# Databricks readiness assessment > Ten questions on governance, pipelines, delivery, cost, operations and AI readiness, scored in the browser, with what to fix first. Canonical: https://www.techfabric.com/assessments/databricks-readiness --- Self-assessment ## Databricks readiness assessment A Databricks platform is ready for more teams, and for AI, when governance, delivery, cost and operations each have a standard and an owner. Answer ten questions about how yours runs today, and you get a score, the areas that will slow you down first, and what to read or do about each. 10 questions · about three minutes · your answers stay in this browser tab #### Where do your tables live? - We are not on Databricks yet - In the legacy Hive metastore - Some workspaces on Unity Catalog, some still on Hive - Every workspace on Unity Catalog #### How is access to data granted? - Everyone who needs data is a workspace admin - Grants to individual users, by hand - Grants to groups, but groups are managed inside Databricks - Grants to groups synced from our identity provider #### How does data get into the lakehouse? - Mostly manual uploads and one-off scripts - Scheduled notebooks, each written differently - Jobs with shared code, but no data quality checks - Declarative pipelines with expectations on each table #### How are your tables organised? - No agreed structure - Raw and reporting tables side by side - Bronze, silver and gold layers, loosely followed - Bronze, silver and gold, with documented gold tables that BI reads #### How do changes reach production? - Notebooks are edited directly in production - Code is in Git but deployed by hand - Git and CI, but one shared workspace - Git, CI and separate dev, staging and production targets #### Who can tell you what last month's Databricks spend was for? - Nobody, we see the invoice - Finance, but only the total - Engineering, from the billing system tables - Each team, from tagged usage in the system tables, with budgets and alerts #### What runs your scheduled work? - All-purpose clusters left running - All-purpose clusters with auto-termination - Job compute, sized by hand - Serverless or job compute under compute policies #### When a pipeline fails at 2am, what happens? - We find out when someone complains - Someone checks the job runs in the morning - Failures alert a shared channel - Failures and late data alert a named owner, and freshness is monitored #### How do people use the data? - Exports to spreadsheets - BI tools reading copied extracts - BI tools reading SQL warehouses directly - SQL warehouses for BI, and Genie spaces for questions in plain English #### Where do machine learning models and agents stand? - None yet - Experiments in notebooks, nothing tracked - Experiments tracked in MLflow, models deployed by hand - Models and agents registered in Unity Catalog, evaluated and served 0 of 10 answered The results ### What each result means 0% to 45% #### Foundations first The platform is not ready for more workloads yet. Governance, delivery and cost each need an owner and a standard before the next project lands on it, or every new pipeline makes the clean-up bigger. 45% to 70% #### Ready to scale, with gaps The core works, and a few areas will slow you down as more teams arrive. Close the weakest areas below first; they are the ones that turn into incidents when usage doubles. 70% to 90% #### Ready for AI workloads Governance, delivery and operations are in place, which is what agents and models need underneath them. The next gains are in Genie, model serving and agents on governed data. 90% and above #### Ready to optimise This is a mature platform. What is left is cost per workload, recovery testing and keeping pace with what Databricks ships each month. FAQ ### About this assessment It measures ten areas that decide whether a Databricks platform can take on more work: Unity Catalog adoption, how access is granted, ingestion, table layers, deployment, cost visibility, compute, operations, how people use the data, and machine learning. Each answer scores from zero to three. No. The scoring runs in your browser and nothing is stored or sent. Close the tab and the answers are gone. Start with the weakest areas the result lists, in that order. Governance and delivery come first because every later workload builds on them. A Databricks Health Check reviews the same areas against your actual workspaces. It means the foundations agents need are in place: governed data in Unity Catalog, deployments through review, and owners for failures. Agents then need evaluation and serving on top, which the Field Guide chapters on MLflow and agents cover.