海角视频

Insights from 海角视频鈥檚 December roundtable on the future of technology in the built environment.

On December 16th, 海角视频 convened three leading voices in architecture, engineering and construction (AEC) technology for a digital roundtable exploring one of the industry鈥檚 defining questions: ‘what do clients genuinely need from AI and automation today?’. Moderated by Patrick Pease, US building performance service leader at 海角视频, the discussion brought together:

• Anthony Hauck, CPO & Co鈥慒ounder,
• Nat MacDonald, Director of Product,
• Dr. Norhan Bayomi, Co鈥慒ounder & CPO,

Spanning everything from data stewardship to domain鈥憇pecific AI agents, the conversation underscored a key theme:

Clients don鈥檛 want 鈥淎I鈥� – they want outcomes. Better decisions. Faster clarity. Lower risk. Higher building performance.

Below is a recap of the most valuable insights for clients, owners, developers and designers navigating today鈥檚 fast鈥慹volving technology landscape.

1. Clients are seeking speed, clarity and confidence in early decisions

Across all three panelists, the most pressing client challenge was consistent: compressed timelines and the need to make high鈥憅uality decisions earlier than ever.

• Hypar customers – primarily architects – need rapid generation of options to support decision鈥憁aking under pressure. As Hauck put it, 鈥淭ime to decisions is getting more and more compressed, and clients want to see more options before choosing a path.鈥�
• TestFit targets the feasibility bottleneck. Traditional two鈥憌eek cycles between architects and developers often yield only a handful of static options. 鈥淲e automate feasibility, so you get the highest and best use instantly 鈥� and move faster into design,鈥� MacDonald shared.
• Lamarr.AI focuses on building owners who lack visibility into envelope performance – a blind spot that limits energy savings and retrofit planning. 鈥淭he building envelope deteriorates, but you can鈥檛 measure that easily today,鈥� Bayomi noted. Lamarr uses drones and thermal analysis to deliver reliable, objective diagnostics within hours.

Takeaway for clients: AI today is most valuable not as a futuristic vision, but as a mechanism to compress timelines, reduce subjectivity, and clarify decisions.

2. Automation vs. AI: what鈥檚 the real difference?

Clients frequently ask whether they need AI – or just better automation. The panel offered a simple distinction:

• Automation is deterministic.
   One task 鈫� one predictable output.
   Best for repeatable workflows and quality control.
• AI is probabilistic.
   It interprets, predicts, or synthesizes – especially useful when there 颈蝉苍鈥檛 a single correct answer.

As Hauck summarized: 鈥淚f there鈥檚 one right answer, you want automation. If there are many statistically likely answers, that鈥檚 AI.”

For Bayomi, the distinction is functional:

• Automation = scale + efficiency
• AI = intelligence + interpretation

3. The rise of domain鈥憇pecific AI agents in AEC

Large language models (LLMs) are now easy enough to customize that firms no longer need proprietary foundational models. Instead, the future lies in domain鈥憇pecific agents trained on a company鈥檚 own data -contracts, specs, standards, details, codes – not the entire internet. Bayomi highlighted the opportunity: 鈥淎EC won鈥檛 build its own foundational models. We鈥檒l fine-tune agents that deeply understand building systems, materials and performance.鈥�

Examples already emerging:

• Agents that read MEP specs and flag inconsistencies
• Agents that map system interdependencies
• Agents that crawl drawing sets and classify conditions
• Agents that identify energy鈥憇aving opportunities in existing buildings

MacDonald added that firms will likely rely on RAG (retrieval鈥慳ugmented generation) to link LLMs to their internal knowledge bases safely.

Key takeaway:
Expect agents that are tailored to your building type, portfolio or project workflow – unlocking real value without generic AI hype.

4. Data ownership and privacy are becoming contractual requirements

All panelists reported a surge in client awareness about data usage, especially for model training.

• Developers and architects now routinely include clauses restricting how software vendors can use project data.
• Tools like Hypar silo data per firm to prevent 鈥渂leed鈥憈hrough鈥� between clients.
• Lamarr.AI emphasizes anonymization and strict ownership boundaries, especially for sensitive sites.

Bayomi noted the cultural shift: 鈥淐lients ask, 鈥榃ill my data be used to train someone else鈥檚 model?鈥� Education is essential – and contracts must explicitly address this.鈥�

Key takeaway:
Clearer data鈥憉se language, stricter privacy controls and greater transparency on how AI tools learn from portfolio information.

5. Lessons from the BIM rollout: Integration matters more than innovation

While BIM promised a 鈥渟ingle source of truth,鈥� the reality fell short. The panel agreed on two big lessons for AI adoption:

• Lesson 1 – Don鈥檛 chase 鈥減erfect data.鈥� Build tools that work with imperfect reality.
  AEC workflows are messy and fragmented; AI that succeeds will embrace that.
• Lesson 2 – AI tools must integrate seamlessly into existing workflows.
  If a model flags a problem but can鈥檛 explain why, adoption slows.

Bayomi emphasized the need for explainability: 鈥淚f AI detects an HVAC issue but you can鈥檛 trace the reasoning, it won鈥檛 be trusted.鈥�

6. Will AI change how buildings are designed and operated? Short answer: yes.

The panel expects AI to reshape the built environment through indirect but powerful forces.

• AI 鈥� Driven Real Estate Patterns: Hauck predicts that AI will uncover hidden drivers of building performance and ROI: 鈥淧atterns in real estate returns will change design goals. One number changing in a spreadsheet can reshape an entire project.鈥�
• Shift to Performance鈥態ased Expectations: MacDonald noted that owners will increasingly ask: “Why am I getting 2D drawings when AI tools can generate dynamic analyses and visualizations in seconds?”
• Predictive, Integrated Building Operations: Bayomi sees enormous potential for operational optimization:
  – Real鈥憈ime energy prediction
  – Envelope鈥慸riven comfort insights
  – Automated adjustment of building systems
  – Reduced need for costly recommissioning cycles

This aligns with broader industry trends, including major acquisitions like AECOM鈥檚 move into AI鈥慹nabled building intelligence, signaling a shift from compliance鈥慴ased design to performance鈥慸riven delivery.

7. The human in the loop is non-negotiable

All panelists agreed: AI will augment expertise, not replace it. Key reasons include:

• Models are trained on the past – innovation still comes from people.
• LLMs struggle with geometry and multi鈥憇tep reasoning, both core to engineering and architecture.
• Liability will always require professional sign鈥憃ff.

As Hauck put it: 鈥淎I is the fastest and stupidest intern you鈥檒l ever have. You must check the work.鈥�

8. Guidance for emerging engineers and designers

The panel offered advice for early鈥慶areer professionals navigating AI鈥慹nabled practice:

• Build a strong domain foundation.
  Physics, building science, systems thinking – AI can鈥檛 replace this.
• Learn to use AI tools thoughtfully, not blindly.
  Bayomi emphasized understanding model limitations and being able to validate outputs.
• Don鈥檛 skip formative early鈥慶areer work.
  MacDonald warned against over鈥憆elying on AI: 鈥淚f firms under鈥慼ire now because 鈥楢I can do it,鈥� they鈥檒l feel that gap in 10 years.鈥�
• Focus on multi鈥慸isciplinary reasoning.
  LLMs struggle with interconnected systems – exactly where human expertise shines.

What clients really need right now

Based on the panel鈥檚 insights, clients today aren鈥檛 asking for AI for its own sake. Clients are asking for:

• Faster, clearer decisions in feasibility and design
• Reliable intelligence about existing building performance
• Better integration of disparate data sources
• Tools that improve outcomes, not complexity
• Assurance that their data remains secure and private
• Human oversight to ensure quality, safety, and accountability

海角视频鈥檚 takeaway is simple and client鈥慶entered: AI is most powerful when it augments expertise, accelerates decisions, and unlocks measurable performance gains – not when it replaces human judgment.

Watch the full discussion below.