AI is in regular use across mid-market workforces, yet capability often lags access. The next stage of upskilling sits inside the workflow itself, organized around the tasks people already do. Here is what that shift looks like and how organizations are designing for it.
Artificial Intelligence is now a normal part of office life. According to Helios HR's 2026 Mid-Market AI Workforce Trend Report, 74% of workers use AI regularly, with 46% using it at least once per day. The tools are already in people's hands. Whether people know how to use them well in their day-to-day roles is a different question.
It would be easy to assume the answer is more training. Yet most organizations are not short on learning material. AI tutorials, prompt libraries, and self-directed courses are everywhere, and Helios HR found that 48% of companies simply encourage their teams to work through that kind of content on their own. Only 5% offer a structured AI training program, and 29% have no framework at all. But more content does not automatically build capability.
What is missing is context. People can open an AI tool and find no shortage of material on how it works, yet still struggle to connect it to the specifics of their own jobs. Part of the reason is that the training methods most organizations lean on were designed for a slower kind of change.
Why traditional training methods can't always support AI upskilling
Learning programs often hit a wall when it comes to AI, simply because the technology is changing so quickly. Traditional frameworks, such as ADDIE and SAM, depend on having approved, teachable materials that can be distributed to learners. When it comes to AI, those materials might be out-of-date before they reach a single employee.
Evaluation systems also struggle to measure AI usage. For example, Kirkpatrick generally assumes standard outcomes, such as whether an employee learned skills that they could apply to their current role. AI tools are so flexible that they can potentially transform a person's role, opening up new ways of working and making existing workflows obsolete. How does an L&D coordinator measure that?
This is why even well-designed AI training programs tend to leave a set of questions unanswered:
- Which AI tool is right for this specific task?
- How does AI apply to this role in particular?
- What business problem does this solve?
- How does this fit into the existing workflow?
- What is approved or safe to use?
- Where is the right place to start?
These questions can't be answered by a standardized training program delivered to all staff. Answering them calls for a new approach to professional development, one that is collaborative, flexible, and built around learning by doing.
5 steps for building a modern AI training program
Traditional training methods focus on teaching people to use software the "right" way, walking them through menu options and helping them make sense of error messages. Working with AI is more collaborative. Instead of returning an error message, an LLM chats back and offers feedback, so people tend to find their own way of working through experimentation rather than instruction.
That changes where the learning needs to happen. Traditional corporate training tends to separate learning from the work, sending people to a session or a module and asking them to apply it later. AI capability builds faster when training sits inside the work itself, organized around the tasks people already do. At that point, it starts to resemble workflow coaching more than classroom instruction. The most practical way to begin is small: a single workflow in one team, scaled out once it proves itself. The five steps below follow that path, each one building on the last.
1. Start with one high-value workflow
Rather than rolling training out to everyone at once, the first step is to pick a single workflow in one department and build around it. The best candidates are tasks that come up often, take real time, and produce outputs the team can already judge for quality. Concentrating on one workflow keeps the early effort focused and creates a clear signal that the later steps can build on. Spreading too wide too soon tends to dilute the wins before anyone can see them.
Workflows that make good starting points include:
- HR teams drafting job descriptions or interview guides
- Recruiting teams shaping sourcing strategies and candidate outreach
- Operations teams producing routine reports or process documentation
2. Design role-based scenarios around it
With the workflow chosen, the next step is to shape it into scenarios built for specific roles. This is where general AI familiarity becomes something people can actually use, because an HR business partner and a recruiter approach the same tool with different goals, inputs, and standards for a good result. A useful scenario spells out the role, the task, the approved prompts, and an example of what strong output looks like.
Each scenario tends to define:
- The role and the specific task it covers
- The tools and prompts approved for that task
- A concrete example of what "good" looks like
Defining "good" up front matters, because without it people are left guessing at quality and the wins become harder to repeat.
3. Run a small pilot with light guardrails
Once the scenarios exist, the next step is to test them with a small group rather than the whole organization. A pilot of five to ten people over a few weeks is enough to refine the scenarios, see what works, and surface the questions that real use throws up. Keeping it small also keeps it manageable, so the effort can sit with an HR lead and a single project owner instead of waiting on a dedicated L&D team.
This is also where governance belongs, kept light and practical:
- A short list of approved tools and the data each one can handle
- Clear examples of strong output to anchor quality
- A simple way for participants to flag questions and gray areas
Light guardrails give people the confidence to experiment, because they no longer have to guess whether they're about to break something.
4. Capture what works and scale it
A pilot only pays off if its lessons travel. As the scenarios settle, the next step is to gather what worked into a playbook the rest of the organization can use: the strong prompts, the role-based scenarios, the governance calls, and the examples of good output, all in one place. Short walkthroughs make it easy to pick up, whether they are five-minute videos or one-page guides.
From there, scaling works best when it stays close to the work:
- Office hours where people share what's working
- Communities of practice that keep the momentum going
- A running library of prompts and proven scenarios
The point is to make individual learning compound, so a useful discovery spreads across teams instead of being reinvented five times over.
5. Keep the program going
The final step answers the problem raised at the start: AI moves too fast for any fixed course to keep up. Rather than treating training as a one-time rollout, the program runs as a continuous loop that adapts as the tools change. The playbook becomes a living asset, updated as new features ship and as people discover new scenarios worth sharing.
In practice, keeping it continuous means:
- Refreshing guidance and content as the tools evolve
- Retiring scenarios and prompts that have gone stale
- Folding new discoveries back into the playbook
Measured this way, success shows up in how people work: whether they are tackling tasks in new and more effective ways than before.
Why one AI win unlocks the next
The advantage of starting small is that a single win rarely stays contained. Once people see AI handle one task well, they start connecting it to the work around it. An HR business partner who saves time drafting interview guides soon wonders what else AI could speed up, from onboarding communications to policy summaries to manager coaching notes. The same pattern shows up in recruiting, operations, and finance once that first scenario produces something visibly useful.
This is why the choice of workflow in step one carries so much weight. A workflow that delivers an early, visible win earns adoption on its own, as people share what they have learned and colleagues ask how to do the same. Helios HR found the effect in how people feel about the technology too: in organizations with a defined AI strategy, 94% of people express positive sentiment toward AI, compared with 54% overall.
What L&D looks like when AI lives in the workflow
As training moves inside the workflow, the role of learning and development changes with it. Producing courses still matters for the things courses do well, and two responsibilities now sit alongside that work.
The larger of the two is continuous enablement, which treats the playbook as a living asset. New features ship every few weeks, and new scenarios surface as people experiment, so the ongoing work is to capture what is working, retire what is not, and keep the guidance current.
L&D also moves closer to HR strategy. Whether AI investment turns into real adoption is a workforce question as much as a technology one, and that puts HR at the center of it. Helios HR found that organizations with successful AI strategies are more than twice as likely to include HR as a key stakeholder, at 72% compared with 31% in those still early in the process.
Underneath all of it is a question employees keep asking: how does this help me do my job better? A training approach that answers that, inside the work and role by role, is what turns access into capability.
Need help with AI upskilling?
AI is changing how work gets done across nearly every function, and the organizations that adapt fastest tend to be the ones whose training meets people inside their workflows rather than away from them. Starting with one workflow, building role-based scenarios, piloting with a small group, and scaling what works is how that capability takes hold. It is the next stage of workforce development, and it is closer than many leaders assume.
Helios HR helps mid-market organizations make that shift:
- AI consulting to design the next stage of capability building
- Strategic HR to align AI planning with your wider people strategy
- Training and development for the programs that sit alongside enablement
- HRIS consulting for the technology layer underneath
- Engagement and performance to update performance expectations for AI-assisted work
Ready to build the next stage of AI capability in your organization? Connect with Helios HR to map out where to start.
FAQ
What is the difference between AI training and AI enablement?
AI training delivers content about how AI tools work. AI enablement equips people to apply those tools to specific tasks inside their workflow, with role-based scenarios and examples of strong output. Most mid-market organizations benefit from both, with enablement layered on top of structured training.
How long does an AI pilot usually take?
A focused pilot typically runs four to eight weeks with five to ten participants. That window is long enough to refine the scenarios, capture what works, and prepare a playbook for scaling, without pulling people too far from their primary responsibilities or losing momentum partway through.
Should HR or IT lead AI training in our organization?
HR plays a central partner role in the most successful approaches. Helios HR found that organizations with successful AI strategies are more than twice as likely to include HR as a key stakeholder (72% versus 31% in earlier-stage organizations). AI capability is a workforce question as much as a technology one.
How do we measure whether AI upskilling is working?
Adoption depth tends to be more informative than access. The most useful measures include how often AI is used in target workflows, the quality of outputs against defined examples of good work, whether teams are creating new scenarios on their own, and the change in time-to-completion for target tasks.
What if we don't have a dedicated L&D team?
The model is built for that reality. A single workflow, a small pilot, and a project owner working alongside an HR lead can carry the early phases. The playbook that comes out of the pilot is what lets the program scale without standing up a full L&D function from scratch.
What if our people are already using AI without approval?
That pattern is widespread across mid-market organizations. The opportunity is to channel that existing energy into approved workflows with light governance, rather than trying to shut it down. The pilot model gives people a sanctioned place to experiment with clear examples of what good output looks like.
Related resources
Helios HR: 2026 Mid-Market AI Workforce Trend Report
McKinsey: The State of AI in 2025: Agents, Innovation, and Transformation