Generative AI Needs a Strategic HR Framework Not Just Tools

The CEO Who Bought the Wrong Kind of Future

Last year, a chief human resources officer at a Fortune 500 company told me she had been summoned to a boardroom. The CEO had just returned from a conference where a vendor demoed a generative AI tool that could write job descriptions, screen resumes, and even conduct preliminary interviews. "We need this," the CEO said. "Roll it out by next quarter." She did. The tool worked. Sort of. Job descriptions came back grammatically perfect but culturally tone deaf. The resume screener flagged candidates who used certain keywords but missed the ones who had actual potential. The interview bot asked questions that were technically correct but made candidates feel like they were talking to a vending machine. Within six months, the company had burned through its AI budget, frustrated its HR team, and alienated a chunk of its applicant pool. The CEO blamed the vendor. The vendor blamed the implementation. The HR team blamed the CEO.

This story is not unique. It is playing out in boardrooms across every industry right now. Companies are rushing to bolt generative AI onto their human resource functions like they are adding a new feature to a software product. They are buying tools. They are not building strategies.

A new paper from Soumyadeb Chowdhury, Pawan Budhwar, and Geoffrey Wood, published in the British Journal of Management, argues that this approach is fundamentally broken. Their central claim: generative AI is not just another piece of HR technology. It is a force that reshapes how work gets done, how people relate to their jobs, and how organizations define value. You cannot manage it with a procurement checklist. You need a strategic human resource management framework, and you need one built on a theory that accounts for the fact that organizations are not machines. They are institutions made of people who resist, adapt, and sometimes subvert the tools you give them.

The Theory That Makes This Work

The authors anchor their framework in something called institutional entrepreneurship. This is a term from organizational sociology that describes how actors within an institution can break existing rules and create new ones. It is not about top down mandates. It is about people inside the system who see a different way and have the power to push it through.

Chowdhury, Budhwar, and Wood argue that generative AI is uniquely suited to trigger this kind of institutional entrepreneurship because it does not just automate existing tasks. It creates entirely new categories of work and relationships. A chatbot that can write performance reviews does not just save time. It changes what it means to evaluate a subordinate. A tool that can generate personalized learning paths does not just replace a training manual. It redefines who is responsible for career development. These are not technical changes. They are institutional ones.

The paper proposes a framework with five interconnected stages: alignment with business objectives, opportunity seizing, strategic resource assessment and orchestration, re-institutionalization, and realignment. Each stage is designed to force organizations to ask uncomfortable questions before they buy anything.

Alignment: Why Your AI Needs a North Star

Most companies skip this step. They see a demo, get excited, and ask IT to make it work. The authors argue that you must first ask: What is this tool actually for? Not in the generic sense of "improve efficiency" but in the specific sense of "this tool exists to solve problem X, which is connected to business goal Y, and here is how we will measure success."

Alignment means mapping the capabilities of generative AI to the strategic objectives of the firm. If your company is trying to reduce turnover by 15 percent, a generative AI tool that writes job descriptions is not aligned. A tool that analyzes exit interview transcripts and predicts flight risk might be. The difference is not technical. It is strategic.

Opportunity Seizing: The Trap of the Shiny Object

This is where the institutional entrepreneurship lens becomes critical. The authors note that generative AI creates opportunities that are not obvious from the vendor demo. A tool that summarizes documents can be used for compliance training. A chatbot that answers employee questions can become a data collection mechanism for understanding workforce sentiment. But seizing these opportunities requires people inside the organization who are willing to experiment, fail, and iterate.

The paper warns against what I call the "feature fallacy": the belief that the tool comes with its use case pre-installed. It does not. The opportunity has to be discovered, often by people who are not the ones who bought the tool.

Strategic Resource Assessment and Orchestration: The Hidden Cost

This is the stage where most implementations fall apart. The authors argue that generative AI is not a plug and play technology. It requires three kinds of resources that most organizations do not have in the right balance.

• ▸Data resources: Clean, structured, and ethically sourced data. Most HR data is a mess. Performance reviews are inconsistent. Employee records are scattered across systems. Training histories are incomplete. Feeding bad data into a generative AI tool produces bad results, but the tool is so fluent that the bad results look convincing.
• ▸Human capital resources: People who understand both the technology and the human dynamics of work. This is not just prompt engineers. It is HR professionals who can articulate what they need, data scientists who can build it, and managers who can interpret the output.
• ▸Relational resources: Trust between the HR function, the IT function, and the business units. If the sales team does not trust the AI that recommends who to hire, they will ignore it. If the legal team does not trust the AI that drafts employment contracts, they will rewrite everything.

The authors found that organizations that skip this assessment stage end up with tools that are technically functional but organizationally useless. They work in the demo. They fail in the real world.

Re-institutionalization: The Part Nobody Wants to Talk About

This is the most provocative stage in the framework. The authors argue that integrating generative AI into HR does not mean fitting the tool into existing processes. It means changing the processes themselves. They call this re-institutionalization: the deliberate dismantling of old norms and the construction of new ones.

Consider performance management. Traditional performance reviews are built on the assumption that a manager can observe and evaluate an employee's work. Generative AI can now generate continuous feedback based on work output, communication patterns, and project completion rates. If you use this tool, you are no longer doing performance management the way you used to. You have changed the institution. The manager is no longer the sole evaluator. The AI is a co-evaluator. That changes power dynamics, trust, and the meaning of a good review.

The authors argue that organizations must be intentional about this. Do not let the tool redefine your institution by default. Decide what you want the new norms to be, and then design the tool to support them.

Realignment: The Loop That Never Closes

The final stage is a feedback loop. Generative AI tools produce data that reveals new insights about how work actually happens. Those insights should feed back into business strategy. If the tool shows that your best engineers spend 40 percent of their time in meetings, that is not an HR problem. That is a strategic problem about how you allocate talent.

The authors emphasize that realignment is continuous. The environment changes. The technology changes. The workforce changes. A framework that worked last year may not work this year. The goal is not to build a perfect system. It is to build a system that can adapt.

What the Research Does Not Prove

The paper is a theoretical framework, not an empirical study. The authors do not run an experiment. They do not survey 500 companies and report a correlation coefficient. They synthesize existing theory and propose a new model. That means the framework is logically coherent but untested. It is a map, not a destination.

The authors acknowledge this. They call for future research to empirically test the framework. They want to know: Does alignment actually predict successful GAI adoption? Does re-institutionalization reduce resistance? Does the framework work differently in different industries or cultures?

There is also a deeper question the paper does not fully address. The framework assumes that institutional entrepreneurship is possible and desirable. But what if the organization is too rigid? What if the culture punishes failure? What if the CEO who bought the tool is the same person who will not tolerate the experimentation required to make it work? The paper offers guidance for the willing. It does not solve for the unwilling.

What This Actually Means

• ▸Stop buying tools. Start building strategies. The single most important decision you will make about generative AI in HR is not which vendor to choose. It is whether you have a clear strategic purpose for the technology. If you cannot articulate that purpose in one sentence, do not buy anything.

• ▸Invest in your data infrastructure before you invest in AI. The quality of your generative AI output is bounded by the quality of your input. If your HR data is fragmented, inconsistent, or ethically compromised, fix that first. A generative AI tool will not clean your data. It will amplify your mess.

• ▸Train your HR people, not just your machines. The authors emphasize that human capital resources are critical. Your HR team needs to understand what generative AI can and cannot do. They need to be able to ask the right questions, interpret the outputs, and challenge the tool when it is wrong. This is not a one time training session. It is a continuous capability build.

• ▸Design for resistance, not compliance. People will resist generative AI in HR. They will resist because they do not trust it, because it threatens their expertise, or because it changes how they work. The framework accounts for this through re-institutionalization. Do not try to force adoption. Design a process that allows people to experiment, fail, and adapt.

• ▸Measure what changes, not just what works. The real value of generative AI in HR may not be the efficiency gains. It may be the new insights you get about how your organization actually functions. Build measurement systems that capture those insights. They are the feedback that will keep your strategy alive.