Kimaru AI

Supply chain decision intelligence — a causal AI platform that simulates physical constraints, runs scenario models, and delivers mathematically proven recommendations to supply chain planners

Most supply chains still run on spreadsheets. Planners export data from the ERP, pull reports from the WMS, reconcile POS figures in Excel, paste everything into pivot tables, and then make multi-million-dollar inventory, pricing, and routing decisions based on what happened last week — hoping the next week resembles it. It rarely does. Kimaru AI is building the decision layer that sits above these systems and does what they cannot: simulate thousands of scenarios in real time, account for physical constraints rather than statistical averages, and deliver concrete, auditable recommendations that planners can review and execute — without replacing the humans who understand the business.

The problem: systems of record that are not systems of action

Enterprise supply chain software has a structural blind spot. ERP, WMS, and SCM platforms are excellent at recording what happened: what was ordered, what was shipped, what was received, what was sold. They are poor at answering the question that matters operationally: given everything that is happening right now — a supplier delay, a weather event, a demand spike, a tariff change — what should we do next, and by when?

The gap between "what happened" and "what to do" is currently filled by experienced planners using spreadsheets, tribal knowledge, and judgment built over years of working a specific category or supply network. That works when the world is predictable. When disruptions arrive — and in the current environment they arrive continuously, from tariff volatility to port congestion to raw material price spikes — the cognitive load on planners exceeds what any individual can process at the speed the business requires. The consequence is reactive firefighting: expensive air freight escalations, stockouts that cost revenue, overstocks that require margin-destroying markdowns, and waste that compounds at every tier of the network.

Kimaru frames this as a $1.6 trillion opportunity loss embedded in global supply chains — the accumulated cost of decisions made with insufficient information, too slowly, based on averages that don't reflect current physical reality.

The architecture: World Models, not chatbots

Kimaru's central product argument is a direct challenge to the generative AI narrative that has dominated enterprise software discussions since 2022. Large language models, the company argues, are probabilistic systems — they generate responses based on statistical patterns in training data. They do not know whether your warehouse is full or empty, what your current machine availability looks like, or whether a supplier delay three tiers back in your BOM will ripple into a production stoppage in six days. Asking an LLM to make supply chain decisions is, in Evan Burkosky's framing, the wrong tool for the job: language models guess based on probability, while operations require simulation grounded in physical fact.

Kimaru's alternative is a Decision Digital Twin: a live model of the client's specific supply chain, built from data ingested from their existing ERP, WMS, TMS, POS, and spreadsheet systems, and updated continuously with external signals — weather, shipping disruptions, tariff changes, currency movements, commodity prices. The Causal AI engine then simulates thousands of potential scenarios against this model, evaluates each against the client's actual physical constraints (machine slot availability, truck capacity, raw material yield, delivery SLAs), and ranks the available decision options by expected outcome. The result is not a prediction — it is a prioritised set of executable recommendations, each with an audit trail explaining why it was generated.

The human-in-the-loop design is deliberate and important. Planners review and approve — or adjust — each recommendation before it is executed. Kimaru then learns from both approved and adjusted decisions, improving the quality of future recommendations over time. The Decision Tracker records every decision rationale for compliance, governance, and audit purposes — which matters significantly in regulated industries and in Japan's corporate culture, where the audit trail for major decisions (the ringi-sho process) is both a cultural expectation and a formal requirement.

Why Japan is the right starting market for this product

Japan's corporate sector faces a compounding challenge that makes it simultaneously a difficult and an urgently receptive market for supply chain AI. The country's manufacturing and retail sectors are among the most sophisticated in the world — precise process documentation, high quality standards, deep institutional knowledge embedded in experienced workers. But that same workforce is ageing and shrinking. Japan's "2025 Digital Cliff" — the moment at which a significant cohort of experienced operational staff began reaching retirement age — created a structural talent gap in supply chain planning roles that cannot be filled simply by hiring replacements who don't yet have the institutional knowledge.

The opportunity Kimaru is positioned to address is therefore not just efficiency improvement — it is knowledge preservation and operational continuity. A Decision Digital Twin that captures the decision-making logic of experienced planners, learns from their adjustments and corrections, and encodes their judgment into a system that can then support less experienced successors is a more compelling value proposition than "our AI is more accurate than your spreadsheet." Japan needs tools that can transfer expertise at scale, and Kimaru's continuous learning architecture is precisely that.

There is also a product-fit dimension to Japan's rule-based corporate culture. Japan's management systems — the ringi-sho approval process, nemawashi alignment procedures, the emphasis on consensus and documented rationale — are often cited as obstacles to AI adoption because they require explainability and accountability at every decision step. Kimaru's architecture, which generates auditable recommendations with causal reasoning and records every decision for governance review, maps directly onto these cultural requirements. An AI system that can explain its recommendation in terms of physical constraints and trade-off analysis, and that keeps humans in control at the approval step, is far more adoptable in a Japanese enterprise than a black-box model that produces outputs without traceable reasoning.

The founders: 22 years of Japan enterprise experience meets 24 years of Panasonic R&D

Evan Burkosky's path to Kimaru is the product of two decades of learning precisely the ecosystem he is now building for. He arrived in Japan with ambitions in seafood imports, then navigated into consulting, marketing, and digital transformation — progressing through roles at J. Walter Thompson, e-Agency Japan, Dynamic Yield (where he built Japan go-to-market for a real-time personalization and product recommendation platform), and Meltwater Japan, before leading enterprise analytics at TechnoPro. The pattern across these roles is consistent: each one involved helping Japanese enterprises translate data signals into operational action, and each one revealed the same structural gap — the absence of a decision layer between insight and execution.

That background produces a founder who speaks Japanese, sells enterprise software in Japanese, understands the ringi-sho process from the inside, and has built the relationships and cultural fluency that foreign enterprise software founders in Japan typically spend their first three to five years acquiring — if they acquire it at all. For a product whose value proposition depends on enterprise trust and process integration, that is not a minor advantage.

CTO Dr. Hareesh Nambiar brings the technical depth that the platform architecture requires. Twenty-four years at Panasonic Japan Research and Development, a PhD from EPFL in the creation of digital humans, and 15 patents in autonomous and intelligent agents is an unusually specific set of credentials for the problem Kimaru is solving. The academic and research background in how machines model and simulate human decision environments maps directly onto the Decision Digital Twin approach — this is not a founder who has read about causal AI and decided to build a supply chain product. It is a researcher who spent two decades studying autonomous agent behaviour and has applied that expertise to a tractable commercial problem.

Alchemist validation and the path forward

Kimaru's selection for the inaugural Alchemist Japan accelerator — a program created in partnership with JETRO, the Tokyo Metropolitan Government, and Mitsubishi Estate to help B2B startups expand globally — and its subsequent admission to Alchemist's main US program (Class 40) represent meaningful third-party validation of the thesis. Alchemist focuses exclusively on enterprise B2B and has a strong track record with supply chain and operations software; selection into Class 40 after the Japan cohort signals that the programme's network saw enough in the product and team to continue backing the company in its US expansion phase.

The go-to-market architecture is designed for speed of initial validation: plug into the stack the customer already uses (ERP, WMS, TMS, spreadsheets), rather than requiring a rip-and-replace commitment. That lowers the procurement barrier for a proof-of-concept engagement significantly, which is critical in an enterprise sales environment where the full sales cycle can otherwise extend to twelve months or more. The Decision Tracker provides the measurable ROI evidence that justifies expansion beyond the initial pilot — cost recovery from improved decision accuracy is quantified and attributed directly to Kimaru's recommendations, giving the champion inside the customer a concrete business case for continued investment.

What would have to be true for this not to work

This is an early-stage investment in a company that is still in the proof-of-concept phase. The claimed outcome metrics — 20–30% inventory turnover improvement, 40% fewer stockouts — are targets grounded in the team's domain knowledge rather than validated across a broad customer base. The path from a compelling demo and a strong accelerator cohort to a referenceable enterprise customer roster requires executing a sales cycle that is inherently long, politically complex, and dependent on champions who can navigate internal approval processes.

The competitive landscape is also substantial. Blue Yonder, Aera Technology, and o9 Solutions are all well-funded incumbents in the supply chain decision intelligence category. Kimaru's response — that it is a lightweight augmentation layer rather than a replacement for existing systems, faster to implement and tuned to last-mile execution — is a plausible differentiation strategy, but it requires the enterprise market to accept the augmentation framing rather than defaulting to the procurement path of consolidating onto a single platform vendor.

We backed Kimaru because the problem is real, expensive, and structurally underserved in the mid-market. The team has the specific credentials the solution requires. Alchemist has validated the enterprise thesis. And Japan's workforce dynamics create a demand environment for decision augmentation tools that is more urgent than almost anywhere else in the world. At the stage we invested, we are backing the team and the conviction as much as the current product — which is precisely when early-stage investors are supposed to act.