Finch

Finch is an AI-assisted building design platform for architects and design teams developing massing, circulation, unit mixes, and floor plans. It connects design geometry with live project data, allowing a team to see how a change affects areas and layout possibilities while the scheme is still being explored. The product is also known by its Finch3D website address.

The central idea behind Finch is to make an architectural firm's design knowledge reusable. A team can develop a plan library and constraints, explore alternatives, refine selected layouts, and move geometry into established design software. Its AI agent, Archie, adds a conversational route to querying project information and performing supported edits. Finch offers a free entry plan, a paid Basic plan for feasibility generation, and an Enterprise offering for more extensive building and floor-plan generation.

Where Finch fits in a design project

Finch is most relevant when a building proposal needs to move beyond a compelling shape into measurable spatial organization. A residential block, for example, must accommodate a unit mix, circulation, entry positions, and internal layouts. Those requirements interact: increasing one apartment's frontage may affect a neighboring unit, while relocating a core may alter the efficiency of an entire floor.

The product overview describes a workflow built around firm standards, layout generation, refinement, and BIM handoff. That makes Finch useful to architects evaluating repeatable building types and teams whose previous plans contain valuable design knowledge. It is less like an image generator that produces a visualization of a possible building and more like a system for manipulating the layout and information behind a proposal.

Start with a defined design question. “Can this footprint support the requested apartment mix with a workable circulation arrangement?” gives the team something concrete to evaluate. A vague request for the best building does not. Record the assumptions behind the study, including the brief, area definitions, and constraints, so the resulting comparison has a meaning beyond the appearance of each option.

Bringing in a mass and its context

Finch integrates with existing design environments, including Rhino, Revit, Grasshopper, and Autodesk Forma. Its Forma upload guide explains how an extension transfers building volumes and context into a selected Finch project and variant. Preparing the incoming geometry is part of the workflow; the software cannot resolve an unclear modeling intention merely because a mass has been uploaded.

For a site study, separate the proposed buildings from surrounding context and choose descriptive variant names. A team might use “Courtyard scheme” and “Two-block scheme” rather than temporary names that become confusing during a client meeting. Check the imported mass before beginning generation: the intended buildings, height relationships, and context should be represented consistently.

In a Forma transfer, the guide distinguishes supported building tools and notes that section settings affect how a line building is interpreted. Details like this matter because one volume interpreted as several buildings may change the structure of the subsequent study. Use a small representative model to establish the transfer process before relying on it for a complex site. Keep the source model available so later changes can be traced to the geometry that created them.

Reading areas and customizing calculations

Finch's data calculation guide describes key figures such as gross floor area, internal area, and usable area, which update as relevant design parameters change. Programs can be excluded from certain calculations, and custom properties can relate a value to area. This supports early comparisons where geometry and numbers need to remain connected.

Area terminology needs agreement before a project team compares options. A client may talk about a target floor area while an architect is working with a different internal or gross measurement. Choose the intended definition and make it visible in the study. Otherwise, an option can appear to meet the brief simply because the two parties are counting different things.

A custom area-based budget property can provide a rough sensitivity study, but it is only as informative as the assigned rate. Use it to ask how a change in total area affects an assumed allowance, then keep that assumption separate from a developed cost estimate. Likewise, an emissions property based on a fixed factor is an input-driven calculation, rather than a material-by-material assessment. Write the factor and scope alongside the comparison so a precise-looking total does not obscure its early-stage basis.

Comparing design variants fairly

Finch supports multiple variants within a project. The variant documentation explains duplication, shared data customization, saved views, and comparison through project figures. A variant preserves a distinct proposal so that exploring one direction does not require overwriting the only version of the scheme.

Change one major question at a time when you want to understand a trade-off. To compare a central corridor with another circulation arrangement, keep the brief and calculation definitions consistent. If the footprint, wall thickness assumptions, and apartment targets all change at once, the result may be interesting but it will be difficult to explain which decision produced the difference.

Use comparable views for the discussion as well as comparable figures. A favorable camera angle can make a weaker proposal look more convincing. Include the practical implications that a metric cannot fully express: a difficult corner unit, a long route from entrance to lift, or a frontage that deserves a more careful study. A good option review combines the numeric comparison with annotated architectural observations, then records why a particular direction should advance.

Generating circulation and a unit mix

The corridor and unit-mix guide describes using a centerline, selected generation areas, stairwell ranges, exit-distance inputs, and weighted metrics. The centerline controls where generated corridors are placed. Users can adjust parameters during generation and then proceed to unit generation when the circulation arrangement is suitable.

These controls turn a broad brief into a more specific search. For a residential floor, establish the circulation logic before trying to optimize every apartment. Decide where the principal entrances and cores should be considered, identify the intended unit distribution, and examine how the selected priorities affect the resulting floor plate. Weighting one metric more heavily can improve that measure while creating an awkward result elsewhere.

Do not stop at the total unit count. Inspect corner conditions, access to each unit, leftover spaces, and the relationship between the core and the rest of the floor. If a result meets the numeric target only by producing a difficult sequence of internal spaces, refine the constraints or the massing. Generation is most useful as an iterative design aid, with the architect determining which proposals deserve further development.

Turning previous plans into a reusable library

Finch's Adaptive Plan Library lets a drawn unit plan be saved for reuse. Constraints can lock selected walls or allow a wall to extend without shrinking, helping preserve particular dimensions as a plan adapts. Library use differs by subscription level, and organization datasets support the Enterprise generation workflow.

Choose library examples for a reason. A well-resolved one-bedroom plan with clear entry and service relationships is more useful than a large collection of unreviewed drafts. Give each example enough context for another designer to understand it: intended occupancy, key dimensions, regional assumptions, and any condition that limits where it can be reused. The quality of a reusable design system starts with the quality of those examples.

Test an adaptive plan against several shapes before treating it as a standard. A layout may remain technically adaptable while becoming less convincing in a narrow footprint or at an unusual corner. Look at the living space, furniture clearances, entrance sequence, and access to daylight after adaptation. A library should help carry design knowledge forward, while still leaving room to recognize when a new situation requires a new plan.

Assigning and refining unit plans

The unit-plan tutorial distinguishes scanning a personal library from Enterprise plan generation. It covers graph rules, filtering suggestions, inspecting fit information, assigning a plan, and editing the result. Fit information can reveal how much rooms adapt and how the suggested plan relates to the unit shape. Matching units can also share linked or mirrored layouts.

Review a plan's spatial logic before accepting it because of a favorable score. Check where a visitor enters, whether circulation cuts through the useful living area, and how storage and service spaces fit the brief. Look at the rooms individually and at their relationships. A plan that fits the outline is still a design proposal requiring judgment.

Linked layouts are valuable for repeatable units, but they also make the scope of a change important. Before revising one apartment, establish whether it is intended to remain a shared template or become a special case. Review all affected units after a linked edit, especially mirrored versions with different facade conditions. This protects the value of repetition without overlooking the local differences that make one unit less straightforward than another.

Archie for project queries and supported edits

Archie is Finch's AI agent for working with design data and supported layout changes. Its documentation includes retrieving building statistics, finding project elements, inspecting room relationships, exporting structured tables, adjusting doors, and handling certain wall or window edits. The agent operates on project information and geometry, rather than serving only as a general architectural chatbot.

Begin with a narrow query before asking for a broad edit. For example, request an overview of units that match a particular area or room condition, inspect that selection, and then decide whether a shared adjustment is appropriate. A clear scope reduces the chance that a bulk change affects units you intended to preserve.

For a door-editing task, specify the relevant units and the desired result in measurable terms. After the change, inspect the doors and adjacent spaces rather than relying on a conversational completion message alone. Consider a prompt as another way of issuing a project instruction, with the same need for review as a manual command. For significant revisions, keep a comparison variant and a record of the accepted changes so the team can explain what happened.

Moving a selected design into BIM and reports

Finch's export documentation describes BIM transfer to Revit, two- and three-dimensional data transfer to Rhino, Grasshopper data streaming, and PNG or CSV output. These routes support different handoffs: a design team may need geometry for further development, while a project meeting may need a clear image and an area comparison.

Plan the handoff before completing the entire study. Ask the receiving team which geometry, classifications, and project information they need, then try the export with a small representative portion. Review object mapping and the destination template before assuming that a transferred model is ready for documentation. Exported geometry is an input to the next design stage, with its own coordination work.

For a client report, show the selected variant with the brief, assumptions, and key metrics that justified it. A CSV alone can lose the design context, while an attractive image alone can hide the trade-offs. Keep the variant name consistent between images, tables, and source models. This makes it easier to connect a later revision with the option that the client actually approved for further study.

A practical residential feasibility workflow

Consider an architecture team studying a modest apartment building. Begin with a prepared mass and a written brief describing the desired unit mix. Confirm area definitions and the assumptions used for wall thicknesses and programs. Establish a base variant, then duplicate it to test a second circulation approach while keeping the brief consistent.

Generate and inspect floor-plate options, looking at circulation, core placement, usable areas, and the quality of the resulting unit shapes. Select a small number of promising options for plan assignment. Use reviewed library plans where they fit, and refine exceptional units individually. If Enterprise generation is part of the team's access, review those suggestions through the same architectural criteria.

Query the project data with Archie where useful, but verify any requested changes in the model. Compare the variants using both figures and annotated plans. Export the chosen geometry to the team's supported design software and record the unresolved questions for the next phase. The result is a more explicit feasibility decision: which option advances, what it satisfies, and what still needs a closer study.

Access, suitability, and limitations

The current plan page offers Free, Basic, and Enterprise access. Free supports a single user with editing tools and building metrics; Basic adds early-stage generation for unit mix and circulation; Enterprise adds firm-wide plan libraries, AI floor-plan generation, custom BIM integrations, and automated design workflows. Check the current plan and onboarding details for the capabilities required by your team.

Compatibility and entitlement are practical limits. The Finch FAQ lists supported software and notes that SketchUp and MicroStation are not currently supported. A firm should confirm its particular software versions and transfer requirements before restructuring a workflow around Finch. Full AI unit-plan generation should not be assumed to be part of free access.

Finch's rules and checks depend on the standards and constraints supplied to the system. A generated layout is not a permit approval or a substitute for professional coordination. Local regulations, structure, services, fire strategy, and project-specific obligations still need appropriate review. The platform is most useful when experienced designers use it to explore and refine decisions with clearer geometry and data.

Frequently asked questions

Is Finch a building image generator? Its main purpose is building geometry, floor plans, design iteration, and project data. The relevant output is a design model and layout information, rather than only a visual rendering.

Can I start for free? Yes. Finch offers a free plan, with generation capabilities differentiated across Basic and Enterprise access.

Can existing firm plans be reused? The Adaptive Plan Library supports saved plans and constraints. Review each example and its assumptions before using it as a repeatable standard.

Can it connect with Autodesk Forma? Finch documents a Forma extension for transferring supported building volumes and context into a project.

Does a generated plan remove the need for architectural review? No. Evaluate the plan, the input rules, and the wider building requirements before advancing a proposal.