Patrick George/Ikon Images
In conversations with executives, I’ve noticed a common pattern: Many are aware of agentic AI coding assistants like Claude Code but assume that these tools are relevant only to software developers. This assumption, while understandable, is rapidly being proved wrong. It turns out that the features that make AI coding tools useful for programmers can be equally valuable for knowledge work that involves no programming whatsoever.
Let’s explore what distinguishes these tools from the AI chatbots that are accessed through a browser or a desktop app (like Claude, ChatGPT, and Gemini) and consider several noncoding examples that convey the broad applicability of these tools for personal and enterprise work.
Knowledge Work With an AI Chatbot Like Claude
Imagine that you have an important meeting coming up with a new sales prospect. You want to be well prepared: You’d like to know about recent company and industry news, what the person and company have been posting on social media, and other relevant background information. This will involve visiting multiple websites and synthesizing everything into a briefing you can review with the team before meeting the prospect.
Since AI chatbots can visit web pages and extract information, you can use them to speed up part of this process. You go to the web interface (or desktop app) of your favorite chatbot and prompt the AI:
Research this person [LinkedIn URL] and their company [company URL]. Find recent news about the company and industry. Summarize what I should know before meeting with them. Create a one-page briefing document.
After some back-and-forth to refine the output and ensure its correctness, you download the briefing document and share it with the team.
This works reasonably well and is certainly faster than doing it without an AI tool. But what happens before your next client meeting? You may need to scroll back through your chat history to find the original prompt, copy it into a new session, swap in the new client details, and iterate through the refinement process again. If a colleague wants to prepare for their own meetings the same way, you’d send them the prompt via email or Slack, along with instructions on how to adapt it.
In other words, conversational AI tools are good at producing a one-time output, but they don’t easily support repeatable workflows. Each time you need to prepare for a meeting, you’re essentially starting from scratch.
Wouldn’t it be nice if you could automate this into a reusable capability — something you or a teammate could run in seconds before an external meeting?
You can. Here’s how.
Knowledge Work With an Agentic AI Coding Tool
With an agentic AI coding tool, you start with a prompt similar to what you’d use with a chatbot, but the tool works directly with the files on your computer.1 It can read files you already have and, most importantly, edit them and create new files. In other words, whatever you can do with your files, it can do too!
This changes what’s possible in important ways.
For example, you can create a folder called “meeting-prep” and prompt the tool:
Research this person [LinkedIn URL] and their company [company URL]. Find recent news. Check my notes folder for any past correspondence. Create a briefing document and save it to this folder.
The tool will fetch the pages, search your notes, and write a briefing. You can then open the briefing file, read it, and see that it is too long, is missing the person’s job history, and doesn’t highlight the specific product line relevant to your meeting. So you can further prompt:
Shorten it to one page. Add a section on their career background. Focus the company news on their enterprise segment.
The tool will make the changes and update the briefing file. After a few rounds, you’ll have something that captures exactly what you want.
This is when things get interesting. Before you close the chat, you can add:
Write an INSTRUCTIONS.txt file that describes step-by-step how to generate this kind of briefing for any person and company.
The folder now contains not just the briefing but also an INSTRUCTIONS.txt document that captures the approach you arrived at through iteration: what sources to check, what sections to include, how long it should be, and what to emphasize. The instructions reflect the refined version of the task, not your first attempt.
Let’s fast-forward to the next time you need to prep for a client meeting. You open a fresh session. The tool has no memory of your earlier conversation, but that’s easily fixed: You can point it toward the folder and prompt:
Follow INSTRUCTIONS.txt to create a briefing for [new LinkedIn URL] and [new company URL].
The tool will read the instructions, fetch the relevant pages, search your notes, and produce a briefing in the format you’ve already refined. The files in the folder, especially the INSTRUCTIONS.txt file, carry all of the background knowledge needed for the tool to “remember” what it should do.
When a colleague needs to prepare for their own meetings, you can share the folder. They’ll see the instructions and can run the same process themselves. They might even refine it further for their own needs, perhaps adding a section on competitive context or integrating notes from a different system.
At any point, if you or a colleague discovers a better way to do something, you can tell the tool to add it to the INSTRUCTIONS.txt file. Even better, at the end of every session, you can ask the tool itself to review the session, extract improvement ideas for you to consider and approve, and then add them to the INSTRUCTIONS.txt file. These improvements will compound over time rather than sitting in isolation in your chat history. And if you find yourself preparing for external meetings very frequently, you can go one step further and use advanced features (such as Skills) to make this process even more efficient.
Three Key Capabilities of Agentic AI Coding Tools
Let’s take a step back from the examples above and examine three interrelated capabilities for knowledge work that coding tools offer.
A form of “memory.” The tool can read what you’ve accumulated in folders, write outputs directly where you need them, edit files incrementally based on your feedback, and work with your existing folder structure. Obviously, this is much more convenient than manually uploading and downloading files to/from a chatbot and copy-pasting prompts from past chat sessions. But, more importantly, your files serve as memory: At the start of every session, you can simply ask the tool to read the relevant files, and it will “remember” whatever is needed to execute the task at hand.
Automation. When source materials change, you can re-execute the same analysis with a simple plain-language instruction, without having to manually repeat the workflow.
Parallel execution. These tools can run multiple tasks simultaneously, with each task operating independently. For example, you can ask the tool to execute the meeting-prep analysis for 10 different clients in parallel. The tool will essentially clone the instructions 10 times, execute each one of them simultaneously, independently of the others, and create 10 briefing documents. This approach will typically be much faster than doing a task multiple times sequentially.
In combination, the increased efficiency these capabilities provide can be transformative for knowledge work.
Building Blocks for Better Knowledge Work
AI coding tools like Claude Code are agentic — they combine several technical building blocks that, together, make the above capabilities possible.
The first building block is multistep reasoning. The tools can break a complex task into a sequence of actions (fetch a web page, extract key information, cross-reference it with your notes, and format a document, for instance) and plan how to execute the actions in order.
The second building block is adaptive execution. As the tools execute each step, they observe the results and course-correct when something fails or produces unexpected output.
The third building block is tool use. The tools can invoke processes that an external system executes on their behalf, such as reading a file, running a command, browsing a web page, or connecting to services like Google Drive or Slack or to databases. The key here is that nearly anything you can do on your computer (such as organizing files, converting document formats, extracting data from images, or querying a database) can be expressed as code (such as terminal commands) that invokes the right tools in the right order. This means that AI coding tools can take action in your work environment without requiring you to manually upload, download, or copy-paste anything.
Three Enterprise Task Examples
Here’s a quick example of “personal productivity” shared by Boris Cherny, the creator of Claude Code. He had a folder with 250 expense receipts that he wanted to organize into subfolders based on expense type and then extract key receipt details (such as date, vendor, and amount) into a summary spreadsheet. Cherny used Claude Cowork (a “knowledge worker” version of Claude Code) to do this with a couple of simple prompts in just a few minutes.
Let’s examine a few enterprise examples that illustrate how these capabilities may increase productivity.
1. Competitive intelligence. Say that a product team maintains a folder containing files on each competitor — positioning, pricing, recent announcements — alongside documentation of their own product. By pointing a coding tool at this folder and asking it to execute the INSTRUCTIONS.txt file in that folder, competitor information can be refreshed with up-to-date information, and a product-feature comparison table can be generated in minutes whenever the competitive landscape changes. The folder becomes a living knowledge base rather than a static deliverable.
2. Marketing campaign versioning. Launching a campaign across multiple audience segments and platforms typically requires dozens of copy variations. By storing brand guidelines, value propositions, and audience personas as files, platform-appropriate copy for each segment can be generated in parallel very quickly. If messaging changes midcampaign, updating just the relevant source files and rerunning the prompt produces a fresh set of variations without you having to start from scratch.
3. Due diligence review. Consultants and analysts frequently work with large volumes of documents — such as financials, contracts, and HR materials — spread across dozens of files. Because coding tools like Claude Code operate locally, they can navigate folder structures, cross-reference documents, and produce structured outputs without the manual uploading and downloading required by browser-based tools. This workflow advantage is significant when you’re working with hundreds of files. However, note that document contents are still transmitted to the tool provider’s API for processing, so organizations with strict data residency or confidentiality requirements should review data-handling policies before using such tools.
Risks of Using AI Coding Tools
What should you be wary of when using these tools? As with any LLM-based product, coding tools may make mistakes and hallucinate. As AI foundation models become more capable, it’s getting harder to find error cases, but perfect accuracy is not guaranteed, and end users still need to own responsibility for the output.
And because these tools operate on your local files, errors can have consequences. An instruction to “clean up this folder” could delete files you intended to keep. Caution is appropriate.
Another potential issue is that files processed by these tools can contain hidden, malicious instructions — a security risk known as prompt injection. A document from an outside source might include text that can manipulate the AI into taking unauthorized actions, such as sending data out or deleting files. Users should constrain the tool’s access to trusted folders and files and review its proposed actions before approving them.
Finally, some of these tools require comfort with the command line, though this barrier is disappearing fast. For example, Claude Cowork doesn’t require users to have command-line knowledge, and it’s likely that more such products will appear in the market soon.
How Business Leaders Should Prepare
The shift from “chat with AI” to “AI operating in your work environment with the ability to do the things you can do” represents a meaningful change in how these tools can positively impact knowledge work. The ability to work directly with the files on your computer is a seemingly modest one, but it vastly expands the potential for automation and enables the valuable accumulation and reuse of institutional best practices and knowledge.
While these tools have historically been marketed to developers and discussed primarily in technical forums, their capabilities are changing fast, and we are at the early stages of their crossover into the nontechnical mainstream. I recommend that business leaders start using them in their own work without delay. You will develop a visceral appreciation for the new automation capabilities that are unlocked, and this understanding will serve you well in helping your organization effectively adopt and use these tools to improve productivity.