Why strategy keeps failing (and why AI won't save you)
On stage at SXSW Sydney, behavioural scientist Ganna Pogrebna and strategist Graham Kenny landed on the same truth. Strategy isn't failing because leaders lack brains or dashboards. It's failing because organisations still misread how decisions actually get made. If you lead, this is your blind spot.
The execution gap you can't fix with another plan
Kenny's point was blunt: strategy lives at the organisational level; execution lives with individuals. When strategy turns into a static plan, it loses punch. Positioning is a company choice. Action is a person's choice.
- Translate strategy into 3-5 role-specific actions per team. Names, owners, deadlines.
- Replace annual plans with quarterly "positioning statements" and weekly execution rhythms.
- Review progress with the people doing the work, not just the people writing slides.
Plan for failure like a Samurai
Pogrebna's take cut earlier in the process: most teams don't plan for things going wrong. Accept the possibility of failure upfront, then pre-plan your next move. That's how you reduce surprise and speed recovery.
- Run a pre-mortem before launch: list top failure modes, triggers, and responses.
- Set explicit kill-switches and pivot criteria. No endless zombie projects.
- Decide who calls the stop, with what data, and within what time window.
The executive myth is cracking
The idea that leaders have all the answers is fading. AI has put research-grade tools in everyone's hands, flattening how work gets done. Leadership is shifting from dictation to co-creation. Resist it and you don't just slow strategy-you drain culture.
- Open-source your thinking: publish decision memos with assumptions and trade-offs.
- Invite bottom-up research and prototypes; make the best idea win, not the biggest title.
- Clarify decision rights: who proposes, who decides, who is informed.
Intuition first. Data makes it safer
Pogrebna called out a common habit: many strategic calls are made intuitively, then justified with data. Flip it. Keep intuition, but de-risk it with AI and analytics that make uncertainty more manageable.
- Write your gut call as a hypothesis. Define what data would change your mind.
- Run small, cheap experiments before big commitments.
- Use AI for scenario exploration and stress tests-not as a stamp of certainty.
Uncertainty is where value lives
Most companies don't have data problems. They have decision problems. Value comes from making better decisions under uncertainty than competitors do. Metrics should reflect outcomes for stakeholders, not internal comfort.
- Customers: service, quality, net value (not just "efficiency").
- Employees: engagement to execution linkage (not perks).
- Partners: reliability and mutual margin.
- Owners/boards: cash generation, risk-adjusted returns, time-to-learning.
AI won't replace responsibility
AI is useful and overhyped. It won't replace boards or accountability. Start with the problem, then decide if AI is the right tool. Current systems are better at creativity than precision, so don't outsource judgment.
- Problem-first checklist: decision type, stakes, acceptable risk, and data quality.
- Co-design with stakeholders; define guardrails before you deploy.
- You don't need an "AI strategy." You need a business strategy that AI supports.
Shadow AI is already here
Employees adopt AI faster than executives because it saves time. That bottom-up use creates new exposure, especially with third-party tools. Many boards can't see how models are trained or where data originates-and that's now a governance issue.
If you need a reference point, map policies to the NIST AI Risk Management Framework and adapt from there.
- Inventory all AI use. Approve safe tools; block data exfiltration.
- Demand vendor transparency: training data sources, evaluation results, red-teaming.
- Keep a human in the loop for material decisions. Log prompts and outputs.
- Prepare an incident playbook: model errors, data leaks, content liability.
Let go to gain control
Pogrebna warned that algorithmic influence is eroding decision independence. Kenny flagged a different risk: AI funding models may not hold. Their personal trade-offs said a lot-she let go of time; he let go of control.
The message is simple: stop chasing certainty and authority. Build judgment and humility. Operate without the illusion of total control-and you'll make better calls, faster.
30-day executive playbook
- Frame one strategic bet with explicit uncertainties and assumptions.
- Run a pre-mortem and set kill-switches with clear owners.
- Launch three low-cost experiments; timebox to two weeks.
- Publish a decision log for your team. Reward disconfirming evidence.
- Stand up lightweight AI governance: tool inventory, data boundaries, approval flow.
Want structured upskilling by role for your teams? Explore curated options here: AI courses by job function.
Your membership also unlocks:
