IBM Korea GM: Data strategy is the bottleneck to AI outcomes
At a press conference in Seoul, IBM Korea General Manager Lee Soo-jung made the point most executives already feel: the lack of a company-wide data strategy is what keeps AI stuck in pilots. "AI has already become a symbol of corporate innovation, but most companies still remain at the experimental stage," she said. The fix isn't another model. It's enterprise data that's accessible, consistent, and governed end to end.
Sharing findings from a recent survey of C-suite executives, Lee said CEOs see enterprise data as the key to capturing generative AI value-yet only 1 percent of enterprise-specific data is used in actual AI products. And only 13 percent of chief data officers said their companies can generate revenue from AI. The biggest blockers: complex data preparation that steals time from model work and limited real-time access caused by inconsistent formats.
What executives should act on now
Lee's message was blunt: make data strategy a core management priority, not a technical side project. "To drive tangible results from AI transformation, companies must establish clear strategies for securing and leveraging AI-ready data and effectively translate more than 90 percent of unstructured data into measurable business outcomes," she said.
- Set ownership at the top: appoint a single accountable leader (often the CDAO) with budget and authority across business units.
- Tie data to money: pick 3-5 profit-and-loss use cases, define KPIs, and back-solve the data needed to move those numbers.
- Standardize the pipes: enforce common schemas, metadata, and data contracts so teams can build once and reuse often.
- Turn unstructured content into fuel: instrument documents, tickets, emails, and recordings into searchable, policy-compliant stores that AI can read.
- Go real time where it matters: expose key data via streaming and APIs; cut latency and batch dependencies that stall AI agents.
- Measure data quality like uptime: track freshness, lineage, completeness, and access SLA-then publish the scorecards.
- Lock down risk: apply consent, retention, and access controls by default; automate audit trails.
Company-wide, not silo-deep
Lee urged leaders to roll out the data strategy across the enterprise with clear performance indicators and tight alignment between business and technology chiefs. AI will underperform if data is fragmented or hidden across systems. AI agents need high-quality data and flexible, real-time infrastructure to do meaningful work.
Governance that earns trust
"Poor data quality or compliance gaps can undermine AI reliability," Lee warned. Put governance at the center-from data sourcing to model outputs. Align policies to recognized standards and frameworks, and make compliance visible to boards and regulators.
- Define approved data sources and usage policies for each AI use case.
- Segment sensitive data; apply role-based access and encryption everywhere.
- Review vendors and third-party data for licensing and security issues.
- Track model input/output lineage for accountability and audit.
Metrics that matter
- Percent of enterprise data that is AI-readable, governed, and live.
- Time-to-data (from request to production-grade access).
- Reuse rate of certified data products across use cases.
- Pilot-to-production conversion rate and revenue impact per use case.
The takeaway for executives: fund the data foundation first. Models are replaceable. Data pipelines, governance, and real-time access are the compounding advantage.
For governance guidance, see the NIST AI Risk Management Framework here. If you're aligning teams and skills to your data strategy, explore role-based AI training options here.
Your membership also unlocks:
