Interview with Evan Burkosky, CEO and Co-Founder @ kimaru.ai

Mr. Evan, can you tell us about your professional background and your current role at kimaru.ai?

Over the past two decades I’ve led analytics and go-to-market teams for enterprise software across Japan and APAC. My career has evolved together with technology, as we moved from early analytics platforms to optimization, recommendation, personalization, and now prescriptive recommendations with Agentic AI. I co-founded digital marketing agency konnichiwa-japan in 2007, focused on multilingual web development, analytics, and optimization. I ran that until 2013 when I joined e-Agency to help build the account development function that helped Fortune-500 companies operationalize GA360 and Optimizely experimentation. As Country Manager for recommendation and personalization platform Dynamic Yield I built the GTM function and led the region to a #1 global sales quarter. At Meltwater Japan, I rebuilt a disrupted sales org and returned it to growth. Most recently, as Head of Data Analytics at TechnoPro, I stood up an executive analytics capability to inform the solutions consulting business. In 2024 I co-founded and serve as CEO of Kimaru AI, where we turn supply-chain data into actionable recommendations on the right product in the right place at the right time.

Evan, can you walk us through your journey, from the early experiences that shaped your approach to operations and AI, to how they influenced your focus on decision intelligence in supply chains?

At e-Agency I led large-scale rollouts of Google Analytics 360 and Optimizely, which taught me that measurement only matters when it triggers a concrete operational step. At Dynamic Yield I built the Japan go-to-market for a real-time personalization and product recommendation platform, and saw that channel gains don’t automatically flow upstream to replenishment, pricing, or markdowns. At Meltwater, a media-intelligence and social-listening company that uses AI and NLP to turn news and social data into structured signals like mentions, reach, and sentiment, I helped rebuilt the Japan sales organization to help clients translate those external signals into predictable actions and results. At TechnoPro I stood up an executive analytics function that delivers decision advisories tied to commercial teams and regular business cadences. Those experiences led me to decision intelligence for supply chains: a decision layer that reads context from ERP, WMS, and POS (and relevant external signals), produces clear, prioritized recommendations with reasons, and writes approved decisions back into the systems operators already use.

Before founding Kimaru.ai, what were some operational or data challenges you observed that made you realize traditional tools like Excel or static ERP systems weren’t enough? Any early lessons or failures that shaped your approach?

• Latency: Weekly spreadsheet cycles trail daily demand and perishability windows.
• Silos: Merchandising, supply, and marketing optimize locally; wins don’t propagate.
• Systems of record ≠ systems of action: ERPs excel at “what happened,” not “what should we do next under current constraints.”

The remedy is a decision layer that proposes ranked actions and writes them back into the systems teams already trust.

Kimaru.ai was created to uncover hidden profits and reduce operational risks without replacing existing tools. Can you share what inspired its founding, and how the early development phase addressed real-world supply chain pain points?

Three persistent blockers:

• Latency: Weekly spreadsheet cycles trail daily demand and perishability windows.
• Silos: Merchandising, supply, and marketing optimized locally; wins didn’t propagate system-wide.
• Systems of record ≠ systems of action: ERPs tell you what happened, not the next best action under current constraints.
  Lesson: Dashboards inform, but operators need ranked, explainable actions that write back into the tools they already trust—otherwise improvements stall.

Leading teams across the U.S., Japan, and APAC comes with its own challenges. How do you ensure alignment, motivation, and effective execution across such diverse teams?

We operate asynchronous-first across Chennai, Tokyo, Austin, and San Francisco. Day-to-day work runs on Slack for communication, Google Workspace for documents and decisions, Asana for planning and tracking, and GitHub for code and reviews. We use light, small-team Google Meetings when helpful and avoid large all-hands. Alignment comes from writing things down: goals, scopes, and updates are captured in Google Docs and linked to the Asana plan so everyone sees the same context regardless of time zone. Execution is kept predictable by keeping work visible in Asana and code reviews in GitHub, and by expecting concise written updates instead of meetings. Culturally, we hire and operate for self-directed management and strict time efficiency, so teams can move quickly without waiting on synchronous touchpoints.

Kimaru.ai promises measurable outcomes like 20–30% improved inventory turnover and 40% fewer stockouts. How did you define this mission of operator-focused decision intelligence, and how does it distinguish your platform from global competitors like Aera or Blue Yonder?

We defined the mission by working backwards from operator KPIs – inventory turns, sell-through, waste and service levels – and built a decision layer that shortens the time between insight and action (markdown timing, store replenishment, routing).

Kimaru is AI-native and agentic: lightweight agents harmonize ERP/POS/WMS and spreadsheet data, run continuous “what-if” decision models, and learn from decision outcomes via a Decision Tracker so recommendations improve over time. That focus on fast, measurable decisioning is the basis for our 20-30% inventory-turnover and 40% fewer-stockouts claims.

Compared with Aera: Aera is a broad, mature Decision-Intelligence Platform with extensive decision modelling, governance and enterprise “crawl” capabilities. Kimaru deliberately trades breadth for speed and operator impact – we are smaller, faster to implement, and tuned to the last mile of execution so managers can act immediately.

Compared with Blue Yonder: we treat Blue Yonder as a systems partner rather than a replacement – Kimaru augments SCM/ERP/WMS/CRM/POS by ingesting minimal context, surfacing ranked, explainable actions, and writing approved decisions back into the customer stack, enabling low-friction pilots and rapid time-to-value.

Kimaru.ai is known for its seamless integration with existing tools like ERP, WMS, and Excel, delivering actionable, real-time recommendations. Beyond integration, what technical or strategic breakthrough allowed you to deliver this speed and intelligence? Which product or module are you personally most proud of?

The product I’m most proud of is our Demand Forecasting module. It is the hardest problem to solve well, and we built it on top of the Decision Digital Twin and modular AI agents so forecasts are produced in the same decision context managers use for actions and constraints, not as isolated numbers.

We engineered continuous learning into the module – AutoML, user feedback loops, and federated learning so the model improves every cycle and adapts to customer-specific patterns.

We also pair forecasting with a Decision Tracker that measures decision outcomes and quantifies value, and the whole flow is integrated with ERP/WMS/POS so forecasts drive ranked, explainable recommendations and writebacks.

That combination – hard forecasting models, simulation in the digital twin, continuous learning, outcome measurement, and pragmatic integration – makes Demand Forecasting the foundation for the measurable improvements operators rely on.

AI adoption in Japanese businesses is increasing, with over 31% of professionals using generative AI in 2025. Can you share examples of how Kimaru.ai has transformed decision-making for Japanese clients, such as in retail or logistics, and the measurable outcomes that surprised you the most? How have these experiences influenced the way you approach product development and client engagement?

In Japan we set pilots against the benchmarks McKinsey reports for AI-enabled supply chains. McKinsey finds that early adopters improve logistics costs by about 15 percent, inventory levels by about 35 percent, and service levels by about 65 percent. Within that frame, our retail and logistics clients used in-system recommendations to make price, replenishment, and routing calls more frequently, which raised on-shelf availability in promoted periods and improved on-time performance without adding headcount. We also focused on demand forecasting quality, since McKinsey shows AI can reduce forecasting error by roughly 30 to 50 percent, and we saw material error reductions when forecasts ran inside our decision context instead of as standalone reports. These outcomes influenced our roadmap: keep recommendations explainable, let teams test changes with a decision digital twin, write approved actions back to ERP, WMS, and POS, and expand scope only after stable results against baseline. The goal is practical gains that sit within McKinsey’s ranges while building a repeatable path from pilot to value.

With Japan’s AI market projected to grow at 33% CAGR by 2033, and AI-driven supply chains becoming central to efficiency and competitiveness, which emerging trends or capabilities do you see as most critical for companies to adopt, and how is Kimaru.ai preparing operators to respond to these changes?

Japan is about 30 percent of our addressable market, and while we are a global company with offices in the U.S., Japan, and India, the combination of a shrinking workforce and long-running productivity stagnation makes rapid AI adoption especially important in Japan. The capabilities that matter most are decision agents that turn data into ranked actions, human-in-the-loop controls so managers can trust and adjust those actions, and low-friction integration to ERP, WMS and POS so value shows up inside tools teams already use. Continuous learning is equally important. Forecasts and recommendations need to get better with every cycle, without exposing sensitive data, which is why we invest in techniques that keep learning close to the customer’s systems.

Kimaru prepares operators by pairing technology with delivery. Forward-deployed engineers build connectors, stand up pilots, and codify local playbooks so customers do not need large internal engineering teams. Our Decision Digital Twin provides context for decisions like demand forecasting, replenishment and markdown timing, so recommendations reflect real constraints and are easy to approve. A Decision Tracker measures outcomes and feeds improvement back into the system, which keeps adoption focused on results rather than theory. The goal is simple. Help Japanese operators move faster, make better calls inside their existing stack, and translate AI into measurable gains in inventory turns, stock availability and service levels.

For young professionals aspiring to work in AI, decision intelligence, or supply chain innovation, what advice would you give in terms of mindset, skills, and navigating a rapidly evolving, data-driven industry?

Learn how to work with AI agents and decision frameworks – not just to automate tasks, but to drive real business outcomes. This starts with understanding how agents plan, call APIs, manage memory, and handle edge cases. But just as critical is the methodology behind those agents. That’s where Decision Intelligence comes in. Originally formalized by Dr. Lorien Pratt, it provides a structured approach to turning data into decisions and outcomes. At Kimaru, we extend this with what we call networked decision digital twins – interconnected, lightweight models that simulate the impact of choices across supply chain functions like forecasting, allocation, and replenishment. Each twin is tied to a measurable business goal, and agents use them to surface ranked recommendations with clear rationale and approval flows. If you’re starting out, focus less on model-building and more on mapping real-world decision points, understanding business constraints, and designing AI that fits within those flows. That’s how you create value – not just predictions.