Performance Support

Contextual Help



Training Employees on Corporate Technology with Generative AI: Recommendations and Limitations

Generative Artificial intelligence (GenAI) and generative pre-trained transformer (GPT) have been all the rage since the public launch of ChatGPT in November 2022. But is it all it’s cracked up to be?

For some things, it’s helpful, but for the most part, it has been blown up in importance. Maybe it’s the start of something bigger, but for now, it’s mediocre help for training.

But that’s not to minimize how it can help you when creating training. Don’t expect it to do a lot for you, and don’t expect everything (if anything) well. At this point, it acts as a mid-grade personal assistant that can help get started, but not much more.

Don’t expect it to be a replacement for anything but bad content at this point, though.

I’m going to cover a few ways you can use GenAI to help you in your journey to train non-IT employees on corporate technology. That’s because that’s our ultimate goal at the techstructional. I’m writing this blog post because I know there’s no threat from generative AI anytime soon, especially for technical training. The quality of content from GenAI is far inferior to anything a human who does a good job can ever compete with on any front.

Don’t expect AI to replace a massive number of jobs anytime soon.

I use GenAI daily for various tasks, but they are mainly on the back end. If anything from GenAI does make it to the front end, it’s heavily edited and unrecognizable from the version provided by our computer overlords. Our most significant use for GenAI currently? Getting ideas for blog post organization. But even that is heavily edited, rearranged, and not recognizable from what AI provides.

As far as how we’re using Generative AI in actual training development, our use is limited. But it can still be helpful for basic tasks that aren’t useful for us to think about and don’t require serious research.

There are far more examples of how not to use GenAI for training employees than there are good uses for it. But, to save room, I’ll cover approximately the same recommendations for each.

How To Use Generative AI To Train Employees on Corporate Technology

Whether you’re using ChatGPT, Google Bard (now Google Gemini), or any of the hundreds of other GenAI options, these are a few ways you can use GenAI to help train non-IT employees on corporate technology.

These uses for GenAI MIGHT save you some time or give you a quicker path to creating better content. That might mean you can spend time on more important tasks for humans or sit back and have another coffee because of your time savings.

Whatever you do with your newfound time is entirely up to you. I’ll use it to stare at the ceiling for a bit and contemplate the downfall of humankind because of GenAI. Just kidding.

Generate Questions

With the right prompting, you can generate decent questions for content. While this likely doesn’t help much in most corporate software training, it could be helpful for security training. That’s because security training for most users is based on ideas and theories rather than actual physical systems or software walkthroughs.

Here’s a good example we had ChatGPT generate for us:

What is the most secure and easy-to-remember password that combines random words, numbers, or symbols, with the first letter of each word capitalized, and a length of at least 12 characters?

A) G0odPa$$word!

B) CorrectHorseBatteryStaple

C) Sunshine12345

D) Qwertyuiop[]\

And the feedback for these questions:

The correct answer is B) CorrectHorseBatteryStaple.

While option A) G0odPa$$word! includes some secure elements such as numbers and symbols, it is a common pattern that hackers can easily guess.

Option C) Sunshine12345 is a weak password as it is a common word followed by a numerical sequence.

Option D) Qwertyuiop[]\ is a simple password and does not include a combination of numbers or symbols.

Option B) CorrectHorseBatteryStaple, on the other hand, is a passphrase composed of four random, unrelated words that are easy to remember but difficult to guess. It also meets the length requirements and includes a mix of upper and lower case letters. This type of passphrase is considered strong and can be used to create a secure password.

And if you’re interested, here’s the prompt used to generate that question:

Write a good question to ask for the most secure and easy-to-remember password with one correct answer and 3 distractions that are around the same length.

But you may want to generate a bunch of questions based on content. If you wanted to do that, you would ask ChatGPT for what you want, including telling it you will input the content. It will generate what you want, so you have to proofread, revise, and then you’re done.

Just don’t abuse the questions because nobody likes a quiz that doesn’t relate to what they learned or doesn’t help reinforce training in a useful way. I have to call that out because it’s an essential point that can’t be made enough.

Create Fake Information

I don’t know about you, but creating fake names and information about fake people in training can be challenging. Generating fake people is necessary for general scenarios that aren’t real-life examples from your organization. That might mean their name, statistics, or history relevant to the training is required.

That’s not an efficient way to spend time because while it’s important, it likely doesn’t impact the training enough to waste hours. It’s possible to give some general directions to GenAI tools and have more details generated. You can even create a whole list of fake names to pull from.

This could be useful for the medical industry, where everything needs to be anonymous, which means no actual patients anyway. GenAI can help you with that by giving it just a bit of context and information.

I used the following prompt to generate topical characters since I’m not creating them for specific training. But, when you develop training, you’ll have to specify what you want to use, and you can tell ChatGPT (or Gemini) that.

Generate a few characters for my training with their age, some general information, and topical health information.

That’s all you have to do, and you can get detailed character information like this:

Name: Sarah
Age: 27
General Information: Sarah is a young professional who works long hours at a tech company. She enjoys exercising and trying new healthy recipes.
Topical Health Information: Sarah is currently dealing with high levels of stress due to work demands, and is seeking ways to manage her stress and maintain her mental health.
Name: Tom
Age: 55
General Information: Tom is a retired factory worker who enjoys spending time with his grandkids and tinkering with electronics.
Topical Health Information: Tom has recently been diagnosed with high blood pressure and is trying to manage his condition through diet, exercise, and medication.

If you want first and last names, ask it!

Summarize Information

Sometimes, it’s necessary to reference other content in training, especially for corporate technology with dense user guides (yuck) or requirements. Instead of simply linking to external content in the training, paste that information into your generative AI tool and ask it to summarize the content.

You can offer up a computer-generated summary for employees so they can get the gist of the information in less time. That will help them get the essential information quickly while distracting less from the training.

Convert Content to Different Formats

If there’s one thing that Gen AI excels at, it’s summarizing and simplifying existing content. That means you could take a written script for a course and ask the tool to summarize it, make suggestions for breaking up the content, or even create topic summaries based on each section. Then, you can use that content in different places or even create executive summaries of training content.

It’s even possible to take a course narrative script and have Gen AI rewrite it into an instructor-led training format.

One thing to consider when using AI tools is that your company’s information is likely proprietary. That means you shouldn’t put it into a system your company does not approve. Tools like ChatGPT are free; once content is put into them, that information becomes public. There are enterprise versions now, but even those must be approved by your company.

Gen AI may not be great at generating original and good content, but it can summarize and reformat current content well. Notice I’ll always call out good. That’s because AI doesn’t create much good content, typically mediocre or bad.

How Not to Use AI To Train Employees on Company Technology

There are a lot more poor uses for AI than there are helpful. One of the best ways to ensure you’re not poorly using AI is to not rely too much on it and revise and review everything it outputs.

My best suggestion is never to have it generate original content. Have AI only work with content you input into it unless you know the topic well and can verify its output properly. It will make stuff up (hallucinate) and act confident when presenting wholly fabricated information.

It will even make up links that are completely fake and never existed and cite it as a source so be very careful.

Don’t Generate Learning/Performance Objectives With AI

This is first on my list because I’ve seen a lot of suggestions for using AI to generate objectives. The suggestion is always to paste in your content and ask for objects.

The huge (massive) problem with this is that objects should never follow the content. The objectives are what people must do to succeed in their jobs; it’s not throwing a bunch of content at them and then figuring out what they could learn from that content later.

You could make no single more significant mistake with AI than having it generate objectives for you.

If you input content and ask AI to generate objectives, you’re misguiding your training, and it’s not likely to focus on performance objectives rather than learning objectives. Objectives must always be one of the first things created after the needs analysis. Otherwise, you’ll end up with a content dump that helps no one and hurts everyone. It will also waste time.

That’s why we always begin with nothing when we begin creating training. You can’t input anything into ChatGPT and expect performance objectives, so how could AI be a good solution for generating objectives?

I’ll tell you how: it can’t!

Don’t use it for objectives to generate content based on what the subject matter expert gives you unless it’s a summary for your personal use.

Replace Human Trainers or Actors

It’s now possible to generate an entire video script and video with absolutely no human input aside from copying and pasting content. We did this when ChatGPT first came out to see what we could develop using only AI.

We generated this video by asking ChatGPT to create the script, pasted that script into DeepWord to generate the fake person video, and then generated captions with HappyScribe.

It was pretty simple to put together, but we don’t recommend it. It’s unpleasant to watch; the people look dead in the eyes, and the mouth is odd and creepily unnatural. It doesn’t add enough value to justify putting anything like this into production.

Not to mention, AI voices are fake-sounding for the most part. But, aside from that there’s no reason to show someone talking if that person is fake anyway, just use the audio.

AI will never replace real trainers with real personalities, faces, authentic voices, and genuine humanity. While we’re not a fan of instructor-led training for technical training, most types of training will benefit from a real person rather than a computer only. Even for narrations in courses with no face, AI voices can’t compare to a real person with personality in their voice.

Have you ever noticed that even all the synthetic voice companies use real voice actors in their marketing videos? That’s because synthetic voices are garbage.

So, don’t replace anything public or user-facing with synthetic voices.

We want personality!

Generating Forward-Facing Content

Don’t generate forward-facing content with AI and then publish it without heavy revisions. We would never publish AI content, even heavily revised, but that’s up to you. It’s going to be rewritten entirely. By forward-facing, we mean content that actual people see.

As mentioned in the beginning, we use AI to generate outlines and then modify those and create 100% unique content based on that outline. So, the AI content is never seen in the training content or even our blog posts. AI should be kept in the background for everything but summaries and rehashed other content.

Replacing Human Interaction

While we’re not a big fan of instructor-led training for corporate IT training, sometimes it does make sense. In the case of security, instructor-led training could be a good option. It allows people to ask questions and get immediate answers.

It’s not possible to replace human interaction with AI, and not even AI-driven chatbots can mimic actual human behavior and personality. That means you can’t replace instructors, experts, and others who help organizations navigate IT changes. People often want to talk to others when dealing with complex topics and changes. AI can’t offer that same type of support; therefore, replacing humans with AI isn’t a good option.

With great power comes great responsibility.

I mentioned earlier that I would offer equal suggestions for ways to use AI and not to use AI. That’s the only reason I’m stopping here. There are infinite ways you should not use AI for training content or training. While it’s a powerful resource for certain things, it can easily be abused, and most AI-generated content is very easy to pick out of other content.

Wrap Up

You may love or hate AI, making you either love or hate some of our do’s and do nots. However, some are unarguable and not a good use for AI to train employees on corporate IT topics. Technical topics are tricky for training sometimes.

All the complexities of IT translate into training for corporate IT systems. Not to mention that corporate technology is typically full of custom and proprietary information. That makes it extra difficult for AI. But, with some of the uses we’ve presented here, you should be safe if you don’t input anything from your organization into any unapproved Gen AI tool.

Avoid relying on AI to replace humans or take over too much (if any) user-facing presentation. Also, don’t let AI guide you down the path of poor training as an order-taker who doesn’t properly analyze training requirements.

AI won’t replace quality work from humans but will augment our daily work.

One thing remains true with AI and will continue for some time: it doesn’t offer the same benefits humans do for training. That is at least true for good training, perhaps not for poor-quality training with which AI might be able to compete.

If you’re looking at ways to improve the performance of your corporate IT training for non-IT employees, we’d love to meet with you to discuss your next project. I know we could make your next tech project more successful, save your employees time, and save your organization money while training employees better than they would have otherwise been.

Schedule a free consultation so we can learn more about your project and how we can help it be a huge success.

Leave a Comment