Book & framework
A New Framework for Organizational AI Strategy
Abhimanyu Gupta
AI Strategy = Goal + Investment + Discipline
Start with the problem. Invest with intention. Build the discipline to make AI last.
Explore the FrameworkThe Sandwich Framework
A New Framework for Organizational AI Strategy
Cover in progress
The central problem
Organizations get pulled toward new models, platforms, and tools before naming the business problem they are trying to solve. The result is expensive experimentation that never resolves into value.
The Sandwich Framework reverses the sequence. It begins with a clear goal, then asks what investment that goal justifies, and finally what organizational discipline is required to sustain the implementation.
The framework
Each layer depends on the one beneath it. Remove any one and the strategy collapses into activity.
Top slice
Discipline
Filling
Investment
Bottom slice
Goal
Make it sustainable.
Value appears only when an implementation survives past the pilot. Discipline is the governance, monitoring, ethics, accountability, and behavioral change that keep an AI system honest and useful once the novelty has worn off.
Choose what deserves resources.
Resources are finite, so strategy is a sequence of choices. Investment is far broader than buying software: it is people, data, infrastructure, expertise, organizational capacity, time, and readiness. Investment should follow strategic purpose rather than technological enthusiasm.
Know why.
An AI strategy begins with a meaningful business problem, stakeholder need, or opportunity — not with a technology looking for somewhere to land. A clear goal names who benefits, what changes for them, and what the organization is willing to decline in order to focus.
Discipline in practice
Organizational AI adoption is not only a technical challenge. People respond to uncertainty, perceived loss, unfamiliar workflows, social cues, framing, incentives, and defaults — often more strongly than to the technology itself.
People weigh what a change might take away more heavily than what it might add.
The current workflow wins by default, even when it performs worse.
The first number or first demo quietly sets expectations for everything after it.
Adoption spreads through visible peers far faster than through mandates.
Small changes to defaults and prompts often move behavior more than training does.
The same rollout reads as a threat or an opportunity depending on how it is described.
Sustainable AI adoption requires understanding how people respond to change.
Value creation
An AI initiative should be able to say which lever it moves: customer value, employee experience, operating cost, or value capture.
The most a customer would give up for the value created.
What the organization actually charges.
What it takes to deliver, including the AI capability itself.
The least suppliers and employees would accept to participate.
Human-centered leadership
Models do not carry an organization through change. Leaders do — by naming a purpose, absorbing ambiguity, and staying credible while the work is still uncertain.
Who is this for?
Explore the ideas
Select any idea to read a short note. These grow into fuller essays over time.
Related work
Program leadership, resource allocation, and organizational change in practice.
Visit LeadershipWeekly writing on the innovation, AI, and implementation questions around the framework.
Visit Ideate Forward WeeklyWhat translating ideas into working solutions looks like at a smaller scale.
Visit InventionsWorking on an AI strategy that has to survive contact with a real organization?
Let's connect