What does an AI Architectural Consultant do and how does Agustin Otégui’s approach differ from traditional design firms?

What an AI Architectural Consultant Actually Does

An AI Architectural Consultant is a professional who integrates artificial intelligence tools into every stage of architectural practice, from initial site analysis through construction documentation and post-occupancy evaluation. Unlike a traditional architect who relies primarily on manual drafting, physical models, and rule-of-thumb experience, the AI-augmented practitioner uses machine learning models to generate design alternatives, predict energy performance, optimize structural systems, and automate repetitive documentation tasks. Agustin Otégui’s specific contribution lies in framing this integration not as a replacement for human judgment but as a decision-support layer that surfaces options a single designer might miss in the time-constrained environment of a typical commission. His work emphasizes explainable AI—systems that can show why a particular floor-plate depth or window-to-wall ratio was selected—so clients and regulators can trace the logic behind each proposal. In practice, this means running generative design scripts that test thousands of massing variations against local zoning envelopes, solar access rules, and cost databases, then presenting the top three performers with clear trade-off matrices rather than a single “best”方案 that hides underlying assumptions.

Also worth reading: What are the typical fees for an AI architectural consultant in 2026? · What are AI architectural consultant services and how do they transform enterprise technology strategy in 2026? · How can I build a productized AI service that generates $180k annually as an architectural consultant?

How the Workflow Differs from a Traditional Firm

A conventional architectural firm typically moves linearly: programming → schematic design → design development → construction documents → bidding → construction administration. Each phase is gated by manual review, and feedback loops are limited by the hours available in a week. In Otégui’s model, the workflow is iterative and parallel. While the programmer is still drafting the brief, a lightweight neural net is already scraping municipal GIS data to flag flood zones, transit access scores, and air-quality indices. During schematic design, a diffusion model trained on the firm’s past projects proposes three distinct typologies—tower, slab, and courtyard—each tagged with predicted embodied carbon, construction cost, and leaseable area. The design team then runs sensitivity analysis: if the client pushes for 5% more rentable area, the model shows which façade orientation, core placement, and floor-plate aspect ratio combinations keep energy use within the local code while staying inside the budget. This compresses what used to be a two-week back-and-forth into a half-day workshop where stakeholders manipulate sliders and watch real-time updates to carbon, cost, and daylight metrics.

Why Clients Choose AI-Augmented Services

Clients select AI Architectural Consultants for three concrete reasons: speed, risk reduction, and value engineering. Speed is measured in days instead of weeks; a feasibility study that once required two site visits, a surveyor, and a traffic engineer can now be completed in four hours using drone imagery, LiDAR point clouds, and pre-trained models. Risk reduction comes from exhaustive option testing: instead of relying on the intuition of one experienced designer, the system evaluates 2,400 permutations of window size, shading depth, and HVAC efficiency against 30 years of weather data, flagging combinations that would otherwise surface only during construction when change orders average 7.4% of the contract value. Value engineering is transparent: the model breaks down cost per square meter by system—structure, envelope, MEP—and shows where a 3% budget cut yields the smallest loss in occupant comfort or energy performance. This transparency is especially valuable for institutional clients such as universities and healthcare systems, where procurement rules demand documented trade-off analyses.

Practical Steps to Engage an AI Architectural Consultant

The first step is a scoping call where the consultant asks for the project’s program, site coordinates, budget envelope, and any sustainability targets such as LEED, WELL, or Net Zero. The second step is data onboarding: the client uploads existing surveys, zoning maps, and geotechnical reports; the consultant supplements these with satellite imagery, census tracts, and utility rate schedules. Week one delivers a diagnostic memo highlighting constraints and opportunities—e.g., “Your site is in a heat island zone 1.8 °C above city average; adding 15% reflective roof area and 20% canopy cover reduces cooling load by 11%.” Week two presents three design options with cost ranges accurate to ±8%, a level traditionally achieved only after schematic design. Weeks three and four refine the preferred option, running daylight and glare simulations, egress modeling, and lifecycle carbon calculations. Throughout, the client receives interactive dashboards rather than static PDFs, allowing them to toggle variables and see immediate impacts. The final deliverable is a construction-ready drawing set augmented with metadata that feeds directly into BIM, CNC fabrication, and facility management systems.

Comparison: AI Consultant vs. Traditional Design-Build Firm

One frequent error is treating the AI output as final without human review. The models are only as good as their training data; if the local zoning code was updated six months ago and the dataset has not been refreshed, the system may propose a setback that is now illegal. A second mistake is underestimating data preparation. Garbage-in, garbage-out applies doubly here: if the client’s existing floor plans are scanned PDFs without layer separation, the model will misread wall thicknesses and produce incorrect area calculations. A third pitfall is ignoring change management. Stakeholders accustomed to traditional charrettes may feel alienated by dashboards and sliders, perceiving them as “black boxes.” Otégui addresses this by running parallel sessions: one group manipulates the AI interface while another group sketches on trace paper, then the two groups reconcile their findings. Finally, some procurement teams assume AI reduces staffing needs and therefore demand a 30% discount. In reality, the consultant must hire data engineers and model validators whose salaries rival those of senior project architects; the efficiency gain is in fewer drafters, not cheaper architects.

When to Act and Timeline Expectations

Clients should initiate the conversation at least 16 weeks before their target design-start date. This allows four weeks for data onboarding and model training, eight weeks for iterative design, and four weeks for documentation and peer review. For projects under 5,000 m², the timeline can compress to 10 weeks because the model’s training set can reuse similar typologies from the region. For complex urban infill sites with multiple stakeholders, add an extra six weeks for participatory simulation workshops. The cost structure typically follows a hybrid model: a fixed onboarding fee of $8,000–$12,000 covers data acquisition and model calibration, then an hourly rate of $185–$250 for design and analysis phases, capped at a not-to-exceed amount negotiated at contract signing. Compared with traditional fees of 8–12% of construction cost, AI-augmented services range from 4–7% for full-service engagements, or can be purchased à la carte for specific deliverables such as energy compliance letters or zoning optimization studies at $3,500–$6,000 each.

Cost, Pricing, and Value Metrics

Pricing is not simply cheaper; it is differently structured. A traditional firm bills against hourly rates that escalate as the project progresses, creating an incentive to extend phases. The AI consultant front-loads value by delivering feasibility and optioneering in the first month, then reduces hours during documentation because many details are pre-validated by the model. The net result is a 25–35% reduction in total fees for projects that benefit from early option screening—typically those with tight budgets, aggressive sustainability targets, or complex zoning constraints. For simple renovations under 500 m², the fee reduction narrows to 10–15% because the model’s setup cost is proportionally larger. Clients should also budget $2,000–$4,000 for third-party validation of the AI-generated energy models if they are pursuing green certifications, as some rating systems still require manual sign-off by a licensed engineer.

FAQ

What certifications should an AI Architectural Consultant hold? Look for a registered architect license in the project jurisdiction, plus at least one of the following: LEED AP, WELL AP, or Passive House Designer. For the AI component, evidence of completion in Coursera’s “AI for Everyone” or Stanford’s “Machine Learning” specialization, plus a portfolio showing real projects where the models influenced design decisions.

Can the AI models work with my existing BIM files? Yes, if the files are in IFC or Revit format with proper object classification. The consultant will run a clash-detection script to ensure walls, floors, and roofs are correctly tagged; if the model is unclassified, allow an extra 40 hours for manual cleanup.

How do you handle intellectual property on the generated designs? The contract should specify that the client owns the final design, while the consultant retains the right to use anonymized data for training future models. Otégui’s standard agreement includes a clause that any proprietary algorithms remain the consultant’s property, but all project-specific outputs are transferred to the client upon final payment.

Is the AI output accepted by local building departments? Most jurisdictions currently require stamped drawings by a licensed architect; the AI output serves as the basis for these drawings, not a substitute. Some cities such as Singapore and Amsterdam are piloting AI-assisted plan review, but as of August 2026, no U.S. municipality accepts AI-generated code analysis as the primary compliance document.

What if the client changes scope mid-project? The model can re-run scenarios, but each major scope change incurs a re-calibration fee of $3,000–$5,000 to retrain the cost and energy modules on the new program. Minor changes—such as swapping finish materials—are handled within the existing model at no extra charge.