SK Group's AI Push Starts With Better Operations, Not Bigger Models
SK Group Chair Chey Tae-won closed the company's annual CEO Seminar in Icheon with a clear directive: fix the fundamentals, then scale AI. Top leaders from across the tech-to-energy group met to align on core competitiveness following a major year-end leadership reshuffle announced on Oct. 30.
Chey's message to executives was direct: operational rigor and domain expertise come first. AI sits on top of that foundation - not the other way around.
The Core Message
"Operational improvement may sound complicated, but it's about getting the fundamentals right," Chey said. He urged leaders to audit internal processes over the past five to ten years to prevent repeating mistakes. "If a company pushes for AI transition without a fundamental foundation, it is bound to fail."
He doubled down on expertise: "Implementing AI without domain knowledge will not solve the problems. We can only gain the upper hand in the AI race when we have solid domain knowledge."
Chey also called for SK to move beyond selling high-performance memory chips and build full customer solutions on AI infrastructure.
Risk, Safety, and Compliance Stay Non-Negotiable
Executives reviewed plans to strengthen safety, health, environment, cybersecurity, and compliance management. For operations teams, that means controls built into daily work, not bolted on after a pilot. If you need a reference model for security, see the NIST Cybersecurity Framework.
What This Means for Operations Leaders
- Map critical value streams. Identify bottlenecks, rework, and wait states. Fix the top three failure points before introducing AI.
- Standardize and assign ownership. Define SOPs, control owners, and review cadence. No process, no model.
- Get your data house in order. Establish sources of truth, quality thresholds, and event logging. Bad data makes smart teams look slow.
- Pick business-first use cases. Start with 2-3 problems tied to P&L (yield optimization, maintenance scheduling, demand forecasting). Assign a single accountable owner.
- Capture domain knowledge. Formalize SME councils, convert tacit know-how into playbooks, and keep a living wiki. Models learn faster when your org already did.
- Build an enablement stack. Process mining for discovery, MLOps for deployment, and observability for drift and incident response.
- Design-in governance. Security, privacy, and compliance checks sit in the workflow, not in a separate queue.
- Think in solutions. Bundle hardware, software, data services, and support into outcomes customers can buy.
Quick Gate Checklist Before Funding Any AI Project
- Baseline metrics exist (cycle time, defect rate, unit cost, service level).
- Data availability and quality are verified, with clear owners.
- Change plan covers training, SOP updates, and comms to frontline teams.
- Success metrics and a 90-day review are written down - and signed.
- Security, privacy, and compliance reviews are complete.
- Integration and rollback plans are defined for current systems.
Upskilling the Ops Org
If your team needs structured paths to apply AI to process improvement, see AI upskilling paths for operations teams.
Bottom line: Chey's stance is practical. Fix processes, strengthen domain knowledge, then layer in AI to compound what already works. That's how you get durable results, not demos.
Your membership also unlocks:
