Lyric.

It's a Tuesday at 7:14 a.m. somewhere outside Atlanta.

A shipment of cocoa is stuck. Not catastrophically - a port delay, a routing question, the kind of small ripple that, if ignored, will leave a manufacturing line idle and a Friday meeting tense. Twenty years ago, the planner staring at this would pull up a spreadsheet, call two people, and guess. Five years ago, she would have opened a planning system she did not design, written for a problem that does not look like hers, and guessed faster.

This morning, she opens Lyric. The model her team built last quarter - the one that knows their network's quirks the way a long-time mechanic knows an old engine - reroutes, reprices, and surfaces three options before her coffee is cold. She picks one. The cocoa moves.

This is a profile of a company that wants that scene to be unremarkable.

A platform company that thinks of itself, slightly stubbornly, as a tools company.

Lyric is an AI platform for enterprise supply chains. That sentence is correct and almost useless - every company in the category claims it. What sits underneath the marketing is more specific. Lyric Studio is a composable environment where data, decision science libraries, configurable workflows, and accelerated compute are pulled into a single canvas. Inside the canvas, supply chain teams build their own apps. Forecasting. Network design. Inventory policy. Transportation routing. Things that used to take six months of consultant time can be wired up in days.

The bet here is unfashionable in 2026. Most AI startups are racing to deliver the end answer - black-box copilots that promise to "do supply chain" for you. Lyric runs the other way. It gives planners and decision scientists the primitives and tells them to build the answer themselves, faster. The marketing word is composable. The honest word is respect.

Supply chain software was built to sell, not to use.

Anyone who has lived inside an SAP APO or a legacy Kinaxis instance knows the joke. The software is powerful, expensive, and miserable. Configuration takes quarters. Customization takes consultants. The people who actually run the supply chain - planners, schedulers, ops leads - are reduced to button-pushers in a system that was sold over their heads, to a CIO who has not opened the tool since.

Ganesh Ramakrishna had seen this from every angle. He had co-founded Opex Analytics, sold it into Coupa (which had also absorbed LLamasoft), and spent years watching brilliant decision-science work get strangled by the surrounding software. The pattern was always the same. A team would commission a model. Six months later, the model would exist. By then, the question had changed. The model would not.

So Lyric's founding premise is almost rude in its simplicity: the planner is the builder. Give her the canvas. Stop selling her the painting.

The second-act team.

You can tell a lot about a company by who shows up on day one. Lyric's founding team is, almost suspiciously, the same crew that built Opex and rode shotgun through LLamasoft and Coupa. Ronan O'Donovan on product. Vish Oza on machine learning. Sarang Jagdale on decision science. Karthik Muthukrishnan on engineering. Sara Hoormann on strategy. These are not first-time founders chasing the AI wave. They are operators who have already lived the failure mode of supply chain software once - and are doing it again, on purpose, with a different bet.

The wager goes like this: large-language-model breakthroughs collapse the cost of two things supply chain has needed forever - synthetic data and natural-language interfaces to math. Combine those with a builder-first runtime, and the moat is not the model. The moat is the platform underneath.

Lyric Studio, or: an IDE for the supply chain.

Studio is the thing everything else hangs from. It is a canvas. On one side, the company's data - ERP feeds, transportation systems, warehouse logs, the spreadsheets nobody admits to. On the other side, a library of supply chain math: forecasting models, network optimization, simulation, inventory policy. In the middle, configurable workflows that knit the two together, plus apps the team can ship to its own planners.

The interface is opinionated about who is using it. Decision scientists get Python. Planners get UI. Engineers get APIs. Executives get dashboards. The same model behind each. Nobody is forced into someone else's tool.

Layered on top is the AI tier. Foundation models tuned to supply chain problems, synthetic data generation (because real supply chain data is famously scarce), and an agentic layer that handles the long-tail of routine decisions - the ones that should never require a human, and currently always do.

Make every supply chain AI-first. The marketing line that turns out to be the actual plan.

Most companies' mission statements survive a single all-hands. Lyric's has been load-bearing. The internal logic is consistent across every product call, customer pitch, and recruiting page: data plus algorithms plus workflows plus accelerated compute equals a supply chain that can think. Take any one of those four out and the equation collapses. Most of Lyric's competitors are missing at least one.

The vision is broader than the product. Ramakrishna talks publicly about democratizing decision science - pulling it out of the consultant economy and into the hands of the operators. It sounds like a slogan until you sit in a planning meeting where the model that ships matters more than the model that wins a benchmark.

The next decade's supply chain will not look like the last decade's.

The argument here is not subtle. Global supply chains are getting more fragile - tariff regimes, climate disruptions, geopolitics, the ongoing reshoring spasm. The companies that handle the next decade well will not be the ones with the biggest planning teams. They will be the ones whose teams can model, test, and ship a new policy in days instead of quarters.

That is the wedge. Lyric is not betting that supply chain becomes easier. It is betting that supply chain becomes a software problem - and that whoever owns the platform on which planners build wins a category that has been ossified since the Clinton administration.

It is also worth noting what Lyric is not doing. It is not selling a chat-with-your-data demo. It is not branding itself as an agent. It is shipping a runtime. The difference is the difference between a toy and a tool, and the customers Lyric has accumulated suggest the buyers can tell.

7:14 a.m., somewhere outside Atlanta. Again.

The cocoa moves. The Friday meeting is not tense. The planner does not become a hero - she becomes ordinary, which is the point. The model her team built does its work and then disappears, the way good infrastructure should.

If Lyric is right, this is what the next decade of supply chain looks like. Less heroism. More instrumentation. Planners who build, models that adapt, and the long quiet hum of decisions getting made faster than they used to be made.

If Lyric is wrong, it will be wrong in interesting company - alongside every other team that thought the planner deserved better tools. But the receipts so far suggest Lyric is not wrong. Just early.