- Upstream Ag Insights
- Posts
- What Claude Code and CoWork Buzz Means for Agribusiness Professionals
What Claude Code and CoWork Buzz Means for Agribusiness Professionals
A look at how it could be used by an ag retail location manager.
Claude Code has had a lot of buzz around over the last couple of months. However, it still has some challenges for the average computer layman (like me).
Recently there was another feature launched: Cowork.
Cowork is an agent-style feature built by Anthropic that moves it’s LLM interface, Claude, from a “conversational assistant” into a system that executes actual work. The feature accesses the files and systems on your computer. It’s designed to take tasks and complete them without you having to prompt step by step, effectively becoming a digital coworker.
Cowork runs inside the Claude Desktop app (currently macOS only and in research preview for Claude Max subscribers — I currently have access to the research preview and have been using this last week).
You can grant it access to specific folders on your computer. Once you describe a task, it plans and executes the work — reading, organizing, editing, and creating files until it delivers finished results. I talked about the future of this in From Chatbots to Digital Teammates: How AI Agents Fit Into Agribusiness, and now there are initial tools moving in this direction.
So, what does it explicitly mean for agribusiness professionals?
I am currently running a GenAi in Agribusiness Survey (Nearly 200 have completed so far, full report will be published the first week of February) and one question was asked surrounding agentic efforts and the level of trust agribusiness professionals have in letting an agent take an explicit action. Today, there is some trust in small tasks being executed, which is where Cowork operates today:

Cowork has the potential to be very powerful for the average agribusiness professional — whether it is organizing your desktop, researching new products, analyzing a spreadsheet, or drafting multiple tailored messages to customers at once.
But the utility will go much further. I wanted to think through how it will impact an ag retail location managers day-to-day as an example, particularly as the agentic capabilities move more directly into software (today it can connect directly into email, or horizontal software CRMs like Monday, design software like Canva and many others while also gaining context from folders on your computer).
To start, thinking about Claude Cowork the right way is important: it is effectively an employee that lives inside your computer. It will be able to operate your web browser. You can show it how you do a task. You can give it your SOPs. You can record the workflow once, augment with context and then it just does it.
Now consider some basic tasks of an ag retail location manager on a May morning:
They log into Agvance, Merchant Ag, or whatever ERP they run.
Look at an order vs. delivered report.
Look at a mountain of shipped product that need to be invoiced.
Try to organize a fleet of sprayers for the day.
Realize some products are short, call three different distributors to find supply.
That’s a lot of decision-making and time required.
Now imagine the location manager gives it access to the ERP, CRM and computer files etc. including access to:
Inventory
SOPs
Forecasts
Ordering workflow
Supplier ordering portals
Fleet dispatching capabilities
Effectively, the entire array of operating systems at an ag retail.
Claude, and even what I imagine ag software companies could build over time, agentic capabilities will be able to:
Compare current inventory against the upcoming customer orders, then flag shortages and draft a PO for suppliers/order from distribution.
Invoices in automated fashion, verifying the split-billing where needed, and emails the invoices to the customer.
Get the agent to scan due payment reports, identifying who is past due, then either writing an email (or text) or updting the point of contact to have a conversation, while inputting touchpoint automatically into the CRM.
Looks at every grower who bought corn seed from you this year but didn't buy a fungicide. It builds a "Gap List" and emails it to the specific agronomist assigned to that account with a note: "Ask GoodEarth Farms about fungicide for his 500 acres of corn."
Theoretically, this can occur every morning at 7:00 AM, or every other day, or once per week, or whatever your preferred cadence.
The Location Manager now focuses on exceptions and edge cases, handling the fires and aligning staff to ensure they have strong follow through on what was served up to them.
There are two crucial considerations here:
Claude is operating like an orchestration or coordination layer. Organizing systems and operations (for more on this, Reshuffle: Who wins when AI restacks the knowledge economy is a useful read) and helping to rethink how businesses are managed.
Physical Reality Needs to Meet Digital Reality - If there are not efforts to ensure inventory is accurate in the shed, conversations with customers are happening and being recorded, there is an incentive structure to drive accuracy etc, then none of this matters. Being realistic about what is necessary is very important, which goes back to what I discussed in November 2025 with Winfield and the need for processes to be in place before AI can be valuable.
Sounds easy — and I know there are always hang ups, but based on my initial usage of Cowork, there is some very real value to be unlocked for anyone that operates with a mouse and a keyboard.
I think we just scratched the surface with the above, too. There are plenty of other opportunities:
Pre-pay booking entry.
Compliance/Safety documentation filing.
Reconciling vendor rebate programs.
Updating price sheets.
As just a few other examples.
If a human currently does it on a screen, it can likely be automated inside the tools you already use.
Claude Code
I think the implications of Claude Cowork is important, but Claude Code also has implications.
Claude Code is an AI-powered developer tool that helps developers write, debug, and manage code by interacting with the Claude AI model directly, allowing it to understand, and modify codebases, run tasks, and even manage commits, all through basic text language commands.
Streamlined coding type capabilities are interesting for a retailer at the head office (or potentially location level).
These Claude tools decrease the cost of analysis because you can code out actual scripts that deliver more complex insights, which changes the potential questions enterprises can afford to ask and dig into, or build new tools to answer specific questions. For example, because the "cost" of a question moved from 2 weeks of 2-3 staff members time to 2 minutes of one person prompting, the nature of the options available to the enterprise change:
Analyzing soil testing and tissue testing data in aggregate to assess proprietary product build out, inventory assessments, training initiatives or marketing efforts.
Running complex if/then scenario’s for rebate programs and product ordering, focused on maximizing rebates. There is the potential to create your own program calculator and tool as well.
Churn prediction, customer segmentation, complex personalized marketing efforts, or software specific needs for local customers and more.
I think there is incredible opportunity to lean in. Just this week I was having a conversation with my friend Lee Vetsch of Ospraie Ventures and we discussed the opportunities for ag retail explicitly to have someone become the go-to person for building out capabilities leveraging tools like Claude Code.
I believe the winners will be companies that build the muscle of day-to-day use, early enough for the gains to compound. While ag is still very much a physical business, there is a significant enough portion that is administrative and knowledge based that brings opportunities for agribusinesses to find an edge.