Plan and improve your Atlan rollout
This guide serves two situations: planning a new rollout (empty or early-stage tenant → adopted implementation) and improving a live one (already live, asking "what's the highest-leverage next move?"). Use the tab that matches where you are.
The fastest route to value: connectors live → enrichment as soon as base metadata is stable → light access → rollout. Don't gate enrichment on a finished governance model; governance emerges from real steward demand. Start the long-lead items (security review, SSO/identity, AI-governance approval) at planning time—they take weeks to months and silently block your go-live date.
- New rollout
- Improve live implementation
The rollout follows five phases. Migration, when present, runs as a parallel track—not a sixth sequential step.
| Phase | Goal | Go deeper in |
|---|---|---|
| 0. Rollout planning | Confirm scope, cohort, readiness, and start long-lead items | This page—Phase 0 below |
| 1. Connectors & integrations | Priority sources crawled, lineage validated, technical users live | Connector & ingestion best practices |
| 2. Glossary & semantic structure | Seed terms; decide metrics catalog vs. business glossary | Glossary & taxonomy best practices |
| 3. Metadata enrichment | Automate enrichment; curate what matters most first | Metadata enrichment best practices |
| 4. Access control & governance | Personas, policies, approval flows—once there are SMEs to own them | Governance, access & personas best practices |
| 5. End-user rollout | Adoption across the scoped cohort with forcing functions | See the Live implementation tab for the diagnostic |
Migration runs in parallel with Phases 1–4 from day one—not after rollout. Challenge the scope first: re-crawl technical assets and lineage rather than migrating them; migrate only the glossary and enrichment that carry real value. See Migration best practices.
Dependency rules
- Connectors live before enrichment (can't enrich assets that aren't ingested) and before lineage.
- Base metadata ingested before enrichment; AI-governance approval can start in parallel.
- Glossary can start in parallel with connectors; linking terms to assets needs both.
- Access and governance come after there's something enriched worth governing—run the access workflow end-to-end at small scale before granting broad edit permissions.
- Broad end-user rollout comes after setup, curation, and access are in place for the cohort.
Use-case variations
Detect your primary use case and adjust—never apply one template blindly.
AI / context-first
Connectors → enrichment early and hard → light access → feed context to AI. Start AI-governance approval immediately. Glossary and heavy governance come later.
Governance / compliance-first
Connectors → sensitive-data classification and personas/policies earlier than usual → enrichment → rollout. Compliance gates are real prerequisites; expect timelines +50–100% longer than a discovery-first rollout.
Discovery / self-service-first
Connectors → enrich the most-searched assets → glossary for shared vocabulary → rollout. Access stays light. Adoption forcing-functions matter most.
Migration-driven
Migration runs as a parallel track from day one. Re-crawl technical assets and lineage rather than migrate them; bring over glossary and enrichment that carry value.
Company size and industry shift this too: regulated industries push governance earlier and add time; smaller teams compress phases and run a lean persona model; large multi-domain estates run independent phase tracks per domain with shared governance in parallel.
Phase 0: Rollout plan
This is where implementations succeed or stall. Frame planning as matching the demand for context (what your scoped users need to find and trust) against the supply of context (what metadata exists today). The gap between them is your implementation plan—the phases raise supply to meet demand for this cohort. Pin down six things:
- Outcome & success definition: the business problem, 30-60-90-day milestones, the executive sponsor, and what "done" looks like.
- Who it's for: the first user cohort and roughly how many; name the second domain so the path can extend.
- Pilot domain & scope: one domain first, not the whole estate, with what's explicitly out of scope.
- Technical readiness: stack documented, priority connector chosen, credentials plus a committed crawl date, SSO/identity status.
- Context readiness: glossary seed set confirmed, who owns steward review, what enrichment scope automation covers, AI/security approval started if AI is in play.
- People & commitments: your internal project lead, named owners, a recurring cadence, and a blockers log.
Readiness checks—confirm these before committing:
- Is the priority source connectable now (credentials, network, security sign-off)?
- Is there a glossary seed set, or does it need importing?
- Is there a named steward to review AI-generated enrichment?
- Are the long-lead items (security review, SSO/identity, AI-governance approval) already started?
- Is the cohort small enough to validate end-to-end before a broad rollout?
Phases 1–5
Each phase has a clear entry condition, the work to do, and an exit signal. Phases can overlap per domain—governance and access often run in parallel once foundations exist. Always confirm the entry condition against your actual state before starting a phase.
Phase 1: Connectors & integrations
- Entry: Priority source identified, credentials and a committed crawl date secured, security review initiated. Start the longest-lead items now.
- Work: Connect priority sources, validate the crawl and lineage, provision technical users.
- Exit: Priority connectors live and crawled, lineage validated on sample assets, technical users can log in and find assets.
Phase 2: Glossary & semantic structure
- Entry: Can run in parallel with Phase 1.
- Work: Stand up glossaries and categories; bring existing business terms in bulk; decide metrics catalog vs. business glossary.
- Exit: A seed glossary exists, the structure is agreed, and a steward owns it.
Phase 3: Metadata enrichment
- Entry: Connectors live and base metadata ingested; AI-governance approval initiated as a parallel workstream.
- Work: Automate enrichment as much as possible (using AI-assisted enrichment where available); enrich the most-used and most-valuable assets first, not everything.
- Exit: In-scope assets meet a quality bar and stewards have reviewed the AI output.
Phase 4: Access control & governance
- Entry: Enrichment underway; you have SMEs and an operating model to own approvals.
- Work: Set up personas and policies; add governance workflows only where there's an owner to run them. Most implementations start with a small persona set—don't over-engineer.
- Exit: The scoped cohort sees the right assets with the right permissions, and any required approval flow has been tested end-to-end with a real user before broad rollout.
Phase 5: End-user rollout
- Entry: Setup, curation, and access are in place for the cohort.
- Work: Roll out to the scoped teams; drive adoption with forcing functions (mandates or integrations that mean users must use Atlan), not just training.
- Exit: The target cohort is active and finding/trusting data, and your Phase 0 success metrics are moving. Then expand to the next domain.
When you're already live and asking "how do you get more value / where are you stuck?"—or "the setup is done, what next?"—the answer is always the same: find the single highest-leverage gap and fix it. Not a twelve-item backlog.
- Diagnose from current state. Review what you actually have across connectors, glossary, enrichment, access/governance, lineage, data products, and adoption—use the completeness and adoption dashboards in your Atlan tenant rather than auditing manually.
- Find the earliest gap, not the loudest symptom. Weak AI answers or low adoption are usually downstream of a missing foundation (thin enrichment, no connectors, unlinked glossary). The highest-leverage move is the earliest gap on the dependency chain (connectors → enrichment/glossary → access → rollout). Thin enrichment is a heavier penalty for an AI-first rollout than a discovery-first one—weight gaps against your use-case emphasis.
- Name one top priority with sequenced next steps, an owner, and a by-when.
- Fix it, then re-run: one priority at a time.
Common stall points → fix
| Stall | Likely root cause | Fix / where to go deeper |
|---|---|---|
| "Setup is done but no one uses it" | Enrichment thin → assets aren't trustworthy or findable | Enrich the most-used assets first, then apply an adoption forcing-function (a mandate, integration requirement, or deadline that means users must use Atlan rather than the old tool): metadata enrichment best practices |
| "AI answers are poor" | Thin context on the assets agents hit | Enrich and link terms on the in-scope assets: metadata enrichment best practices + glossary & taxonomy best practices |
| "Lineage looks empty" | Lineage miner not run or misconfigured (separate from the crawler), or sources not yet cataloged | Check whether the lineage miner ran: lineage best practices |
| "Users see too much / the wrong things" | No persona scoping | Start with two or three personas: governance, access & personas best practices |
| "Glossary exists but feels useless" | Terms not linked to assets | Link terms to assets: glossary & taxonomy best practices |
| "Stalled after the pilot" | No second domain, no forcing function | Re-run rollout planning for the next domain—this page |
| "Governance is heavy / stuck" | Over-engineered before there was demand | Simplify; govern only what's enriched: governance, access & personas best practices |
Common pitfalls
- Sequencing from a template instead of current state. Always check what's already configured and find the next gap, not the whole plan from scratch.
- Gating enrichment on a finished governance model. Governance emerges from steward demand—get value flowing first.
- Leaving long-lead items until go-live. Security review, SSO/identity, and AI-governance approval take weeks to months; start them in Phase 0.
- Boiling the ocean. One pilot domain first; name the second, but don't try to onboard the whole estate at once.
- Treating a stalled rollout as a training problem. A stalled rollout usually needs a forcing function and a curated foundation, not another workshop.
- Chasing the loudest symptom. Fixing a late-stage symptom rarely helps when an upstream foundation is missing—fix the earliest gap.
Troubleshooting
Process and meta symptoms—"what do I do" and "why is a phase stuck." For capability-area symptoms (setup unused, weak AI answers, empty lineage, useless glossary), see Common stall points → fix in the Live implementation tab.
| Symptom | Likely cause | Next move |
|---|---|---|
| "The team doesn't know what to do next" | No grounding in current state | Check current state against the five phases; name the first gap, name that phase, take the first two or three concrete steps. |
| "A phase feels blocked" | An unmet entry condition or an unstarted long-lead dependency | Check the phase's entry condition and whether security/SSO/AI-governance approval was started early enough. |
| "The rollout keeps slipping" | No named single owner or milestone | Assign one owner and one concrete next milestone; keep the blockers log visible and reviewed in every working session. |
| "Stuck deciding how to sequence two connectors" | Both feel equally urgent | Pick the one that unblocks the most enrichment and the most users; do the other second. |
| "Enrichment isn't running" | Missing lineage, missing steward, or AI-governance approval not in place | Confirm connectors are live and lineage is validated; confirm AI-governance approval status; assign a steward before re-enabling enrichment. |
To assess completeness or adoption per area rather than auditing manually, check the completeness and adoption dashboards in your Atlan tenant—the available reporting depends on your release.
See also
- Connectors & ingestion: connector choice, crawl scoping, and lineage validation.
- Glossary & taxonomy: glossary and taxonomy structure and metric definitions.
- Metadata enrichment: automating and curating metadata enrichment at scale.
- Governance, access & personas: personas, policies, and access design.
- Lineage: lineage generation, coverage, and troubleshooting.
- Migration: migrating off a legacy catalog in parallel with your rollout.