AI can make a good workflow faster. It can also make a weak data foundation more dangerous. If the information behind an AI workflow is outdated, duplicated, incomplete or poorly governed, the automation may simply produce mistakes at higher speed.
Data foundations are not a luxury before AI. They are part of making AI useful.
AI needs reliable context
An AI assistant can summarise, classify, draft or recommend based on the information it receives. If that information is wrong, the output can be wrong in a convincing way. If access boundaries are unclear, the output can expose information to the wrong people. If data is scattered across tools, the AI may rely on incomplete context.
Twoday’s AI-ready data article is relevant because it frames data readiness as business-critical. AI value depends on the quality, availability and governance of the data behind it.
Common data problems
Companies often have useful data, but not in a form that supports automation:
- customer records differ across systems;
- spreadsheets contain manual corrections;
- documents are stored without consistent metadata;
- product data is incomplete;
- reporting depends on exports;
- the same field has different meanings in different systems;
- permissions are based on informal practice;
- historical data is not cleaned or labelled.
These problems do not prevent all AI use. But they decide which use cases are safe, which require preparation and which should not be automated yet.
Product data as an example
IMPACT Commerce’s AI-powered PIM article is a useful example because product information is a clear data-quality problem. AI can help enrich, localise and manage product data, but the workflow depends on a structured foundation: attributes, categories, source ownership, approval and validation.
The same pattern appears in other domains: customer cases, legal documents, support tickets, operational logs, financial records and internal knowledge bases.
Data foundations include access control
Data quality is not only accuracy. It also includes permission, provenance and accountability.
A readiness review should ask:
- Who owns the data?
- Which system is authoritative?
- Who can access it?
- Can permissions be enforced technically?
- Can the AI retrieve only approved sources?
- Can outputs cite or reference source material?
- Can changes be reviewed and audited?
This is where data, security and workflow design meet.
Build the minimum foundation needed
A company does not need a large data platform before every AI project. It needs enough structure for the specific workflow. For a knowledge assistant, that may mean approved document sources, metadata, access rights and retrieval testing. For document processing, it may mean input templates, validation rules and exception handling. For reporting automation, it may mean stable data pipelines and agreed definitions.
The practical question is: what foundation does this use case require to be useful and safe?
Warning signs
AI is likely premature when:
- no one knows which data source is authoritative;
- users disagree on what fields mean;
- sensitive documents are mixed with public material;
- data access is not role-based;
- reports are manually reconciled every time;
- outputs cannot be checked against sources;
- there is no owner for data quality.
These issues should become roadmap items before production automation.
Memory(One) perspective
For Memory(One), data foundations sit between Integration & Platforms and AI, Data & Automation. Useful AI requires connected systems, clear data flows, access control and human review. The right starting point is often not “build an AI tool.” It is “prepare the workflow and data foundation that will make the AI tool worth building.”
Sources and inspiration
- Twoday — AI-ready data becomes business critical: https://www.twoday.com/blog/ai-ready-data-becomes-business-critical
- IMPACT Commerce — AI-powered PIM benefits: https://impactcommerce.com/insights/12-advanced-benefits-of-ai-powered-pim/
- NNIT — Importance of systems integration: https://www.nnit.com/insights/articles/importance-systems-integration
- NoA Ignite — Getting your customer data platform off to the right start: https://noaignite.com/insights/getting-you-customer-data-platform-of-to-the-right-start/