Concepts assembled and contextual framework developed by Michael J. Negrete, PharmD. Built on the foundational scholarship of Rita Gunther McGrath and related organizational strategy research. Drafting and editing supported by AI tools. © 2026 Michael J. Negrete
As corporations accelerate the adoption of artificial intelligence, many risk mistaking ubiquity for strategy. The rallying cry to "implement AI everywhere" may sound visionary, but without discipline it often leads to scattered pilots, fragmented investments, and little measurable value. Rita Gunther McGrath's discovery-driven planning, first introduced in the Harvard Business Review in 1995, offers a disciplined alternative: a method that turns uncertainty into structured learning and measurable progress.
Rather than assuming success, discovery-driven planning starts with the premise that assumptions must be surfaced, tested, and refined through evidence. This philosophy is especially powerful for AI initiatives, where both the technology and the business context evolve rapidly. As McGrath observed, "The only plan is to learn as you go."
Why Discovery-Driven Planning Matters for AI
AI initiatives frequently fail not because of technical weakness but because of managerial overconfidence and poor assumption management. Organizations often act as if data quality, model performance, and user adoption are certainties rather than hypotheses to be tested. McGrath's framework reframes these uncertainties as assets: each assumption is a potential insight waiting to be validated. In this way, discovery-driven planning transforms risk into a managed learning process, replacing traditional project planning with experimentation and feedback loops.
From "AI Everywhere" to Strategic Focus
A broad directive to implement AI everywhere tends to produce unfocused experimentation. To make AI adoption both strategic and sustainable, organizations must translate this mandate into a focused, discovery-driven portfolio. That requires connecting AI investments directly to enterprise goals, testing assumptions early, and building governance that learns over time.
Five Key Actions to Translate the Mandate
| # | Action | Purpose |
|---|---|---|
| 1 | Anchor to strategic outcomes | Identify 3–5 core business objectives and align AI initiatives accordingly. |
| 2 | Map value vs. feasibility | Use a 2×2 grid to identify where AI can deliver the highest strategic impact given current organizational readiness. |
| 3 | Formulate testable assumptions | Convert beliefs into hypotheses. Example: "We believe predictive maintenance will reduce downtime by 25% if data reliability is at or above 90%." |
| 4 | Establish an AI Portfolio Board | Form a cross-functional team to vet, fund, and monitor AI projects. |
| 5 | Institutionalize learning loops | Capture lessons from each pilot and apply them systematically to future scaling decisions. |
This disciplined translation of ambition into evidence-based governance ensures that AI investments drive both learning and value creation.
The Five Discovery-Driven Disciplines Applied to AI
McGrath's original framework identified five disciplines essential to managing innovation amid uncertainty. Each translates directly to AI initiatives and guides design, testing, and scaling decisions.
| Discipline | AI Application |
|---|---|
| 1. Define Success | Articulate measurable business outcomes before any modeling work begins. |
| 2. Benchmark Context | Compare AI maturity and performance against peers to establish realistic goals. |
| 3. Map Capabilities | Assess gaps in data infrastructure, governance, and skills. |
| 4. Surface Assumptions | Explicitly document what must be true for success and test those assumptions systematically. |
| 5. Create Checkpoints | Use learning milestones to decide whether to continue, pivot, or stop a project. |
Metrics Before, During, and After Implementation
AI initiatives should be measured across three distinct phases: readiness, validation, and impact. The risk in most organizations is that measurement begins too late — after deployment, when it is no longer possible to distinguish signal from noise. Establishing baselines before work begins is not optional; it is the foundation of any honest evaluation.
| Phase | Focus | Example Metrics |
|---|---|---|
| Before | Readiness & Baseline | Data completeness, governance maturity, employee AI literacy, baseline KPIs, ethical compliance. |
| During | Learning & Validation | Model accuracy, bias audit scores, adoption rate, hypothesis validation rate, learning velocity. |
| After | Impact & Sustainability | ROI, net promoter score, cost savings, model drift rate, retraining cadence, stakeholder trust index. |
The Balanced AI Scorecard
The Balanced AI Scorecard complements discovery-driven planning by ensuring that performance, capability, ethics, and trust evolve together. No single dimension is sufficient on its own. An initiative that delivers ROI while eroding employee trust or producing biased outputs has not succeeded.
| Quadrant | Focus | Example Metrics |
|---|---|---|
| Value Creation | Business impact and ROI | ROI, revenue growth, productivity gain, cost reduction. |
| Capability & Learning | Internal development and adaptability | Number of AI-trained employees, retraining frequency, data quality improvement rate. |
| Governance & Ethics | Risk, fairness, and compliance | Bias audit frequency, explainability score, adherence to ethical AI standards. |
| Adoption & Trust | User and stakeholder confidence | User adoption rate, employee engagement with AI tools, trust index, reputation score. |
Leading the Discovery-Driven AI Transformation
Executives must treat AI not as a destination but as an evolving portfolio of learning experiments. They should reward transparency, validated learning, and the willingness to pivot based on evidence. This mindset turns uncertainty into competitive advantage.
The leadership question shifts from "Did it work?" to "What did we learn that improves our next iteration?" That is the essence of discovery-driven transformation.
Conclusion
AI transformation demands humility, discipline, and curiosity. McGrath's discovery-driven planning offers a rigorous framework for transforming uncertainty into advantage. By institutionalizing explicit assumptions, staged learning, and continuous feedback, leaders can make AI both strategic and sustainable.
The goal is not to apply AI everywhere.
The goal is to build organizations that learn everywhere they apply AI.
McGrath, R. G., & MacMillan, I. C. (1995). Discovery-Driven Planning. Harvard Business Review, July–August.
McGrath, R. G. (2013). The End of Competitive Advantage. Harvard Business Review Press.
McGrath, R. G. (2020). Discovery-Driven Growth. Knowledge at Wharton Interview Series.
Wikipedia (2024). Discovery-Driven Planning. en.wikipedia.org/wiki/Discovery-driven_planning