A couple of years ago, I wanted to implement an HR case management system. A month into demos and vendor selection, the IT team came to me and said that we should use AI to create the system.
As in many other companies, our leadership had issued the mandate “Every department must implement AI.”
So we did. We took my original functional requirements and built them out with AI. And it worked.
I was very proud of it at the time. And it was indeed a great cross-functional project. But I am a bit embarrassed by it now.
Because, apart from being cheaper and a bit more glitchy, it did nothing more than a standard HR case management system would.
See, I understood the processes in depth but did not know enough about the technology to be able to see what else is possible. And my IT partners understood AI but did not know enough about the work to propose something bolder. And this being our first big internal AI project, none of us thought to step back and take the time to redesign the workflows. We had deadlines and expectations, so we just went with the requirements we had.
And this is what happens in many organizations right now. The people who know the work do not know the technology. The people familiar with the technology do not understand the processes. And leadership, who understands neither, sets targets that are fueled not by real needs but by the fear that the company is falling behind on the AI front. And to make it even more fun, ownership is vague and unclear.
So you end up with agents on top of agents that either recreate the wheel or bring only marginal value without creating real disruption.
In order for AI to realize its promise and for organizations to truly gain from it, they should treat it not as a technological transformation but as an organizational transformation.
Stop Mandating AI, Start Mandating Problem Resolution
AI implementation should not be a goal in itself. It should serve a specific purpose. This purpose should be defined first at the strategic level and then at the tactical level. And each AI project should focus on resolving a specific problem or achieving a specific goal.
Build a Cross-Functional AI Governance Group
This group should include operations, process owners, data/tech, HR, and risk. People who understand the process, people who understand the technology and people who understand the legal, ethical and people implications. This will be the “translation engine” of the organization that will make decisions on what AI projects are worth the time and investment and will bring real value to the organization.
Not everyone needs to know how to build AI agents. But everyone should know:
- What AI can and cannot do
- What makes a process automatable
- How data quality affects outcomes
Over time, organizations should create roles in each function that understand both the process and technology in depth and can serve as the translator between process, technology and strategy.
Redesign Processes Before You Automate Them
We are all limited by what we know and what we are used to. For AI to bring true value, we have to let go of our preconceptions of what a process should look like.
We should start with the input for a process and the desired output and redesign the full process based on the capabilities of AI. If done well, this will not only eliminate process steps. It will shift decision rights, responsibilities and hand-offs.
Manage the Downstream Impacts
Full process redesign has an impact on roles, KPIs and communication flows. These need to be determined and managed at the project phase and not as an afterthought.
HR should be part of every big AI project to assess the impact on people, roles and org structures and manage those thoughtfully.
Manage the Human Side
AI implementation should be treated as any other large organizational transformation.
Employees should be involved early and included in decision-making whenever possible. Communication and training should be executed with the same effort and attention as in any other large change initiative.
Resistance and fears should be addressed, not discarded or ignored.
Without these steps, organizations are not implementing and utilizing AI. They are performing AI. And they risk missing the great opportunity these technologies bring to not only make things faster and cheaper, but also make them truly better.