Upstream Ag Professional - June 14th 2026

Index

Thank you for being an Upstream Ag Professional member!

No audio edition this week due to connectivity outages in my home office.

One of my problems with AI has been the constant need for re-prompting. Each time I started a new research initiative, I felt like I was starting from scratch.

So I was spending a large chunk of time re-delivering context: research or business goals, constraints, the target outcome, or all the information I wanted it to take into account at any given time. I would type all of it in, eventually get to good answer, but then I still had two problems:

I took these problems to my friend Patrick Walther, Director of AI Product at AgVend, who has helped built AI systems at AgVend and VisorPRO, for help and said he had built out a system to overcome those problems, plus it was able to get better over time with skills and “put me into the top 1% of AI users”. He called it a Personal OS and he helped me build one for myself.

The process took a detailed approach to building the fundamental systems on my computer, along with a detailed process to understand me, my goals, ambitions, problems, and anything I’ve ever written and how I think about the world into an integrated folder of markdown files that are always connected to the model of my choice (I use AI via Cursor). It enabled AI understand me continuously instead of starting from scratch.

As a non-technical person, the thought of it initiall intimidated me, however, the process was straight forward and the outcome for me has been next level — Less prompting, the system makes suggestions, and takes action based on my values, the business context behind Upstream, my goals for the year, and in the context of everything else going on that I have given it access to. It briefs me the way a chief of staff would, before I ask — and reallocates several hours of administrative burden into higher quality thinking time.

Patrick and I want to show you how it works, and then point you to two ways to do it yourself.

Register for the webinar here.

Get for more information here.

There is a ton of talk about AI strategy at the enterprise and company level, however, the starting point for me has been building the operating system to help me understand how to get AI platforms working for me, which has given me more thinking time back, plus helped me better understand how these systems can apply to larger enterprise strategy.

I think a Personal OS is the perfect starting point for those wanting to improve their output with AI systems.

Index:

For decades, the primary mechanism for crop protection has remained unchanged: broadcast applications moving chemistry from a tank through the nozzle. But despite 60 years of incremental improvements in machinery and chemistry, the systems remain ripe for improvement. Data from MIT and the BioScience Journal suggests that anywhere from 30% to 75% of applied product never hits its intended target. To improve these numbers, the industry cannot rely on a single "silver bullet." Instead, the future of crop protection lies in stacking software-enabled decision support, computer vision, and droplet physics on top of traditional nozzles and formulations, the industry is moving from field-level management to precise, plant-level execution.

In the full breakdown, we dive into:

The Evolution of Spraying - How various technologies came to the market over the last 60+ years to improve spray application outcomes across optimizing agronomic, economic, and environmental outcomes.

The Tech Stack Framework - Breaking down the distinct layers of modern application technology and how they interact and add value.

Look at Companies and Their Capabilities - A look at how companies like AgZen, DriftSense, MagrowTec, John Deere and others add value at their layer of the tech stack.

Last week, I shared Building a Go-to-Market Strategy in Agriculture. It generated a lot of feedback about direct-to-farmer dynamics and how to build a distribution approach that works.

However, a few comments focused on a different question came in: once you have the distribution channel, how do you get the most out of it?

An individual from a biostimulant company had a question many are challenged with: But, how do you actually get a distribution channel to actually prioritize your products?

It’s a great call out cause it can be a conundrum.

Distribution agreements tend to get celebrated, but they are just the beginning. It reminds me of a venture capital raise — the raise gets celebrated, but the real work begins once the money is in the bank. Distribution is the same, once the agreement is signed, you're at the beginning of the harder work: generating demand, training sales people, staying top of mind, supporting end customers, and earning renewed interest year after year.

This piece builds on last week's article and looks at execution after you have secured market access. I don't pretend to have all the answers, but below are a several considerations that should begin to improve your odds.

A retailer can carry your product, like your brand, give you warehouse space and time with their people, even make you the exclusive provider in a segment, and your product can still sit idle in the warehouse.

You have to activate the channel sales rep by sales rep, grower by grower, season after season.

Check out the full overview that provides a framework for helping derive better results with channel partners.

Below is coverage of two raises this week: Leaf Agriculture and SWARM Engineering. They are both related to AI and analytical infrastructure across the ag industry. One notable tie between them is the similar investors, including S2G, Radicle Growth and Trailhead Capital in SWARM, all of which have previously invested in Leaf. On top, the strategic aspect: AgRogue, a fund invested in by Land O’Lakes, invested in SWARM, and Leaps by Bayer lead the Leaf round. I think it illustrates the interest in AI and analytical infrastructure for agriculture, along with the acknowledgement that ag specific experience required to make AI successful in agriculture. In my conversation with Leaf CEO, Bailey Stockdale, he shared how valuable it is for their business in having deep agriculture experience to deliver to their ag based customers.

Leaf Agriculture announced this week a $13 million Series B co-led by Leaps by Bayer alongside a group of industry investors. The round brings Leaf's total capital raised to roughly $29 million since 2021. Existing investors also participated including S2G Ventures Trailhead Capital, SP Ventures, Spero Ventures and Cultivian.

For those new to Leaf, the company has been workng to build the data infrastructure layer for agriculture. Ag data is difficult to work with and many companies in ag do not have baseline skills to structure and leverage data.

The press release compares Leaf to what Stripe is to payments or Plaid is to banking — they sit invisibly behind a lot of the digital experiences that retailers, input companies, crop insurers, and others are building on top of farm data. They process data covering more than 20% of global crop acres annually and power some of the most recognizable digital platforms in the industry, including, but not limited to, Bayer's FieldView, Syngenta's Cropwise, and BASF's xarvio.

For most of its history, the company was heavily infrastructure focused, cleaning, normalizing, and serving data so engineering and product teams at large ag companies could build on top of it. That gave them a foundation. But CEO Bailey Stockdale told me about a pattern they kept seeing where they’d work with IT and product teams, but then when they’d connect with the head sales at their partners for example, they would run into a situation where they’d seen the data, but hadn’t found a way to leverage it.

Moving from data to driving decisions is what Leaf has been working towards with them leaning much harder into dashboards, mapping tools, and consultative analytics work that gets closer to the actual business problem.

As Bailey put it to me, "We are their edge. Our goal and our job is to make sure that our customers win in their very competitive environments."

That comment resonated with me and I suspect it will resonate with many of you — I have frequently talked about finding information advantages in ag retail to be successful, whether it is for your sales people or operational people. Leaf helps enable that advantage by delivering insight from previously disparate data sources.

The repositioning to deliver actionable insights that deliver an edge also comes through in how Leaf has rebuilt internally. Bailey said Leaf has spent the last 12 months rebuilding the underlying systems because, in his words, "it was just too slow for agents."

Bailey’s notably avoids over indexing on AI, instead saying that he wants to ensure Leaf is always primarily focused on solving a core business problem for each customer, and the tech will follow. That said, he sees Leaf as the critical piece of bringing the value of AI to agriculture, and shared how Leaf thinks about earning the right with each customer to not only be a key piece of their AI stack, but also be their AI partner.

Check out the full article including use cases, business model and more in the link above.

SWARM Engineering, a decision intelligence company for ag, food and manufacturing, raised $10 million in an oversubscribed Series A. The round was co-led by S2G Investments and AgRogue Growth Partners, with participation from Radicle Growth, Grit Road Partners, Middleland Capital, Open Prairie, Serra Ventures, and Trailhead Capital.

The press release states that the capital will be used for accelerating SWARM’s operational AI roadmap, scaling go-to-market, and deepening integrations with ERP and supply chain systems — the latter I believe is important since their limitation is going to be depth and quality of data and they need data from various systems (more on that later).

The list of investors is impressive with S2G, Radicle, Open Prairie, Trailhead, and the rest being some of the most established names in agtech investing. Given that many ag specific investors converged on one company, it seemingly signals that the underlying thesis behind SWARM is widely held across the investor base.

SWARM positions itself as operational AI purpose-built for agrifood and manufacturing. The platform combines AI agents and optimization algorithms into a single system that ingests real-time data, runs scenarios across thousands of variables in minutes, and surfaces recommended decisions for supply chain, workforce, and logistics. I highlighted this as an optimal starting point for agribusinesses using AI back in 2023.

The leadership team of SWARM comes out of Microsoft, Palantir, Google, and UiPath, which gives technical credibility. One thing I wonder is if they will take a page out of the Palantir playbook and embed “forward deployed engineers” into businesses — this is maybe more of a challenge for verticalized players because there is a lot of overlap / competition, but will be interesting to see nonetheless.

The differentiator according to CEO Shail Khiyara in the release is that SWARM is built on the “operational ontology” of these industries, rather than a generic AI platform that learns the business over time.

Ontology is important in the world of AI.

Ontology, in the AI and computer science sense, is a formal specification of concepts within a domain and the relationships between them. It defines a shared vocabulary, like the categories of entities, their properties, and the rules connecting them, structured so that AI can reason effectively over the data.

For agriculture, that means the system already understands the structure of the industry, such as how a field belongs to a grower, that a grower buys from a retailer, that a retailer is supplied by a distribution center, that demand at the retailer is shaped by weather, disease pressure, planting intent, and application windows, and that each of those variables has its own set of dependencies. The ontology is encoded with how the industry makes decisions before the platform ever sees a customer’s data. It is sort of like context.

Generic large language models have to learn this structure through usage, often occurring slowly and incompletely while limiting value creation in the interim. A platform that starts with the ontology baked in should compress time-to-value and reduce the volume of data and trial-and-error required to produce a useful recommendation, and reduce poor recommendations.

Ontology has to be built and maintained by people who genuinely understand the industry, which is one place where SWARM is trying to emphasize their differentiation.

For a look at why the AgRogue connection is compelling with SWARM, check out the full article linked in the heading.

This is an interesting perspective from Erik Benson surrounding “Right to Intelligence.”

He says the following (note: the post is longer, I selected specific lines that I deemed most relevant):

I think it’s important to first consider the chain of data to value, because to me iron doesn’t inherently generate intelligence — the sensors associated with it generate data and that data is collected, structured and analyzed, which is where intelligence comes in. I think of it something like this: