Parametric Engine with the AI Brain: Turning What Your Firm Already Knows into Your Next Design

Design is a creative act, and it should stay one. A blank page is where good ideas begin, and no tool should take that away. But a blank page is a costly place to redraw a hotel suite, a clinical wing or a kitchen layout that your firm already solved. The best moments for creativity come when the proven parts are already in hand and the team can spend its energy on what is new about the project.

At Skema, we think starting from scratch deserves a refresh. Begin with your firm's own proven assemblies, then reshape them freely to fit a new site, a new program and a new idea. You keep every decision your teams have already made, and you still get to explore. That belief shaped how we built our engine, which pairs two kinds of intelligence: an AI brain that learns from your past work, and a parametric engine that applies rules the same way every time.

What AI in Design Needs to Get Right

As AI takes on more design work, two questions matter most. Does the system remember what your firm already knows? And can you trust it with that knowledge?

Many tools generate generic 3D shapes from scratch. They can look impressive in a demo, but the output has no connection to your standards, your assemblies or the decisions your teams refined over years. A different approach starts from proven, fully specified content and adapts it. That is the approach we took.

Trust belongs in the same conversation. Design knowledge represents years of investment and experience, and it deserves the same governance as any other critical business asset. Our partners at AVAIL describe the bet this way: as AI tools do more of the work inside a firm, they will need a governed source of information rather than guesses. We share that view, and it is why this topic sits at the center of our work with AVAIL.

The Parametric Engine

A parametric engine builds objects from rules and variables. Change one value, such as the height of a wall, and every connected element updates: window spacing, supports, adjacent rooms. Tools like Grasshopper and Revit work this way. The result is precise, repeatable and reliable.

Inside Skema, the parametric engine is a structured analytical model organized by building levels and constraints. It resolves program, code and adjacency rules together and keeps key metrics for the building type current as the design changes. Because the logic is rule-based, the outcome is deterministic, meaning the same inputs always produce the same design.

The AI Brain

We give an AI agent the parametric solvers as tools and skills, and it configures them to reach an outcome you describe in plain language. Skema reads a firm's completed Revit projects and harvests whole assemblies into a Design Catalog: room layouts, wall assemblies, equipment arrangements and their metadata. We manage chunks of buildings, so a hotel suite or an arrangement of equipment travels as a single reusable unit with its relationships intact. Each project adds to the catalog, so the body of knowledge compounds over time.

When you apply an assembly to a new footprint, you are able to instruct the brain to morph it. It stretches, squeezes or angles the geometry to fit the site while preserving the spatial logic inside it. We store the full native data of each assembly, so information you attached to it, such as equipment data from AVAIL or parameters linked to a database, comes back out in the model Skema generates.

Why the pairing matters most on the hardest problems

Data centers show the value of this pairing clearly because they are among the most complex building types we design. A data center layout runs several solvers at once: electrical equipment, rooftop cooling, the data hall, the mechanical gallery and circulation. Each solver carries 40 to 50 parameters, which adds up to roughly 300 settings that trade off against each other. Change one and the other four move.

Even an expert would struggle to remember which of 300 settings to adjust for a given result. An AI brain handles that well. AIs never forget anything. The agent learns what the 300 settings do, and you state the outcome you want.

The same approach applies wherever layout drives the result. Hotels and apartments depend on spatial arrangements. Hospitals and schools involve complex spatial layouts. Quick serve restaurants are shaped by the kitchen, which then drives parking and the drive-through. If the approach works on the most demanding case, other complex building types benefit from it too.

An AI brain working from placeholders can produce a plausible layout. Working from real equipment, with real data and real connection points, it produces a layout engineers can build on.

Real content in the loop: the AVAIL + Skema integration

That is the purpose of the integration between AVAIL and Skema. AVAIL Gateway extends AVAIL's governance capabilities to third-party applications and AI agents through APIs and MCPs, and Skema is the first AVAIL Gateway integration. Teams can find the right equipment, with the right data, inside a Skema project and let the solvers use it, without chasing files across systems. The integration enables new workflows and removes friction between the content a firm governs and the design work that depends on it.

From concept to a native Revit model

Teams can explore options quickly in the browser, run what-ifs, and edit assemblies by stretching them. When an option is ready, Skema generates a native Revit model in under a minute, built from your own families and office standards. The goal is a model you do not have to fix afterward, because a model that needs repair has not saved you time. Skema works alongside the design tools teams already use, so people keep working where they are comfortable.

Trust is part of the design

Our goal is to help you build durable design intelligence: knowledge that compounds over time, accelerates future work and remains a lasting competitive advantage. That requires confidence that your intellectual property stays under your control. Here is what you can expect when you use Skema:

• Your models remain yours. We make no claim of ownership over your project data or intellectual property.

• Your data is not used to train shared AI models. The work your team creates is not added to a centralized AI training corpus.

• AI works for you. Skema's AI features help organize, structure and surface your organization's own design knowledge so your teams can reuse it more effectively.

• Your expertise stays yours. The experience you have built across years of projects is not shared across the industry.

Business rules you embed in the solvers stay yours as well. What makes a result unique is your catalog and your rules.

Where this is heading

Our data center configurator is the most developed example today, and we are developing other building types over the coming months. Each new typology will use the same two-part approach: a brain that remembers, and an engine that holds the rules.

If you would like to see these connected workflows and help shape what comes next, let us know. Go to skema.ai/AVAIL. Sign up to get early access.