Test Before You Build: Performance Prediction

Building performance simulation used to belong exclusively to engineers with expensive software licenses and months of training. AI has changed that equation. Architects, project managers, sustainability leads, and urban planners can now run meaningful energy, daylight, and thermal analyzes using plain-language prompts, no simulation software certification required. This lesson covers what building performance simulation actually means in an AI-assisted workflow, which tools handle which tasks, and how to get results you can use in client meetings, planning submissions, and design reviews.

7 Things to Know Before You Start

1. AI does not replace certified energy modeling software (EnergyPlus, IES-VE, DesignBuilder) for code compliance submissions, it accelerates early-stage decision-making and interpretation of results.
2. The most useful AI tools for simulation support right now are ChatGPT Plus (GPT-4o), Claude Pro, Microsoft Copilot integrated into Excel/Word, and Autodesk Forma, each with different strengths.
3. You do not need to understand the math behind U-values or solar heat gain coefficients. You need to know what questions to ask and how to describe your building clearly.
4. AI can read and interpret simulation reports, paste in a table of EUI figures and ask for a plain-English summary with recommendations.
5. Autodesk Forma (formerly Spacemaker) is the most purpose-built AI tool for early-stage massing, daylight, and microclimate analyzis. It runs directly in the browser.
6. Garbage in, garbage out still applies. Vague building descriptions produce vague analyzis. Specific inputs, floor area, orientation, climate zone, glazing ratio, produce actionable outputs.
7. Building performance simulation with AI is most powerful at three project stages: concept design (massing alternatives), schematic design (envelope decisions), and post-occupancy (interpreting energy bills against benchmarks).

What Building Performance Simulation Actually Covers

Building performance simulation is an umbrella term for several distinct analyzes. Energy simulation models how much electricity and heating fuel a building will consume annually, expressed as Energy Use Intensity (EUI) in kBtu/sq ft/year or kWh/m²/year. Daylight simulation models how natural light penetrates and distributes across interior spaces, critical for occupant wellbeing, LEED credits, and reducing artificial lighting loads. Thermal comfort simulation models whether occupants will feel too hot or cold based on air temperature, radiant heat, humidity, and air movement. Each analyzis requires different inputs and produces different outputs.

Most non-technical professionals interact with simulation results rather than run the simulations themselves. A mechanical engineer produces an EnergyPlus model; you receive a 40-page PDF. AI tools are exceptionally good at this translation layer, turning dense technical output into decisions. Claude Pro and ChatGPT Plus can read simulation reports, identify the three highest-impact design changes, compare your building's EUI against ASHRAE 90.1 benchmarks, and draft the sustainability narrative for your planning application. That is where most architecture professionals will find immediate, practical value.

• Energy Use Intensity (EUI): Total annual energy per square foot/meter. Lower is better. ASHRAE 90.1-2019 sets code minimums by building type.
• Daylight Autonomy (DA): Percentage of occupied hours a space receives sufficient daylight without electric light. Target: 50%+ for LEED credit.
• Useful Daylight Illuminance (UDI): Refines DA by excluding hours where daylight is too intense (glare). Target range: 100–3000 lux.
• Thermal Comfort (PMV/PPD): Predicted Mean Vote and Predicted Percentage Dissatisfied. ASHRAE 55 standard metrics for occupant comfort.
• Solar Heat Gain Coefficient (SHGC): How much solar radiation passes through glazing. Lower SHGC = less heat gain, important in hot climates.
• Infiltration Rate (ACH): Air changes per hour through the building envelope. Tighter envelopes = lower heating/cooling loads.
• Carbon Intensity: kgCO₂e per m² per year, increasingly required for planning submissions in the UK, EU, and major US cities.

Click to enlarge

Quick Reference: Simulation Types and AI Tool Fit

How AI Fits Into the Design Workflow

Think of AI as a performance consultant available at every project stage, not just when you commission a formal energy model. At concept stage, Autodesk Forma lets you sketch massing alternatives and immediately see how each option scores for daylight and solar exposure. You do not draw in Forma the way you draw in Revit; you push and pull simple volumes and the platform runs the analyzis automatically. This takes minutes, not weeks, and gives design teams real data to present at early client reviews instead of intuition and precedent alone.

At schematic and design development stages, AI shifts into interpretation and comparison mode. Your structural or MEP engineer may run the formal EnergyPlus or IES-VE model. You receive results. ChatGPT Plus and Claude Pro can then help you understand what those results mean for specific design decisions, should the south facade glazing ratio drop from 60% to 40%? What is the energy penalty of the client's preferred curtain wall system? How does the proposed building compare to ASHRAE 90.1 baseline? These are questions that previously required another engineering meeting. Now they take three minutes and a well-constructed prompt.

1. Concept Stage: Use Autodesk Forma to test massing alternatives for daylight, solar, and wind before committing to a scheme.
2. Schematic Design: Use ChatGPT Plus or Claude Pro to interpret engineer-produced simulation outputs and identify the highest-impact envelope changes.
3. Design Development: Use Cove.tool AI to run parametric energy studies on glazing ratios, insulation levels, and HVAC system options.
4. Planning Submission: Use Claude Pro to draft energy and sustainability statements from simulation data in plain language aligned with local planning policy.
5. Construction Documentation: Use Microsoft Copilot in Excel to track specification compliance against simulated performance targets.
6. Post-Occupancy: Paste 12 months of energy bill data into ChatGPT Plus alongside the design-stage EUI target, ask for a gap analyzis and likely causes.

Click to enlarge

Prompting AI for Performance Simulation Support

Understanding Climate Zones and Why They Change Everything

Climate zone is the single most important contextual variable in building performance simulation. The same building design performs radically differently in Phoenix (Climate Zone 2B, hot-dry) versus Minneapolis (Climate Zone 6A, cold-humid). ASHRAE divides North America into eight climate zones; the EU uses a similar system under EN ISO 13790. When you prompt an AI tool for performance guidance, including the climate zone is not optional, it is the difference between useful advice and dangerously wrong advice. If you do not know your project's climate zone, the US Department of Energy's Building Energy Codes Program website provides a postcode lookup tool.

Climate zone affects every major design decision: how much insulation you need, the ideal glazing ratio for each facade orientation, whether natural ventilation is viable, the appropriate SHGC for windows, and what HVAC system type will perform most efficiently. AI tools handle climate zone adjustments well when you provide the information clearly. Claude Pro and ChatGPT Plus both have strong knowledge of ASHRAE climate zone requirements and can compare design options against zone-specific baselines without any additional software. Urban planners working on masterplans that span multiple climate zones should specify this explicitly in every prompt.

Click to enlarge

Part 1 Cheat Sheet

• Building performance simulation covers: energy (EUI), daylight (DA/UDI), thermal comfort (PMV), solar, wind, and carbon.
• AI tools for non-technical professionals: Autodesk Forma (massing/daylight), ChatGPT Plus and Claude Pro (interpretation/writing), Cove.tool AI (parametric energy), Copilot in Excel (compliance tracking).
• AI cannot produce certified compliance documents, use it for decisions, interpretation, and communication.
• Always include: building type, floor area, climate zone, and facade orientation in every performance-related prompt.
• EUI (Energy Use Intensity) = annual energy ÷ floor area. Lower is better. ASHRAE 90.1-2019 sets code baselines by building type and climate zone.
• Climate zone is the most important contextual variable. The same design performs completely differently in Phoenix vs. Minneapolis.
• Autodesk Forma runs daylight, solar, and wind studies on massing models in the browser, no simulation training required.
• AI is most valuable at three stages: concept massing decisions, interpreting engineer-produced simulation reports, and post-occupancy gap analyzis.
• Paste simulation reports or energy bills directly into ChatGPT Plus or Claude Pro, they can read and interpret tables of data.
• For climate zone lookup: US DOE Building Energy Codes Program website provides postcode-level climate zone identification.

Key Takeaways from Part 1

• AI democratizes building performance insight, project managers, sustainability leads, and urban designers can now engage meaningfully with simulation data without engineering training.
• The highest-value use of AI for most non-technical professionals is interpreting existing simulation outputs, not generating new simulations from scratch.
• Autodesk Forma is the purpose-built tool for early-stage AI-assisted performance analyzis; ChatGPT Plus and Claude Pro handle interpretation, writing, and comparison tasks across all project stages.
• Specifying climate zone, building type, floor area, and orientation in every prompt is the single most important habit for getting accurate, actionable AI responses on building performance.
• AI-generated performance figures are decision-support tools, not regulatory compliance documents, know the boundary between the two.

You now know what building performance simulation is and why it matters. This section gets practical: which AI tools connect to simulation workflows, how to read the outputs they generate, and how to use those outputs to make real decisions, in client meetings, design reviews, and budget conversations. No engineering degree required.

7 Things Every Planner and Architect Should Know About AI Simulation Outputs

1. AI simulation tools generate probability ranges, not single fixed answers, a result showing '68–74°F interior temperature' is more trustworthy than one showing exactly '71°F'.
2. Most AI-assisted platforms (Cove.tool, Autodesk Forma, TestFit) produce outputs as dashboards, not raw data, you read charts and scores, not spreadsheets.
3. Energy Use Intensity (EUI) is the single most important number in building performance reports, it measures energy per square foot per year and allows apples-to-apples comparisons between buildings.
4. A simulation is only as good as its inputs: wrong floor area, wrong climate zone, or wrong occupancy schedule will produce confidently wrong results.
5. AI tools can run hundreds of design variations in minutes, what used to take an engineering firm two weeks now takes an afternoon inside tools like Forma or cove.tool.
6. Simulation results are negotiating tools, they give you hard numbers to justify design choices, push back on value-engineering cuts, and satisfy green building certification reviewers.
7. Carbon and energy are related but separate metrics, a building can have low operational carbon (efficient HVAC) but high embodied carbon (concrete-heavy structure), and AI tools now track both.

Reading AI Simulation Dashboards Like a Professional

Most architects and planners first encounter simulation outputs as colorful dashboards inside tools like Autodesk Forma or cove.tool. These dashboards compress complex physics into readable scores. The key is knowing which numbers actually drive decisions. EUI sits at the top of every energy dashboard for good reason, it collapses heating, cooling, lighting, and plug loads into a single comparable figure. A typical US office building runs around 50–80 kBtu/ft²/year. A high-performance building targets below 30. LEED Platinum requires you to demonstrate significant improvement over a baseline. AI tools calculate this automatically when you input building geometry, location, and use type.

Daylight autonomy scores appear alongside energy metrics in tools like Forma and ClimateStudio. This number tells you what percentage of occupied hours a space receives adequate natural light without electric lighting. A score above 50% is considered good practice for offices. Below 30% signals a daylighting problem, narrow floor plates, deep plan layouts, or insufficient glazing. AI tools generate these scores as color-coded floor plan overlays, making it easy to spot underperforming zones and adjust window placement or interior layout before construction documents are issued.

• EUI (Energy Use Intensity): kBtu/ft²/year, lower is better; benchmark against CBECS national averages for your building type.
• Daylight Autonomy (DA): % of hours with sufficient daylight, target 50%+ for occupied spaces.
• Carbon Intensity: kgCO₂e/m², tracks operational and embodied carbon separately.
• Thermal Comfort Score: based on ASHRAE 55 standard, flags overheating or underheating risk zones.
• Peak Demand (kW): highest single-hour electricity draw, affects utility costs and grid connection fees.
• Window-to-Wall Ratio (WWR): % of facade that is glazing, directly impacts both daylight and heat gain.
• Solar Heat Gain Coefficient (SHGC): how much solar energy passes through glass, lower values reduce cooling loads in hot climates.

Using AI to Compare Design Options. The Parametric Approach

Parametric simulation is where AI genuinely changes the design process. Instead of simulating one design and accepting it, you define a range of variables, orientation, glazing percentage, shading depth, insulation value, and let the AI run every combination. Autodesk Forma does this at the massing stage. Cove.tool does it for envelope and mechanical system choices. The output is a matrix of options ranked by performance. You pick the cluster of options that meet your energy target and fit your budget, then refine from there. This replaces the old workflow of waiting two weeks for an engineer's energy model.

For urban planners, parametric simulation extends to district-scale analyzis. Tools like Forma can model how a proposed building affects wind comfort and daylight access for surrounding streets and existing buildings. This matters enormously for planning approvals in dense urban areas. Cities including London, Singapore, and New York increasingly require shadow and wind studies as part of planning applications. AI tools generate these studies in hours rather than weeks, and the outputs can be embedded directly into planning submissions, community engagement presentations, and environmental impact reports.

1. Define your performance targets first. EUI goal, daylight minimum, carbon budget, before opening any simulation tool.
2. Set your fixed variables (site, program, budget) and your flexible variables (orientation, glazing ratio, facade material).
3. Run the parametric study, in Forma or cove.tool, this means setting ranges for each flexible variable and clicking 'analyze'.
4. Read the output matrix, look for the 'Pareto front': designs that are best on energy without being worst on cost.
5. Flag the top three to five options for design team discussion, never present a single 'AI-recommended' design without alternatives.
6. Document the simulation assumptions, climate file used, occupancy schedule, HVAC system type, so results are defensible in review.
7. Re-run after each major design change, simulation is not a one-time exercise, it should update with the design.

Click to enlarge

Translating Simulation Results Into Client and Stakeholder Language

Common Simulation Mistakes That Produce Misleading Results

Simulation outputs carry authority because they look precise. A chart showing '42.3 kBtu/ft²/year' feels more certain than an engineer's rough estimate. That precision can be misleading. The most common error in AI-assisted simulation is garbage-in-garbage-out: using default occupancy schedules that don't match the actual building program. A simulation assuming a standard 9-to-5 office schedule will significantly underestimate energy use for a 24-hour hospital, a school with evening community use, or a mixed-use building with retail on the ground floor. Always verify that occupancy inputs match the actual intended use of every space.

A second major mistake is treating early-stage simulations as final engineering calculations. Forma and similar tools are designed for concept-stage decisions, orientation, massing, rough envelope performance. They are not substitutes for the detailed energy models produced by mechanical engineers during design development. Using a concept-stage EUI figure to make final HVAC sizing decisions or to commit to a specific energy cost guarantee in a client contract is a serious professional risk. AI simulation tools are most powerful when used at the right stage of design, for the right level of decision.

Click to enlarge

Part 2 Cheat Sheet. Building Performance Simulation Reference

• EUI = Energy Use Intensity (kBtu/ft²/year), your primary performance benchmark; US office average is ~50–80.
• Daylight Autonomy, target 50%+ for occupied office and classroom spaces.
• Autodesk Forma = best entry point for concept-stage massing and site analyzis.
• Cove.tool = best for energy modeling and green building certification support.
• Parametric simulation = running many design variants automatically to find the best-performing options.
• Climate file = 8,760-hour weather dataset; must match your actual site location.
• Always document simulation assumptions, occupancy schedule, climate file, HVAC type, equipment loads.
• Concept-stage results carry ±15–20% uncertainty, never use them for final cost guarantees.
• Embodied carbon and operational carbon are separate, track both for full sustainability reporting.
• Use AI outputs as negotiating tools, they justify design decisions with numbers, not just opinions.
• Benchmark against CBECS national averages before presenting any EUI result to a client.
• Label early-stage simulations clearly, 'Concept Estimate' not 'Final Energy Model'.

Key Takeaways From This Section

• AI simulation dashboards are readable without engineering training. EUI, daylight autonomy, and carbon intensity are the three numbers that matter most in most project conversations.
• Parametric simulation lets you test dozens of design options in the time it used to take to test one, this changes the economics of performance-led design.
• The right tool depends on your project stage: Forma for early massing, cove.tool for energy and compliance, ClimateStudio for detailed daylight.
• Misleading results almost always come from wrong inputs, not wrong AI, verify occupancy schedules, climate files, and program assumptions before trusting any output.
• Simulation results are most powerful when translated into client language: dollars saved, certification points earned, and occupant comfort improvements.

AI-assisted building performance simulation is no longer reserved for engineers with specializt software. Non-technical professionals, project managers, sustainability leads, urban planners, and client-facing architects, can now use conversational AI tools to interpret simulation outputs, generate performance briefs, and communicate energy and comfort trade-offs to stakeholders. This section consolidates the core concepts into a reference you can use on active projects.

1. AI tools interpret simulation data, they don't replace certified energy modeling software like EnergyPlus or IES VE.
2. Natural language prompts can turn complex simulation reports into plain stakeholder summaries.
3. Thermal comfort, daylighting, and energy load are the three metrics most commonly flagged in AI-assisted reviews.
4. Climate zone directly affects which design strategies AI will prioritize in its recommendations.
5. Passive design strategies (orientation, shading, thermal mass) consistently outperform active systems on cost-per-unit-saved.
6. AI can help you ask better questions of your engineering consultants, not replace them.
7. Free tools including ChatGPT (free tier), Google Gemini, and Microsoft Copilot can handle performance brief drafting and data interpretation tasks today.

Reading Simulation Outputs Without an Engineering Degree

Most simulation reports arrive as dense PDF documents or spreadsheet exports filled with acronyms: EUI, WWR, SHGC, ACH, PMV. For non-technical professionals, these outputs are nearly unreadable without context. AI tools like ChatGPT or Claude can act as translators. Paste a section of the report into the chat, describe your role and project goals, and ask for a plain-language interpretation. The AI will flag which numbers are concerning, which meet standard benchmarks, and what design changes are typically associated with improvement.

The key is giving the AI enough context. A raw number like "EUI: 87 kBtu/ft²/yr" means nothing without knowing the building type, climate zone, and target standard (ASHRAE 90.1, LEED, Passive House). Tell the AI all three. A well-framed prompt produces a response that identifies whether 87 is good, average, or poor for your specific building type, and lists the most likely causes if it's high. This turns a technical report into an actionable brief in under two minutes.

• EUI (Energy Use Intensity): Total energy per square foot per year. Lower is better. Benchmarks vary sharply by building type.
• PMV (Predicted Mean Vote): Thermal comfort scale from -3 (cold) to +3 (hot). Target range is -0.5 to +0.5.
• SHGC (Solar Heat Gain Coefficient): How much solar heat passes through glazing. Lower values reduce cooling loads in warm climates.
• WWR (Window-to-Wall Ratio): Percentage of facade that is glazing. Higher WWR increases both daylight and heat gain risk.
• ACH (Air Changes per Hour): Ventilation rate. Too low risks air quality; too high increases heating/cooling energy.
• DGI (Daylight Glare Index): Measures visual discomfort from excessive brightness. Values above 22 indicate glare problems.

Using AI to Compare Design Options

Design teams routinely generate multiple simulation runs, different glazing ratios, shading configurations, or mechanical systems, and need to compare them quickly. AI tools excel at structured comparison tasks. Feed two or three scenario summaries into ChatGPT or Claude and ask it to rank them against specific criteria: lowest energy cost, best thermal comfort, fastest payback period. The AI won't run the simulations, but it will organize and interpret the data your engineers provide in seconds.

This is especially useful when presenting options to clients or planning committees who aren't equipped to read technical tables. Ask the AI to rewrite the comparison as a one-page executive summary, a pros-and-cons list, or a recommendation memo. You control the format. A comparison that takes an engineer 30 minutes to write up can be reformatted for a non-technical audience in under three minutes with a well-structured prompt.

1. Collect scenario summaries from your engineering consultant, even rough bullet points work.
2. Open ChatGPT, Claude, or Google Gemini (all free tiers sufficient for this task).
3. Paste the scenario data and specify your evaluation criteria (cost, comfort, carbon, resilience).
4. Ask for a ranked comparison with brief reasoning for each rank.
5. Request a second output formatted as a client-ready summary, plain language, no jargon.
6. Ask the AI to identify which assumptions have the most impact on the ranking.
7. Use the output as a first draft, review with your engineer before sharing externally.

Communicating Performance to Non-Technical Audiences

The gap between simulation data and stakeholder decision-making is primarily a communication problem. Planning committees, clients, and building owners rarely understand kBtu/ft²/yr, but they do understand monthly utility bills, occupant sick days, and carbon equivalent comparisons. AI tools can translate between these languages instantly. Ask ChatGPT to convert an EUI figure into estimated annual energy cost (provide the floor area and local utility rate), or to express a carbon reduction as an equivalent number of cars removed from the road.

Visualization is the next frontier. Tools like Canva AI and Microsoft Copilot can take your AI-generated summaries and help format them into presentation slides or infographic layouts. You're not building dashboards, you're formatting a clear narrative. The combination of AI-interpreted data plus AI-assisted formatting means a single sustainability lead can produce board-ready performance reports in a fraction of the time it previously required a specializt firm to deliver.

Key Takeaways

• AI tools translate simulation metrics into plain language. EUI, PMV, SHGC, and ACH can all be explained and benchmarked in seconds with a well-framed prompt.
• ChatGPT, Claude, and Google Gemini (all free tiers) are sufficient for interpretation, comparison, and communication tasks, no specializt software needed.
• Always give the AI context: building type, climate zone, target standard, and audience. Generic prompts produce generic output.
• Passive design strategies (orientation, shading, thermal mass) consistently deliver better cost-per-unit-saved than active mechanical upgrades. AI recommendations will reflect this.
• AI-generated summaries are communication tools, not engineering certifications. Licensed professionals must verify anything used for compliance or contract purposes.
• The most powerful use case for non-technical professionals: turning a 40-page simulation report into a one-page stakeholder brief in under five minutes.
• Microsoft Copilot inside Excel can summarize simulation data without copy-pasting, valuable if your data workflow lives in Office 365.
• Combining AI interpretation with Canva AI or Copilot formatting creates board-ready performance presentations without a specializt communications team.