The conventional approach to enterprise AI workforce development follows a familiar sequence: select a platform, run a procurement process, configure the system, then train employees on how to use it. The logic seems reasonable. You need the technology in place before you can train people on it.
The problem is that by the time employees get to training, the critical decisions have already been made without their input. The workflows, the data structures, the use cases, the assumptions baked into how the system was configured: all of it was decided by a smaller group with different information and different daily realities than the people who will actually use the system.
Sharma Vedula, Head of Solutions at Data Society, describes what this produces in practice:
“If you haven’t invested in change management alongside the technical rollout, you’ll hit resistance that looks like a technology problem, but it’s actually a people’s problem.”
Reframing AI workforce development as a people problem before a technology problem changes nearly everything about how it should be approached.
WHY AI FAILS THE WORKFORCE TEST AT SCALE
The dynamics that make AI pilots succeed are precisely the dynamics that disappear when you scale. Vedula describes this pattern with clarity:
“A pilot runs in a controlled environment. You’ve got a small team, handpicked data, and engaged sponsor, and a high tolerance for imperfection. The moment you scale, all of that goes away.”
At scale, you are dealing with data quality that nobody cleaned up before the pilot. You are hitting edge cases the pilot never encountered. You are working with users who were not part of the design process and have zero patience for friction. The workforce development program that works for 25 engaged, self-selected employees will not automatically transfer to 2,500 employees with different levels of comfort, different job requirements, and different prior exposure to the tools.
This is why the organizations that get AI workforce development right start with people, not platforms. The technology decisions flow from a clear understanding of who will use it, how they currently work, and what capability they actually need to develop.
Related reading: AI Upskilling vs. AI Implementation: Why Enterprises Need Both and in the Right Order
https://datasociety.com/ai-upskilling-vs-ai-implementation/
THE HHS MODEL: BUILDING INTERNAL CAPABILITY FIRST
Data Society’s work with the Department of Health and Human Services illustrates what a people-first AI workforce development model looks like in practice.
The COLLAB program was not designed to deploy technology and then train the people who would use it. It was designed to build internal capability first, across 25 employees from NIH, the Office of the Secretary, and other parts of the agency, so that when AI tools were applied at scale, the workforce already understood the data, the workflows, and the decisions they were supporting.
The results from that first cohort were concrete: capstone projects generated more than $500,000 in annual cost savings and freed up four full-time staff. But the number that tells you the most about whether the program actually worked is this: when it expanded, the program received 450 applicants for 30 available spots.
Employees do not compete aggressively for training programs that they perceive as box-checking exercises. They compete for programs that they believe will genuinely advance their capability and their careers. The demand signal is a quality signal.
Related reading: AI Training for Government Agencies: How Public Sector Organizations Are Closing the Data Science Skills Gap
https://datasociety.com/ai-training-government-agencies/
THE CHANGE MANAGEMENT LAYER MOST ENTERPRISES SKIP
Change management in the context of AI workforce development is not a communications plan or a rollout checklist. It is the ongoing process of keeping people in the loop, gathering their feedback, and building the trust that makes adoption possible.
Vedula is specific about what this looks like:
“The more you loop in your teams upfront not after the fact the more you will get a better buy in and you’ll get more patience through from them because they would understand whereas if you bring them at the end they may not have the patience to adopt it because at the end of the day adoption is the key because you are bringing changes to the company the way things are done now.”
Adoption is not a technology metric. A system can be perfectly implemented and still have near-zero adoption if employees were not brought in early enough to understand why it exists, how it will affect their work, and what they are expected to do differently.
The change management work happens before the platform goes live, not after. It means involving employees in the use case prioritization process, sharing what was decided and why, being transparent about what the system can and cannot do, and creating clear channels for employees to surface problems once it is in use.
WHAT GOOD AI UPSKILLING PROGRAMS LOOK LIKE
One of the practical challenges of AI workforce development at scale is that employees arrive with very different levels of comfort and capability. A workforce development program that assumes everyone is starting from the same place will fail the employees at both ends of the spectrum.
Data Society’s approach when working with large, complex organizations, such as the City of Dallas across 42 city departments, reflects this reality. As Vedula describes:
“Meet people where they are, build the connective tissue and design for the next two years, not just this quarter.”
The practical implication is that workforce development programs need to be tiered, adaptive, and built around the real variation in employee readiness. Foundational AI literacy for employees who are new to the tools, applied upskilling for employees who are ready to use AI in specific workflows, and advanced capability development for the employees who will become internal champions and trainers.
Getting this right requires investing in the diagnostic work upfront: assessing where the workforce actually is, not where you assume it is.
Explore Data Society’s AI upskilling programs: https://datasociety.com/upskilling/
Frequently Asked Questions
The most common failure mode is treating workforce development as a final step after the technology is already in place. When employees are not involved in the design process, when training is generic rather than role-specific, and when there is no ongoing support structure, adoption rates suffer and the investment does not produce measurable outcomes.
Start with a clear picture of where the workforce actually is: different departments and roles will have very different readiness levels. Design tiered programs that meet employees where they are. Build in a change management layer that involves employees before the tools go live. Create an ongoing feedback cadence so the program can adapt to what employees are actually struggling with.
There is no universal answer, but the organizations that treat AI workforce development as a one-time event rather than an ongoing capability consistently underdeliver. The initial capability-building phase might run weeks to months, but sustaining and deepening that capability requires a continuous development model built into how the organization operates.
Change management is not a separate workstream from workforce development. It is the foundation that makes workforce development work. Bringing employees in early, communicating what is changing and why, and creating channels for ongoing feedback are what convert training into adoption. Without change management, the most technically sophisticated workforce development program will produce underwhelming adoption.
BUILD A WORKFORCE THAT IS READY FOR AI, NOT JUST TRAINED ON IT
There is a meaningful difference between a workforce that has been trained on AI tools and a workforce that has genuinely developed AI capability. The first can demonstrate tool familiarity. The second can apply AI to real problems and adapt as the tools evolve.
Data Society builds AI workforce capability with enterprises and government organizations from the ground up. Talk to our team about designing a workforce development program that produces real adoption and measurable results: https://datasociety.com/contact/
