AI Only Replaces Bad Managers, Says Quant Leader Ernest Chan

AI Only Replaces Bad Managers: What Ernest P. Chan Wants Leaders to Hear

Ernest P. Chan-founder and chief scientist of PredictNow.ai and non-executive chairman of QTS Capital Management-is a leading voice in applying AI to investing. His message to managers is simple: you can't outsource your thinking to AI.

AI exposes weak judgment. It rewards clear thinking, strong process, and disciplined decision-making. If your management habit is to guess, delegate the hard part, and hope the tool is "smart," AI will highlight the gaps.

What AI Actually Replaces

• Managers who treat AI as a magic 8-ball instead of a decision partner.
• Teams that skip hypotheses, baselines, and metrics-and jump straight to prompts.
• Leaders who ignore model limits, data quality, and risk-then blame the tool.

What Great Managers Do Instead

• Define the decision boundary: what AI informs, what humans own.
• Start with a hypothesis and a baseline strategy before touching a model.
• Treat data as evidence, not decoration. Track confidence, not just outputs.
• Use AI for scenario testing, risk checks, and speed-not final judgment.
• Keep a clear audit trail for how conclusions were reached.

A Simple Operating System for AI-Backed Decisions

• Frame: State the decision, objective, constraints, and acceptable risk.
• Baseline: Write the "human-only" plan first. This gives you a benchmark.
• Model: Select the tool that fits the job (classification, forecasting, retrieval, generation).
• Scenarios: Run multiple paths. Compare outcomes vs. baseline and best/worst cases.
• Stress test: Change key assumptions and re-run. Note where results break.
• Decide: Make the call. Record rationale, uncertainty, and next review date.
• Feedback loop: Log results, update assumptions, and adjust the process.

Guardrails That Keep You Out of Trouble

• Model risk management: document purpose, limits, data lineage, and monitoring. See the NIST AI Risk Management Framework for a clear structure here.
• Human-in-the-loop: humans verify and approve consequential outputs with defined checkpoints.
• Data discipline: remove leakage, bias, and stale inputs. Version your datasets and prompts.
• Clear KPIs: measure decision quality (hit rate, drawdown, cycle time), not tool usage.

Signals Your Team Is Using AI Well

• Faster decisions with equal or better outcomes-and a written rationale.
• Lower variance in results because risk is modeled, not guessed.
• Fewer "surprises" because scenarios were explored up front.
• Ongoing model monitoring with triggers for retraining or rollback.

Practical Moves You Can Make This Quarter

• Run one high-value decision through the operating system above. Compare against last quarter's approach.
• Create a one-page model card template for every AI tool your team touches.
• Set a weekly "assumption review" meeting. Update models when reality shifts.
• Add a decision post-mortem log. Track what you believed, what happened, and what changes.

Why This Matters

AI won't save weak leadership. It amplifies your thinking-good or bad. The managers who win treat AI like a disciplined partner: clear roles, clear checks, and clear accountability.

If you need structured, practical upskilling for managers and operators, explore focused learning paths such as AI Learning Path for Training & Development Managers or the AI Learning Path for Project Managers. Executives and IT leaders may also find the AI Learning Path for CIOs useful for governance and strategy.

About Ernest P. Chan

Ernest P. Chan is founder and chief scientist of PredictNow.ai and non-executive chairman of QTS Capital Management. He is widely recognized for applying machine learning and quantitative methods to investments.