Why We Invested in Myca
Myca is building an AI-native operating system for agribusiness, a “supervisor in every field rep’s pocket” that turns the voice notes, photos and WhatsApp chats where agribusiness actually runs into validated, auditable workflows. The promise is not better reporting. It is fewer layers of human middleware needed to run the same business, with tighter control.
“Fifteen steps. Five teams. Three layers. Three weeks. All to answer one question.”
Here’s that question. A mid-sized agribusiness leader asks it every month:
“We need to win on chilli margins next month. What will flows and prices look like?”
The data exists. The ERP has it, the CRM has it, a spreadsheet somewhere has it, and a field rep sent a photo of it on WhatsApp last Tuesday. But it is scattered, stale and not trusted, so a simple question becomes a three-week relay race through human middleware. By the time the answer lands, the market has moved on.
This is not a story about slow people. It is a story about an industry whose software records everything and executes nothing. It is also the problem three founders have spent the better part of a decade inside, and the one they have now set out to solve with Myca. We are excited to back Arnav, Seamus and Tejas to build the execution layer of agribusiness. Three things drew us in: a problem that quietly drains margin at scale, a product built for the real world rather than the demo, and a founding team that has already scaled this business once.
1) A Tax Nobody Budgets For
Most mid-to-large agribusinesses are not undigitized. They run ERPs, CRMs, scheme engines, inventory tools, a patchwork of field apps and a BI layer. Yet inside the day-to-day operation, work still moves through human middleware. Field reps report through WhatsApp groups, calls, handwritten notes and photos. Supervisors reconcile and interpret it. Mid-office teams clean it again for planning and risk checks. Managers finally decide, on inputs that are late, imperfect and context-poor.
Systems record activity, but they do not drive work. So execution quality depends on people, and growth means adding headcount.
The economics make this painful. A representative harvest procurement business with roughly $100M in topline runs at about 3% to 3.5% margin: ~75% to 80% goes to COGS, ~10% to SG&A and ~2% to technology. In input distribution, aggregators earn ~5% to 7%, and even manufacturers top out around 12% to 15%. In businesses this thin, the founders estimate that 30% to 40% of SG&A is consumed by repetitive coordination: planning, supervising, correcting, validating, reporting, escalating and complying.
Scaling compounds the problem. Going from 100k to 200k tons of procurement typically means opening a branch and hiring layers of supervisors and coordinators, roughly $650k of spend before a single incremental ton is procured. For every ~10 field people, another layer is added to keep the machine running. This is not an inefficiency at the edges. It is the operating model.
Regulation is adding weight. Traceability and sustainability regimes such as EUDR, RSPO and FSMA-style expectations create audit trails, document checks and verification loops that are still human-heavy, while credit and risk decisions remain spreadsheeted and tribal in many organisations.
2) A Deficient Status Quo
The gap has persisted not for lack of software, but because each category of software solves a different problem than the one that matters.
ERPs are excellent systems of record but weak at field-first workflows, multimodal capture and real-time validation. They do not do the work. CRMs manage relationships, but were not built for low-bandwidth environments, voice and image-heavy inputs, or seasonality-driven planning. Point solutions tend to fail in one of four ways: they capture data without improving its quality, demand behaviour change that field teams never adopt, become expensive customised projects on top of ERPs, or stay too horizontal to encode agribusiness context and audit needs.
The result is the Frankenstein stack: ERP, CRM, custom field apps, spreadsheets and WhatsApp. Our conversations surfaced about how much budget this stack already absorbs, and it scales with the size of the organisation. Smaller players spend roughly $50k to $150k on field optimisation and execution tooling. National-level players spend roughly $500k to $750k. Larger players spend in the millions, and multi-country enterprises spend multiple millions. That is the budget pool Myca can consolidate.
Agribusiness does not need another dashboard. It needs a system of execution.
3) From System of Record to System of Execution
Myca sits between the systems of record and the frontline, and it works where field teams already work: WhatsApp, voice and photos.
Picture the loop. A field rep gets a dynamic beat plan, with visits prioritised and prompts on what to push, whether a scheme, a product or credit terms. The rep shares a photo of a blackboard price, an invoice or a quality reading, and Myca validates it in real time. If a number looks wildly off-market, it asks a contextual question before the data enters the system. The supervisor gets an oversight console rather than a report: what is happening, where data is missing, which exceptions need approval and what the system recommends. Planning becomes dynamic, capture is validated at source, decisions are explainable, actions are orchestrated, and governance is native rather than bolted on.
Get Eximius Ventures’s stories in your inbox
Join Medium for free to get updates from this writer.
The product is built as a modular suite on a common core, sequenced around wedges that are urgent, easy to adopt without ripping out systems, measurable in ROI and expansion-friendly:
- The entry module covers beat planning, capture, nudges and supervisor oversight. It sits at the centre of the coordination burden and becomes the data engine for everything else.
- A controls and compliance layer extends Myca into governance, audit trails and verification workflows.
- Once field activity is structured and validated, credit and risk decisions become something the system can inform and govern, moving Myca closer to the flow of money.
- Procurement execution covers planning, quality validation, price decisions, logistics and audit, workflows that are structurally consistent across commodities and geographies.
- Commerce and downstream integration is a longer-term expansion once Myca is the execution layer.
The commercial model follows the same logic. Myca lands with the entry module on a per-user subscription alongside a platform fee for the underlying OS layer, with adoption ramping across a customer’s field force rather than switching on overnight. From there it expands across workflows, users, business units and geographies. Credit and embedded finance add a volume-linked revenue stream, so monetisation scales with the customer’s business and not only with the number of seats. Land where execution pain is most acute, then expand into the rails that touch money and compliance.
What impressed us most is how the system is built. Many products in this category fall apart by being either too LLM-driven, so they hallucinate and cannot be audited, or too rule-driven, so they break against real-world messiness. Myca sits deliberately between the two: a structured understanding of each customer’s business, models used where interpretation is needed and deterministic logic where precision is needed, and a full record of both what the system recommended and what actually happened. That is what makes it trustworthy in regulated and credit-adjacent workflows.
This is a routing and execution system, not a prompt wrapper. And treating WhatsApp as a first-class product surface, rather than a nice integration, is a deliberate bet on adoption. In field-heavy markets, adoption is the real battleground.
4) Founders Who Have Lived Inside the Problem
Myca’s founders are not agri-curious builders. They ran the operation they are now productising, across inputs on credit, harvest procurement and the mid-office layers between ERP records and ground truth. Their roles are also cleanly orthogonal: product and workflow abstraction, systems architecture, and go-to-market. For a company trying to become the execution layer of an industry, that triangle matters as much as pedigree.
Arnav Pandey (CEO) was most recently CPO at Jiva, the Olam-backed agritech venture, where over roughly five years the business scaled from zero to $120M+ in GMV while running a full-stack model: inputs on credit, then harvest collection and sale into mills. He has seen what breaks at scale, why it breaks and what the operational workarounds look like. Before Jiva, he was one of the earliest product managers at Locus (the first PM, later owning solutions), the logistics optimisation company later acquired by Ingka Group, IKEA’s parent. The pattern carries over directly: constraint-heavy workflows, routing and planning under real-world messiness, and the discipline of moving from “models” to “systems people trust.” His earlier engineering stints, including mobile backend work, show up in Myca’s WhatsApp-first, low-friction choices.
Seamus Tardif (CRO) brings the muscle for the central go-to-market risk in agribusiness software: how to sell without becoming a services company. He was a cofounder and head of growth at Jiva, and came to it from BCG X, where he worked on the early venture definition with Olam, built the MVP and saw the full arc from concept to scaled operations. He has also built ventures with partners such as Singapore Airlines and the World Wildlife Fund, and run his own startups earlier. That background shows up in crisp, operator-led narratives that enterprise buyers understand, and in the discipline to separate pilot enthusiasm from a repeatable sales motion. He is the founder most likely to keep Myca thinking in playbooks, ROI framing and expansion paths rather than custom builds.
Tejas Dinkar (CTO) was CTO at Jiva and before that CTO at Quintype, a multi-tenant SaaS platform used by newsrooms across India and the Middle East, a serious reliability and scale environment. Earlier, he was at C42 Engineering (acquired by Gojek) and ThoughtWorks, and he founded Nilenso, a firm known for engineering craft and long-lived, modular software. An “agentic OS” dies quickly if the architecture is shallow, non-auditable or fragile under enterprise constraints. Context profiles, orchestration layers and dual ledgers require someone who has built real systems before, not just shipped prompts.
All three worked together at Jiva, which means the team has already been through one full cycle of building and scaling a field-heavy agribusiness. They know where the pain is because they carried it.
5) Early Customer Signal Points to Pain, Not “AI Curiosity”
Our diligence calls across India and Indonesia, spanning input-heavy and procurement-heavy businesses, told a consistent story. Execution is trapped in manual coordination, and the cost is measurable. Buyers modelled about 10 days of lost time from delayed inventory decisions, a 15% to 20% gap in receivables performance, and a potential gross margin shift of ~15% in a single geography from better retailer prioritisation. Others described price collection that reaches decision-makers with a 2 to 3 hour lag, order conversion stuck at 40% to 45%, and beat planning still run on Excel and Google Forms. One had already paid for a generic CRM and dropped it because the workflow did not fit.
Two things stood out. Budgets exist when value is framed around working capital, conversion, margin or control, and the question buyers ask is not “do we need a CRM” but “what share of the improvement belongs to the software.” And the appetite for a WhatsApp-first, low-friction surface was unprompted and consistent across geographies.
6) What Sets Myca Apart
Plenty of companies will point AI at agribusiness. Three things make Myca different. It is built around execution, not insight: most tools in this space collect data and report on it, while Myca aims to be the place where plans are made, actions are taken and decisions are recorded. It is designed for trust: structured context, deterministic logic alongside models, and an audit trail by default, which is what enterprises in regulated and credit-adjacent workflows actually need. And it meets field teams where they already work, which makes adoption a feature of the product rather than a change-management project.
The prior generation of agri software often grew through program and project budgets, optimising for coverage metrics such as farms digitised or reports delivered. Myca is built the other way around: operator-led, workflow-native, fast to deploy, and measured on direct operational outcomes like conversion lift, fewer missed visits, better credit discipline and less supervision overhead. It does not sell “AI.” It sells execution outcomes.
Agribusiness in the Global South runs on thin margins and heavy coordination. Its systems record activity but do not drive work, and every increment of growth is bought with another layer of supervisors, coordinators and spreadsheets. Myca takes aim at that coordination tax with an execution layer that lives where field teams already work, validates data at the source, and leaves an auditable trail for every decision.
The team is what makes us believe it can be done. Arnav, Seamus and Tejas have already built and scaled a full-stack agribusiness together, and they combine lived operator context with product abstraction, systems depth and go-to-market rigour. The early customer signal is consistent across geographies and business models, and the product is designed to grow from the field into the flow of money and compliance.
We are proud to partner with Myca and look forward to helping them become the system of execution for agribusiness.
This article was produced in collaboration with Preeti Sampat
