Knowledge Workers are generally understood to be anyone who uses their brain primarily to get their work done. In the modern context, this means they probably spend a lot of time sitting in front of a computer, though, it doesn’t have to always be that way. Still, there is an interesting divide in the knowledge worker economy when it comes to AI: tech and non-tech roles.
Tech roles are probably the most impacted by AI through the use of LLMs for writing code. Of course, it goes far beyond writing code at this point: you can practically spin up an entire development team of just Agents who will ping each other back and forth and work on separate work trees to get the job done. No worries if you don’t know what these words mean, since this article is about understanding why there is a divide between these roles.
What I’m here to tell you is this: AI will never replace the human parts of work. The conversations we have between each other, where inference is subtle. Judgement calls that need a person. Ethics that can only be felt by a human heart, with an inner voice telling you “this isn’t right”.
However, AI can be an extremely good tool to use to get those decisions written down, formatted, and disseminated. Honestly, one of the most annoying parts about being a knowledge worker: you have the knowledge, and it’s a lot of work to get it where it needs to be.
For this article, I want to go over a few use cases that may help, that I have either personally worked on, or learned from my peers in this space.
Marketing#
Developing your content market fit: basically, determining who your target audience is since, let’s face it, your product likely solves a similar problem to other existing products, and find out what they’re talking about. AI can help you scrape this information from places like TikTok or other social media at scale until you build a better idea of what your customers are actually talking about. An analysis of your ideal customer base has never been so accessible.
Setting up your content calendar and publishing: there are plenty of tools that exist now that expose the ability to post on different platforms through AI. My favourite is PostProxy, based in Germany, which you can also integrate into your own app if you’re interested in exposing that ability securely for your end customers. But even for your own internal processes, it’s cheap to use, and the documentation is AI-friendly. Just tell the AI exactly what you’re trying to achieve, and set up your .env with your API key (ask the AI to tell you how if you don’t know, but never give your API key to an AI model).
SEO: AI can be a game changer for SEO. Since I’ve handed over the reins (I approve everything, and I steer my goals, but it’s an AI coding agent actually updating all my titles, tags, and other metadata) my stats have greatly improved. Clicks are still hard to generate these days with google summaries, but my impressions are up 51% for the current 28 day period, and I’m writing this at the end of August. You know, the month where most of Luxembourg is on vacation. I’m excited for September.
Video editing: you can edit videos programmatically. If you have all the pieces you want to put together, and an idea of how they should be stitched or what the flow is, an AI agent can actually do the cutting and piecing together. It takes a while to set up your agent in a way that works for you, but once you do, it’s wonderfully freeing to simply shoot the footage and then hand it over for processing.
Sales#
Creating a dashboard of reminders: through ingesting all your communications (please do this locally, or with an AI inference provider who gives you a DPA), you can set up a dashboard to remind you who you haven’t replied to, what ideas have been bouncing around between you and your colleagues, when your meetings are, what sales cycles need more pressure, etc. I have my own Paperwork tool set up that helps me plan my days. Everything is local, so I’m not worried about compromising anyone’s data, and it doesn’t reply for me, because I know how to say what I want to say. What it does do is relieve some mental labour from my own already-full schedule.
Customer Service#
Live querying of company documents and procedures: I hate AI chatbots for customer service. They’re actually getting pretty accurate, but when they’re wrong, it’s an incredibly frustrating experience for the end customer. It’s much better speaking to a person who is actively trying to solve the problem than an AI agent who is stuck in a hallucination loop. Using AI to build a RAG Graph system that can ingest all your company’s rules, product info, customer info (though make sure you are able to keep this securely constrained to only the people who should have access), and other information can help a human customer service representative perform faster and better than a human with a traditional knowledge base setup.
Product#
Using AI coding agents for small product changes, not architectural ones: let’s say the CEO calls the Product Manager and says the buttons are too small. This becomes an “emergency” in a lot of orgs because of the chain through which this request was passed. Meanwhile, the dev team is surviving on Red Bull and pizza and trying to get an actually important feature out for a sales demo tomorrow.
With a properly set up CI/CD pipeline that includes AI agents accessible by key non-dev team members, you can make small front-end changes without disrupting anyone’s already stressful workday.
I’m not saying to hand over the reins to a non-tech team member, but I am saying that resizing a font 2 points shouldn’t be put on the shoulders of someone with 15 years of experience coding, which it often is.
General#
Keeping your dashboards up to date: whether you work in marketing, sales, product, customer service, or any other part of a business, you probably have a dashboard that some upper management is on you to update quarterly, monthly, or (shuddering to think about) weekly.
You can’t just say “Hey, AI helper, update my dashboard for me” and expect a good output. But what you can do is say “Here are all the places the data lives. Here are the tools to access them. Here is a template with the expected output. And here is who to send it to, and how often.” With context, tools, guardrails, and a clearly defined definition of what “done” looks like, today’s AI models are actually able to do this type of work quite efficiently.
Formatting documents: this is the simplest thing on the list but it’s just crazy to me how few people are doing this. I hate Microsoft Word. You know who’s really good at using it? AI agents. Let them do it. You wrote the content: now step back and let it flow.
In Conclusion#
Notice that in no place did I put here content creation, strategy, responding to emails, or anything that should require a human-level understanding of the business itself. You know your business, or at least, I hope you do. AI is a tool to help you execute faster. If you know what “good” looks like, you can free up a lot of time using AI.
It can be hard to know where to start with AI especially if you’re only using its basic capabilities. The best place to start, I find, is face-to-face with someone, a human, who can relate to your pain points and understands how to use AI for them.
If you’re interested in doing this with me, feel free to get in touch. I offer 1-1 and team sessions.