Upscaling Your People: Advanced AI Training

Are your employees using AI every day but barely scratching the surface of what it can do? Have you invested in AI training only to watch people quietly go back to their old habits within weeks?

In this article, you'll discover a structured framework you can model for transitioning your team from basic AI users to strategic AI users.

Why Advanced Employee Training Benefits Businesses More Than Large-Scale AI Initiatives

Most companies are making a single large bet on AI: one multi-million-dollar application built by outside vendors. They pour resources into that initiative while ignoring the fact that if their employees can't understand, contribute to, or build upon what's been created, it's unlikely to succeed.

But if the collective AI knowledge within an organization sits at roughly level three, and the initiative being deployed requires level eight or nine capability to operate and maintain, nobody in the company can meaningfully contribute. The only person who knows what's going on is the person who was hired to build it—and when that person leaves, the initiative collapses.

The alternative to the single large bet approach is advanced training for everyone.

The goal of advanced AI training isn't to turn everyone into a developer. It's to take each person from wherever they are right now and move them steadily toward building tools that solve the problems only they fully understand, because they're the ones doing the work every day.

When 50, 100, or 200 employees each build their own, even relatively simple, tools built inside ChatGPT, Claude, or Gemini—the cumulative effect outpaces a single custom application in both speed and cost. Each person solves the problems they know best, because nobody understands a job's friction points like the person doing the job.

As people reach levels five and six, something else happens: their ideas for larger, more sophisticated AI applications become genuinely useful to the organization. They've built enough on their own to understand what AI needs to work—what data needs to be connected, what the model needs to know, where the edge cases are. That collective intelligence makes the organization a far better client to any outside builder they eventually hire, and reduces dependence on vendors in the first place.

The broader shift John describes is one from AI resistance to AI curiosity. When employees build something real, they see the power firsthand. They feel like they're contributing something meaningful. Confidence grows, resistance falls, and the organization starts generating AI ideas from the inside out—rather than waiting for leadership to hand down the next initiative from above.

#1: Establish Two Types of Governance Before You Begin Training

You’ll need a system for monitoring two things simultaneously.

The first thing to monitor is the progression of an employee’s AI skills. You’ll need to benchmark how long each employee’s tasks take now and how long they take after training. The before-and-after measurements serve as evidence that training is producing real results.

The second thing to monitor is what employees create with AI during and after training, because security and oversight must scale as skills progress. When employees use AI for simple queries or to write blog posts, oversight requirements are minimal. However, when employees begin running agents connected to external databases, the security and oversight requirements grow substantially.

John's framework keeps those two tracks moving in parallel, so organizations aren't scrambling to impose controls after the fact.

Pro Tip: For organizations that want to provide employees with access to multiple frontier AI models within a secure, compliant environment, John recommends platforms like BoodleBox and NebulaONE. Both are built with HIPAA and FERPA compliance and give employees access to multiple AI models through a single secure interface without the data exposure risks that come with using consumer-facing tools on company networks.

#2: Format Hybrid, Goal-Centric Training

A training failure pattern John sees consistently is two-fold.

First, there is a reliance on self-guided study. The majority of employees left to work through self-guided training on their own don't get past the first couple of modules. They're busy, and the demands of their existing jobs don’t go away because you’ve asked them to take on AI training.

To address this failure point, John runs a hybrid model of recorded modules that employees can work through on their own schedule, along with live office hours a couple of times each week. Without that live human connection, momentum dies quickly, and the program produces a very expensive set of partially watched videos.

Second, there is a lack of personal relevance. Without a compelling, personally relevant goal to work toward, employees often treat the training as background noise and quietly return to their old habits.

To neutralize this failure point, John helps each person identify five to ten specific things they could realistically build in AI, before any training begins. We’ll touch on this more fully later in this article. This pre-ideation step gives people a reason to go through the training. They’re already thinking, “If I could get AI to do this thing that drives me crazy, that would be a win!” and so they engage in entirely different ways than someone who is told by their boss to go watch some videos.

Every person who completes training should leave with at least one tool, workflow, or prompt system that saves them at least three hours per week.

#3: Run Two Team Assessments  

John runs two distinct assessments before training begins.

Map Your Team’s AI Capability Against Four Stages of Mastery

John organizes AI capability into ten levels, grouped into four distinct stages: literacy, fluency, mastery, and stewardship.

Literacy (Levels 1–3): Employees at this stage understand what AI is, what it can and can't do, and how to use it safely. They know how to ask a clear question, refine their prompt when the output falls short, and evaluate whether the output is reliable. They don't blindly accept the first answer.

Fluency (Levels 4–6): This is where people start using AI regularly inside their actual job, improving work quality and speed. They've begun building simple tools: a custom GPT, a Claude project, or a structured prompt they share with teammates. This is the stage where real business impact starts to appear.

Mastery (Levels 7–9): An employee at this level is building repeatable workflows, connecting tools, using reusable prompt systems to solve ongoing problems in their specific role, and beginning to work with AI agents. At this stage, the level of governance and security oversight required also increases, as employees connecting to external data sources or running API calls need closer monitoring.

Stewardship (Level 10): The final level is where someone is managing both people and AI systems. Stewards oversee employees who have been authorized to build and run agents, and they're responsible for making sure AI is being used properly at an organizational level. John notes that no one in his training programs has reached level nine or ten yet, largely because security practices haven't kept pace with what AI can now do.

98% of employees at every organization John trains are at level three or below. To determine where everyone in your organization is, create a questionnaire with about 20 questions designed to understand everyone’s individual AI skill level.

The first 17 should include questions such as: Have you built a knowledge base? Have you created a prompt and shared it with teammates? Have you built a custom GPT? Have you built a Claude project? Have you built an agent? Have you connected two different workflows together? Have you ever turned on secure features in ChatGPT? Can you explain when to use AI and when not to?

Each of the final three open-ended questions should require employees to submit an actual sample prompt. You need to see exactly how someone structures an instruction.

John uses this assessment to produce a visual heat map showing where an organization's AI capabilities are currently concentrated. As training progresses and people are retested, you should see those dots shift into levels four through six. A smaller group might even push toward seven and eight. That movement gives leaders concrete evidence that training is producing a return on investment.

Determine Your Team’s Role Type Makeup

The assessment, similar to the PAIE assessment, presents employees with 15 to 20 questions in which they read a scenario and identify which response option is most like them and which is least like them. The forced choice prevents employees from simply picking the answer they think their manager wants to hear.

The assessment identifies each employee's primary working style across four categories:

• Doers execute. They love getting work done and thrive when given clear tasks.
• Administrators organize. They build and follow rules and love creating structure.
• Innovators generate ideas. They think creatively and love conceptual problems.
• Connectors build teams. They thrive when helping others work well together.

John uses this breakdown for two purposes.

First, it helps him anticipate where each person will need more support during ideation.

Administrators and Doers tend to think inside existing constraints, so when it comes time to imagine a custom GPT or Claude project they'd want to build, they often get stuck. Innovators generate ideas quickly but may need help grounding them in something practical. Knowing these details in advance lets John and his team provide each type with the right direction.

Second, he uses PAEI to build what he calls an AI council, the internal group that oversees AI adoption across the organization. The composition of that council directly determines what kind of culture emerges.

A council composed entirely of Administrators produces overly restrictive policies that get locked down before they ever gain traction. Innovators are essential because they're the ones who champion AI, get people excited, and keep the organization moving forward. Administrators balance that energy with the necessary “not so fast” logic that prevents the organization from moving recklessly. Without Innovators, nothing gets adopted. Without Administrators, nothing gets controlled.

The council needs all four types, or it will be structurally biased from the start.

#4: Give Your People a Problem to Solve Before They Watch a Single Video

As noted, a common reason AI training fails isn't the curriculum. It's that employees enter training without a personal stake in what they're going to build.

Asking people what they want to build produces blank stares. Asking them to name something in their job that drives them up the wall produces immediate, specific, energized answers. John’s team starts the upskilling process by asking every employee one straightforward question: What do you do every week that is repetitive, slow, frustrating, or mentally draining?

The next step is what John calls the Perfect Day Exercise. Employees are asked to imagine their ideal workday—specifically, what are all the tasks they'd love to hand off to someone else, with confidence that those tasks would be done with excellence? That wishlist becomes the raw material for what they'll build.

Once an employee has identified their candidate task or process, the conversation moves to two diagnostic questions. First: could AI realistically help with this? Second—and this is where most people stop short—should we simply speed up the existing process, or should we redesign the process entirely because AI changes what's possible?

John describes this distinction as the difference between bolting AI onto something old and asking, “What would this process look like if it were designed with AI from the start?” The second question consistently surfaces the bigger wins.

The tools employees build don't have to be technically complex. A custom GPT, a Claude project, a document analyzer, or a structured prompt workflow are all within reach for someone at the fluency level. The goal isn't novelty. It's practical time savings, repeatable quality, and confidence.

Three examples from John's training programs illustrate the range:

Patent Analyzer: A chemical industry professional who files 20 to 30 patents per year was spending $30,000 annually in legal fees. He built a patent analyzer that could review a patent he was preparing to file, cross-reference it against existing patents to identify potential conflicts, and help him rewrite the application before handing it to his attorney. His legal fees dropped by 90%, and he eliminated a $15,000 software subscription entirely.

Home Construction Cost Estimator: A woman working in real estate built a home construction cost estimator that delivered estimates within 3% of a $20,000-per-year software application she had been paying for. She hadn't started with that idea. She came in planning to build something that analyzed market buying signals, but as she worked through the training, she pivoted to the tool that would actually solve her biggest problem.

RFP Assessment: An office furniture CEO who works with businesses filling 20,000–40,000 square feet of space came through John's training already familiar with ChatGPT.

His sales team received 350-page RFPs to bid on large commercial projects, and just determining whether to bid the go/no-go decision took three to six hours per document. If the decision was yes, the team would put two and a half people on it for two to three weeks to write the response. The result: they could realistically bid on only three projects per year, each worth between $250,000 and $1.5 million.

By the end of training, the CEO had built a tool that could digest a 350-page PDF, surface the furniture-relevant sections, and deliver a go/no-go recommendation in 20 minutes. If the decision was yes, the same tool helped him generate a complete RFP response in two hours with one person himself. He immediately saw that bidding on three to five projects per month, instead of three per year, was now possible.

Other Notes From This Episode

• Connect with Michael Stelzner @Stelzner on Facebook and @Mike_Stelzner on X.
• Watch this interview and other exclusive content from Social Media Examiner on YouTube.

Where to subscribe: Apple Podcasts | Spotify | YouTube Music | YouTube | Amazon Music | RSS

✋🏽 If you enjoyed this episode of the AI Explored podcast, please head over to Apple Podcasts, leave a rating, write a review, and subscribe.

---

Stay Up-to-Date: Get New Marketing Articles Delivered to You!

Don't miss out on upcoming social media marketing insights and strategies! Sign up to receive notifications when we publish new articles on Social Media Examiner. Our expertly crafted content will help you stay ahead of the curve and drive results for your business. Click the link below to sign up now and receive our annual report!