From AI to Agentic to Autonomous Supply Chain Planning: Four Expert Perspectives

We recently hosted a Deep Conversation on the path from AI-assisted planning to agentic and increasingly autonomous supply chain planning. It brought together four supply chain leaders, each approaching the shift from a different angle.

Alex Koshulko (Streamline) explored autonomy and context, while Adam Basson (FlexChain) examined the changing economics of value creation. Ray Li (NeoEdge) brought a mid-market perspective, and Mauricio Dezen (Logica OPS) showed how better planning translates into measurable financial results. 

Here’s a recap of the discussion and the key insights each expert shared.

Alex Koshulko: Context is still the bottleneck to autonomous planning

Alex Koshulko sees autonomous supply chain planning as a matter of when, not if. Supply chain teams have adopted each major wave of AI development, from assistants to agents, and autonomous execution is the next stage.

The biggest obstacle is giving AI enough context to make the right decisions. At Streamline, that context comes from ERP data, user input, and simulation results, but each source still leaves gaps. Human input remains essential because people can add business knowledge and nuance that AI does not yet capture on its own.

Turning those inputs into reliable decision support is also a knowledge-management challenge. Companies need ways to bring fragmented expertise into one place and connect technical teams with the people who understand the business on the ground.

That keeps people central to setting the goals and constraints that guide autonomous decisions. For Alex, one of the most important safeguards is making sure AI understands the financial consequences of each decision before it acts.

“We must make sure that AI understands the financial consequences of any decisions that it makes. And if we succeed in making it understand this, it will be much safer to let AI do anything autonomously.”

Autonomy will therefore come in stages. Alex expects planning to move from 90% autonomous to 95%, 98%, and beyond as more agents are added.

Adam Basson: Why operational improvement matters more now

Adam Basson showed why better planning now matters more financially. From a private equity perspective, the economics of value creation have changed significantly over the past few years.

Between 2010 and 2021, private equity deals could often be financed with 60–70% low-cost debt. A company bought at an 8x EBITDA multiple could later sell at 11x or 12x, while low interest costs made it easier to pay down debt during the holding period. Under those conditions, roughly 5% annual EBITDA growth could be enough to reach a 2x to 3x return.

Today, interest rates are around 4–5%, while borrowing costs can reach 8–9%. Reaching the same return can now require roughly 12% in operational cost reduction or revenue growth instead of the 5% previously needed. In Adam’s words, “12 is the new 5.”

That leaves less room for quick fixes such as headcount freezes, software license audits, or minor procurement renegotiations. Operational savings have a particularly direct effect: every dollar taken out of costs flows directly into EBITDA, while only a share of each additional revenue dollar reaches the bottom line.

That makes supply chain transformation a much more important driver of enterprise value. But warehouse automation, freight renegotiation, network redesign, and similar initiatives depend on understanding demand and the inventory footprint first.

“If you don’t have the right understanding of your inventory and how much you’re actually going to sell as a business, all you’re doing is figuring out a way to lose money more efficiently.”

For Adam, demand forecasting and supply planning therefore come before broader supply chain improvements. They give companies the visibility needed to decide where broader operational improvements will actually create value.

Ray Li: Where mid-market supply chain planning breaks down

Ray Li drew on his work with small and mid-cap companies that had outgrown their existing planning processes. The symptoms differ by industry, but the underlying problem is similar: manual workflows, disconnected data, and homegrown systems become harder to manage as the business grows.

He illustrated that with three recent client examples. 

• An industrial manufacturer sourcing globally faced tariffs, supply disruption, and the challenges of a China Plus strategy. The company also struggled to manage part numbers and metadata consistently across a complex supply chain. 

• A fast-growing personal care chain relied on homegrown forecasting and procurement processes and carried around 60 days of inventory, even though supplier lead times to individual stores were only two to three days. The extra stock tied up capital without eliminating stockouts or improving profitability.

• For a casual dining chain, the challenge was different. The business had to plan around ingredients with limited shelf life, while its own ABC priorities did not always align with those of its distributor, which managed inventory across a much broader customer portfolio.

Across these companies, Ray saw the same need: a practical way to connect data, workflows, and guided decision-making without introducing an enterprise system that is too expensive or overengineered. Mid-market businesses are looking for solutions that are affordable, quick to implement, easy to integrate, and capable of delivering a fast payback.

In the fireside discussion, Ray also outlined where he would start. First, connect data across sales, supply chain, procurement, and finance. Then improve demand forecasting, where the data is often more standardized, and use that foundation to make better inventory decisions.

Mauricio Dezen: What better planning looks like in financial terms

Mauricio Dezen focused on the financial impact of more responsive supply chain planning. In his experience, the less visibility and reaction time a company has, the more inventory buffers it needs to protect the business.

AI can reduce the need for those buffers by recalibrating demand and safety stock sale by sale, location by location, and SKU by SKU as new data comes in. Mauricio described forecasting as becoming more like a “cockpit of calibration” than a periodic exercise the entire company depends on.

The same logic applies to inventory optimization. Companies can apply different replenishment strategies by SKU and location, while ABC classifications adjust as sales patterns change and items move between categories.

The financial results can be significant. One mid-market company had budgeted around $500,000 per month to transfer inventory between East and West Coast warehouses because stock was frequently in the wrong locations. That budget later fell below $20,000 per month.

Another fast-growing company started with an inventory-to-revenue ratio of 25%, meaning it held 25 cents of inventory for every dollar of revenue. Within a few months, that ratio fell to 13%, freeing around $80 million in cash. For Mauricio, this is where supply chain optimization starts to fund growth.

Mauricio also shared the example of a $50 million mid-market company that reduced inventory from 30 weeks to 20 and then to 10. The change released $18 million in cash, while the company canceled 80% of its purchase orders within six months.

“That amount of free cash flow means you don’t have to pay interest to the banks. You’re self-funding, because the money is already there, it’s just misallocated.”

For Mauricio, supply chain optimization can become a source of funding for growth. He has also seen companies use better demand visibility to give suppliers a clearer view of expected orders. This can lead to better commercial terms and more effective planning on both sides.

Top insights 

• Start with the business problem, not AI. Define the operational or financial outcome first, then select the technology and scope around it.

• Treat adoption as a human-led process. AI still needs business logic, calibration, goals, constraints, and oversight, especially as automated decisions begin to scale.

• Don’t assume a successful pilot will scale as-is. Business units and operating environments may require their own training, rules, and context.

• Bridge domain and technical expertise. AI teams need access to the knowledge held by supply chain experts, while subject-matter experts need enough understanding of AI to guide how it is used.

• Build trust through early wins. Smaller, measurable implementations can prove value, strengthen leadership support, and create confidence before broader adoption.

• Look beyond efficiency. Better planning can improve inventory performance, working capital, supplier predictability, commercial terms, and decision-making across the business.

Progress toward autonomous planning depends on the combination of better technology, better data, shared business knowledge, and disciplined adoption.