopNight predicts; suggests
  Customer reviews; accepts
  Sync-back applies tag to cloud
  Cloud is now the canonical tag store

  All tools (ZopNight + others) read from cloud
  Single source of truth
  No divergence

ALTERNATIVE:
  ZopNight as source of truth
  Cloud tags secondary

  Works if only ZopNight needs the tags
  Most orgs prefer cloud-as-truth

EFFICIENT BULK ACCEPT (when right):
  All high-confidence predictions
  All in one environment (e.g., all env=prod predictions)
  Same type pattern

WHEN BULK IS WRONG:
  Mixed confidence levels (some moderate)
  Mixed environments (would mass-mistag)
  After uncertain inference period

NEVER BULK-ACCEPT below 85% confidence
  Risk of mass-mistagging
  Each reject is more friction than each careful accept

T+0      Open Insights → Auto-Tagging
         245 predictions across all resources

T+5min   Filter: confidence ≥ 95%
         84 high-confidence predictions
         Mostly env=prod tags on production-named resources

T+6min   Select all 84 → Bulk Accept → Sync-back enabled
T+30s    Tags applied to ZopNight database
T+30s    84 cloud-side ec2:CreateTags calls
         All succeeded
T+30s    Tag-coverage jumps: 62% → 76%

T+15min  Filter: confidence 85-94%
         78 moderate-confidence predictions
         Open each; review the inference panel briefly

         Decision per prediction:
           Match operator's knowledge: 65 → Accept
           Doesn't match (resource is different env): 13 → Reject

T+30min  Tag-coverage now at 87%

T+1 MONTH  Discovery picks up new resources
           New predictions surface naturally
           Operator runs through them weekly
           Coverage climbs to 92-94%

T+90 DAYS  Stable at ~93% tag coverage
           Quarterly review cleans the long tail

□ STEP 1: Open Insights → Auto-Tagging
  Total predictions: _____
  Filter: confidence ≥ 95%

□ STEP 2: Bulk Accept high-confidence
  Select all → Accept (with sync-back if appropriate)
  Verify tags applied in ZopNight: __________
  Verify tags applied in cloud console (if sync-back): __________

□ STEP 3: Review moderate (85-94%)
  Pick 5 to review
  For each:
    Inference makes sense? □ Yes □ No
    Accept / Reject decision: __________

□ STEP 4: Calculate impact
  Tag coverage before: _____%
  Tag coverage after: _____%
  Resources tagged in this session: _____

□ STEP 5: Schedule next session
  Date: __________
  Goal: process moderate confidence remaining