Welcome to the 127th edition of Upstream Ag Professional!

Index

1. New Upstream Ag Professional White Paper! AI and Agribusiness Software: From Systems of Record to Systems of Action

2. What Claude Code and CoWork Buzz Means for Agribusiness Professionals

   1. Principled AI Use

3. The Theory vs. Reality of Nitrogen Insurance

4. From Yield to Efficiency: Agtech’s Next Big ROI

5. Inari Lays Off 64 in Indiana

6. Bayer Expands Agricultural Portfolio with Advanced Biostimulants

7. Precision Planting Unveils ArrowTube

8. John Deere Rolls Out See & Spray Gen 2 and Upgrades Sprayer Lineup

9. Satya Nadella on AI’s Business Revolution: What Happens to SaaS

10. Other Interesting Ag Article (13 this week)

This week’s audio edition can be found here and covers the following:

1. What Claude Code and CoWork Buzz Means for Agribusiness Professionals

2. John Deere Rolls Out See & Spray Gen 2 and Upgrades Sprayer Lineup

Thank you for being an Upstream Ag Professional member!

1. AI and Agribusiness Software: From Systems of Record to Systems of Action - Upstream Ag Professional

Over the last couple of months I have been looking at vertical software companies from other industries, like healthcare, services, restaurant management and construction and their approach, offering and vision surrounding artificial intelligence integration into their offerings.

The aim was to get a baseline for what agribusiness software companies might do, and look at some of the products they could offer to ag retails and crop input manufacturers and position themselves moving forward.

This week I am releasing a White Paper exclusively for Upstream Ag Professional members that shares some of the learnings from other industries, what that means for agribusiness software strategy, how AI could impact the offering and where we are headed. At the end I look at some of the capabilities and considerations ag retailers need to be aware of today.

Index of the White Paper

1.  Why AI Matters Now

• Why Vertical Beats Horizontal

• The Agribusiness-Specific Context

2. Lessons from Other Industries

• Procore and the Connected Platform

• Toast and Multi-Product Expansion

• Service Titan and Atlas

3. The Current Agribusiness Software Landscape

• Core Systems of Record (ERPs)

• Engagement and Action Platforms

• Data and Intelligence Providers

• Agronomy and Supply Chain Platforms

4. The Evolution: Systems of Record to Systems of Action

• Phase 1: Systems of Record

•  Phase 2: Systems of Engagement

• Phase 3: Systems of Action

• The Context Graph

5. What This Means for Agribusiness

• For Ag Retailers

6. Interface Concepts for AI in Ag Software

• Design Principles for Ag AI Interfaces

•  Interface Concept

7. Strategic Recommendations

8. Conclusion

A few takeaways

• Systems of record (ERPs like Agvance, Merchant Ag) remain foundational but are insufficient alone. The next competitive advantage comes from systems of action, towards systems of intelligence with embedded AI that execute workflows autonomously.

• AI agents represent a new tool that can handle repetitive cognitive tasks: processing orders, dispersing recommendations, coordinating logistics, and managing customer communications and advertising initiatives.

• The agribusiness software landscape is fragmented across ERPs, agronomic software, CRMs, and various specialized tools. AI can create an coordination layer that connects these systems.

•  Interface design matters enormously. The winning platforms will meet users where they work: in truck cabs, warehouses, field offices, and manufacturing facilities, not necessarily a chat-based interface.

• First-mover advantage is significant but not insurmountable. The key is moving deliberately: mapping workflows, piloting solutions, and building institutional capability. 

The full paper can be downloaded at the link in the heading.

2. What Claude Code and CoWork Buzz Means for Agribusiness Professionals - Upstream Ag Professional

Key Takeaway

• Anthropic’s "Cowork" feature is a shift from conversational chatbots to autonomous agents that can access local files and business software (ERP/CRM) to execute complex, multi-step tasks like inventory management and invoicing. The feature can be valuable for agribusiness professinals and evolution of them could deliver unique efficiencies for agribusiness professionals.

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:

1. Claude is operating like an orchestration or coordination layer. Organizing systems and operations and helping to rethink how businesses are managed.

2. 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.

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.

For more on Claude Code, check out the full link in the heading.

Principled AI Use

As an aside, I have been getting asked a lot on how I use AI, or how I think about using it. I am not the best person to answer any of those questions, but one thing that I have found useful for myself and Upstream is creating principles that drive when and how I use AI.

1. Don’t use AI to confirm your biases without considering alternative views. Stay curious.

2. Don’t replace critical thinking with AI. Writing is thinking.

3. Outputs from AI should never be the final output. Fact-check everything that you discover via AI systems.

4. Prompt the system at the level you expect the output to be delivered.

5. Always consider ways to add incremental context to an AI system.

3. From Yield to Efficiency: Agtech’s Next Big ROI - Granite Creek

Related to the above on AI capabilities becoming an orchestrating layer, I was reading about SeedBox Solution this week in the above article.

Connecting assets, whether seed boxes, or fertilizer bins, anhydrous facilities, or bulk glyphosate systems becomes valuable for integrating data into logistical decision making, inventory planning, servicing etc.

Subscribe to Upstream Ag Professional to read the rest.