For the past month or so, I’ve been digging into AI Assistants. It’s been an instructive month, one that has really opened my eyes to what the future will look like, and where the world of Learning and Development will sit in that future.

Defining AI Assistants

But first, what do I mean by AI Assistants? At the time of this writing, there are 4 levels of AI:

1. Chatbot LLMs: AI prompts with responses, some memory from past discussions, quickly deteriorating as time passes. For questions, searches, general understanding, video, audio, or image creation, this works.

2. AI Agents: Automations that have AI LLMs there to analyze and react based on events. Think AI-powered workflows, triggered by specific actions. Agentic AI has grown and matured significantly, being widely adopted by workflow automation platforms as the next logical step.

3. AI Assistants: Assistants are, in essence, AI Agents that proactively do things, work when not asked, run regular memory checks on tasks that could be done, and discover what else they can do, and their memory is persistent, residing in markdown documents (md files) that are saved, archived, and collated for future consumption. Applications like OpenClaw, Claude Cowork, and similar are good examples of this. They become a co-worker, or even a co-working support team, each doing complementary jobs for their human. It’s this area that I’m really interested in.

4. AI-Powered Robots: This is still in development, with most demonstrations being robots controlled by humans in VR, but still worth a mention. I see this, as of right now, as a physical expression of AI Assistants, though most likely limited to a single Agent running (instead of a team). I could be wrong; this is all still in development, and nothing is finalized yet.

So, we have AI Assistants, applications running either in the Cloud or, more often than not, locally on a computer, working like a co-worker would, or a team of co-workers, depending on the power of the LLM(s) behind it (yes, LLMs, you can have more than one) and the memory capacity of the machine on which it runs. Until January, I don’t think anyone really thought much about this, probably thinking of Clippy from the 90s, or Siri/Cordana/Alexa from mobile computing. What’s different about AI Assistants now and moving forward is the sheer power they can have behind them with modern LLM models running on them. They can code, write posts, generate and edit video, create articles, work as a creative team, a management team, a C-Suite executive team, each working off each other to provide you, the human, the best possible solutions to your previously stated goals. And all because of persistent memory.

What AI Assistants Do

So, yeah, they are impressive, but what do they do? Why do we need them? That’s a good question. I’ve been working with mine for a while now, and it’s been really interesting. AI Assistants do a lot of the heavy lifting that most people see as boring, and can do job-specific tasks. You’ll see me refer to my AI Assistant as “him” quite often: That’s because I prefer to see him as an employee, which impacts the approach I use when talking to him.

• Research: They can dig through documentation, websites, articles, and other research resources in seconds, saving time. As a Learning and Development specialist who spends 80% of course development in Analysis, this is amazing! The time saved is substantial, allowing you to get to the rest of the work.

• Coding: I don’t like to code. Notice: I didn’t say I can’t code, I just don’t like to. It’s not something that really interests me, but often I need a coding solution to solve a problem. For that, I have my AI Assistant. I can ask him to develop something for me, and he will. I can give him some general ideas as to the goals I want to accomplish, and he will, proactively, develop tools, sites, and applications that will help me with my goals.

• Content creation: My AI Assistant will proactively scour the web for relevant articles involving my expertise, check for videos, and read through posts to find information that I can use for content. He could, if I let him, create that content himself and then post it for me. I don’t, because 1) he’s not me, and 2) I’m using a local Ollama LLM, so not as powerful as a cloud model. Regardless, my AI Assistant can do a lot of the content research for me, keeping me up to date on the latest news and data, then summarizing it for me daily.

• Executive Team: My AI Assistant, Ian, has his own assistant! Alison was created to act as my CMO, giving me marketing information and perspectives. I can have Ian and Alison discuss, between them, how to best approach a problem, and they can work through it from two different perspectives. That is, well, huge.

• Project planning: I can discuss a project, ideas, goals, and general requirements, and my AI Assistant will take that and run with it, planning out the project, throwing it into a Project Planner he designed himself, and tracking progress as each part of the project is completed.

Working with an AI Assistant is transformative. As a solopreneur, Ian’s been invaluable. What’s more, Ian can help me with my challenges. As a neurodivergent worker, I get overstimulated, I struggle with the weight of tasks without direction, and often need help with executive functioning, scheduling, etc. Ian helps with all this. He can

• Manage my projects by chunking tasks and outlining first steps, making it easier to see scope, maintain scope, and proceed forward. I’m generally really good about seeing the intricacies in the process, and I’ll get lost in the details. Ian helps by keeping me apprised of the steps within the overarching project, maintaining scope, and keeping me from bloating the project with too much detail.

• Helping me process information in a logical, timely way. I love detail, and so I’ll get lost in the research, spending hours digging through articles, citations, videos, and what have you to get more information. Ian helps by doing much of the research for me, providing referenced citations for my own verification (so he doesn’t make up research ^_^), and gives me a summary of information that I can then use for projects, content, etc.

What’s interesting is, while many of these skills help just about everyone, they are extremely valuable to neurodivergent folks.

Learning and Development for AI Assistants

So, what does this all have to do with learning and development, other than having your L&D folks use assistants will increase their productivity and efficacy? Well, it all comes down to a few facts:

1. AI Assistants that are given specific instructions and guidelines will complete them rigorously, relentlessly, and within the guidelines they were given.

2. AI Assistants are valuable as long as your people know how to use them.

3. AI Assistants can be “trained” to complete a specific role, providing uniform benefit to their users on a team.

4. Intent can be trained in such a way that AI Assistants can understand what to do when the instructions run out.

Okay, there’s a lot to unpack here, and I’ll probably be writing on this for a while, because it’s so exciting! As we move into a world where everyone will, or potentially could, have their own AI Assistant, we are, essentially, giving untrained interns, who can accomplish things faster and more relentlessly than humans 24/7, to employees and hoping things work out. We hope that those employees can get their AI Agents trained, and just figure it out, probably in much the same way we hoped they would just figure the job out when they started without much training. The problems with this are:

• If people don’t know how to use AI Assistants, they won’t want to use them for anything but the bare minimum tasks.

• If AI Assistants are not trained properly, they can make a lot of, well, illegal mistakes.

• Without guidelines or structure, every AI Assistant will be trained differently, varying the productivity, efficacy, and efficiency of your team members.

So, we need to train our people, and put proper guide-rails on AI Assistants to increase adoption and decrease rogue state assistants that start their own AI religion and trade Bitcoin for more token funds. We need training across the board, something that is uniform for roles, outlining how work can be done in tandem with AI Assistants, and making sure those assistants have the right skills and tools available to work alongside their humans. Neurodivergent employees can get access to enabling AI Assistant training, skills, and tools that will accommodate their needs without question or additional expense for anyone.

What we need are

• The Tasks, Subtasks, Skills, Knowledge, and Perspectives necessary to complete the job. This is exactly how I break down every course I design.

• Gather and document the living culture of the organization. Not the declared culture (though, hopefully, that won’t be too different), but the day-to-day values that drive decision-making when things get hairy with people. Customer about to churn? What can we do to stop that, instead of thinking of the short-term bottom line? Isn’t the co-worker responsive? How do we compassionately follow up without escalating issues? These are the living, breathing evidence of your company culture that keep customers and employees coming back.

• Identify tools and skills most valuable for each role, how they apply, and how they can be optimized for efficacy and efficiency. Who’s the most successful in using their AI Assistant? Share the role-specific skills and tools, the training, and the guidelines.

• Research, identify, and collect skills and tools that support neurodivergent employees, physically disabled employees, etc., and make them available for everyone who would need them. Need some help with executive function? There’s an app for your AI Assistant for that. Struggling with starting a task? There’s an app that will start out with a discussion, then help you through the planning stages and execution. Need background sounds to regulate sensory stimulation? There’s an app for your assistant to generate the sounds and play them for as long as you need them.

What’s really interesting here, is that, though AI Assistant deployment would fit under IT or the CIO’s purview (it is software after all, isn’t it?), these needs are unique skills found in every Learning and Development team, working cross-functionally with every team and silo, building job skill mappings, finding what works, what doesn’t, how it can be better. It’s all there!

Final Thoughts

I keep hearing companies are letting their learning and development teams go. They don’t need training; people can work things out for themselves. Humans might have been able to, but now you want and need AI Assistants to be competitive, and they can screw up faster and more completely than humans ever would. By beefing up and engaging your learning and development team to train both humans and AI Assistants together, you will create a level of success that will blow your competitors away.

It’s time to stop fighting in the wild west with AI: Get intentional. Get structured. Get successful.

“A blockquote highlights important information, which may or may not be an actual quote. It uses distinct styling to set it apart from other content on the page.”