Single speaker, screen share on the live deck. This is the verbatim script: say the words exactly as written, press → at each new block. About six minutes end to end. Green boxes are the lines that must land word for word.
Hi everyone. I'm Ben, I'm the CTO at Nomad. I want to take a few minutes to talk about what we're doing at Nomad and what we're building at the moment. We're building a system called Orchid, and today I'm going to walk you through exactly what Orchid does.
For the purpose of this demo I'm going to use a manufacturing company. It could be any manufacturing company, and just to reiterate, Orchid is not limited to manufacturing. Manufacturing just got the nice icons for the demo.
[pause for the laugh, then straight into the bridge below]
So, when we look at a manufacturing company, what we're really talking about is the model: what it is, why it matters, and how you use Orchid to build decision-centric systems in the real world. This platform you can see is where that model lives.
The model is the nouns and verbs that make up your business. Here's the first noun: a plant. And it's not a box on a diagram, it's a live object. Its status, its energy, its issues, right now.
Now the verbs. I'm supplying plants from warehouses, and warehouses prepare orders. The relationships are first-class, and the goal is to capture how your business actually operates, not how other systems need it structured so that they can work.
And it keeps going. Orders ship to customers, customers generate revenue, revenue updates the forecast. This is the ground truth of the whole operation: every object with real state, every relationship named.
So what do I need to make decisions on top of this? Three things. I need the data, I need the knowledge, and I need the actions. The data is the current state of my business. The knowledge is how my people actually think about it. And the actions are what we can take to affect the real world.
In Orchid, everything you ingest resolves into four primitives: entities, knowledge, trackers and actions. That's the whole model. There is no fifth.
Data first. I'm connecting to all my enterprise systems: homegrown, off-the-shelf, legacy, new web tools, it doesn't matter. Orchid's collectors give you out-of-the-box connectors, or you pull straight from Snowflake, Databricks, BigQuery, data that's already sitting in your lakes, and it resolves into entities in the model.
And one thing we do differently: permission resolves at ingest. Whoever was allowed to see something in the source system, that's stamped on as it comes in, so the model never shows anyone what they couldn't already see.
Then the knowledge: the part of your business that never made it into a system. It might be an Excel spreadsheet. It might be rules of thumb, or how your best planner actually decides, the exceptions that live in people's heads.
Orchid captures that as a first-class part of the model, through structured voice sessions with your people, and every assertion is anchored back to its evidence. So it connects to your systems, and it connects to your people.
The last piece is the actions. If I've got the data and the knowledge, what can I actually do? Whether it's SAP and I want to create an STO to move product across my network, or it's on the plant floor, a financial system, a webhook: legacy, on-prem, edge or cloud, those are modelled as the actions I can take. You can see them appearing on the model now.
And here's the rule that is not negotiable: agents draft, people send. Nothing reaches a client and nothing posts to a ledger without a human sending it. That's a term in the customer contract, not a setting.
Bring the data, the knowledge and the actions together and I have a model that is rich: a structured, current, permission-aware picture of how my business is operating. Teams explore it and build on it, and the insights bubble up from the model itself.
And this is the important bit: the model is the product. The agents are configuration on top of it. There's no workflow builder and no per-customer code. A pack that works for one distributor works for the next one.
Now we add the reasoning. The agents can reason about your business because they have the model: the context, not just the data. They use the knowledge, how your people think. They can call a deterministic model. And they draft the action to write back.
So instead of someone swivel-chairing from a dashboard into another tool, Orchid does the complex work in the back end and puts a drafted action in front of a person. They review it, and they send it. Because the LLMs were not trained on your business. The model is what gives them that context.
And last, the API. Orchid is API-first, everything goes through it. The mobile app for the queue on the plant floor, the desktop app, a custom app for your customers. And there's a generated, typed client of your model, effectively an SDK of your business, so every surface runs off the one model. That's what improves usability, speed and time to value.
So all of this comes together with one goal: AI and humans working together on the model. Over time the agents carry more and more of the work, and your people stay on the send button.
[optional, as you click the chip] And remember I said this isn't just manufacturing? Same Orchid, different model. Nothing else changed.
So that's Orchid. One living model of your business: the data, the knowledge and the actions, with agents doing more of the work and your people on the send button. We're building it right now at Nomad, and we're looking for a small number of companies to build it with. If anything you've seen sounds like your business, come find me afterwards, I'd love to talk. Thank you very much.