Introduction
AI is marketed as the future to replace humans.
This marketing is not happening anytime soon in the architectural industry, but the application of AI in assisting workers is available right now.
AI is capable of dealing with mundane jobs that take human time and involve repetitive processes. Humans make mistakes during these processes due to the mundane, repetitive nature of doing something the same way over and over again.
This makes AI in its current form ideal for working as a personal assistant to every person in the Architect Practice.
This could boost productivity and reduce stress levels of Architects and fit the criteria of
More Time = Money + less stress
This Integration report identifies tasks an AI can do within a company and how it may assist in an Architectural Practice, improving productivity and efficiencies, plus further advancements in the future.
Index
· Understanding AI
· Cost involved with AI
· AI Network Structure
· AI Role
· Assistant
· BIM library Search
· REVIT
· Wellbeing monitor
· Future possibilities
· Pros / Cons
· Conclusion
Understanding AI
AI is a process of digital input of text or voice, converting it to text before the process is carried out.
Example:
Input:
“Can you explain what benefit AI can add to an Architects Practice”
AI Response:
“Certainly! Artificial intelligence (AI) has the potential to revolutionize
various industries, including architecture. Here are some benefits that AI can
add to an Architect's Practice:
1. Automated Design Processing: AI can help automate repetitive and time-consuming tasks involved in the design process, such as generating multiple design options based on user inputs, analysing building codes and regulations, and optimizing building layouts for energy efficiency and cost savings.
2. Enhanced Customer Service: AI-powered chatbots can help architects provide 24/7 customer support, answer common queries, and even assist clients in generating design concepts based on their preferences.
3. Accelerated Building Information Modelling (BIM): AI can streamline the BIM process by automating tasks such as creating 3D models, generating construction documents, and performing clash detection between building systems.
4. Improved Project Management: AI can help architects manage projects more efficiently by predicting project timelines, identifying potential issues early on, and optimising resource allocation.
5. Enhanced Sustainability Analysis: AI can analyze building designs and suggest improvements for energy efficiency, water conservation, and waste reduction. This can help architects create more sustainable buildings that meet the needs of clients and society as a whole.”
6. Increased Collaboration: AI-powered tools can facilitate collaboration among architects, engineers, contractors, and clients by providing a shared platform for communication, file sharing, and project management.
7. Competitive Advantage: By leveraging AI, architects can differentiate themselves from competitors, attract new clients, and establish themselves as leaders in the industry.
8. Improved Accessibility: AI-powered tools can help architects create more accessible buildings by analysing building codes and regulations, identifying potential accessibility issues, and suggesting design solutions to address them.
9. Enhanced Data Analysis: AI can help architects analyse large datasets related to building performance, energy consumption, and occupant behaviour. This can inform design decisions and improve the overall performance of buildings.
10. Innovative Design Solutions: AI can generate novel design solutions by analysing patterns in building forms, materials, and layouts, allowing architects to explore new creative possibilities.
By leveraging these benefits, architects can streamline their workflows, improve
project outcomes, and provide better services to clients. As the field of AI
continues to evolve, we can expect even more innovative applications in the future.
The AI response doesn’t consider the user scale and cost of hardware or user costs in its answer.
AI can only respond from what it has learned, has access to or is directed to say.
All AI models can be trained specifically for a role which it is required to do.
A large general model like ChatGPT would cost far too much for all the computing units required by Nvidia or AMD to do AI processing, this would be a massive investment which most medium and large Architectural Practices couldn’t even consider, not counting the amount of electricity this requires to run all this hardware.
AI can be scalable, the large corporate companies are not marketing this information.
This is because they want tech firms to control the market.
AI can be run on clusters of low-power mini computers with low-power AI processing units, which during non-core business hours, can be used to train the AI in new roles, processes, etc.
This is limited against high-powered professional processing units but still possible and far more cost-effective than using computers and graphics cards (GPU’s).
Small clusters of mini PC's with AI processing units, make light work of large jobs.
Working with other software, AI has to be given access to input into external software and servers. This then allows it to apply inputs like a person using a computer.
Process example:
Search
Input Voice/Text - AI processes the question - Responds Text/Voice or command
Software Process
Input Voice/Text - AI processes the Question - Responds by completing the command via permission
Cost AI
AI can cost a lot of money if a large-scale commercial AI is required. These track cookies and collect user information and internet searches.
Image generation, video generation and many more tasks,
These are the major commercial AI, available online for free:
·ChatGPT
·Claude
·Gemini
·DeepSeek
·Grok
·META
and many more.
These free commercial AI’s are not good for protecting an Architect's Practice or client information.
In the case of a medium-sized Architectural Practice, a large commercial-sized AI isn’t needed.
Reasons:
Firstly, there is limited user access to AI, only staff in the Architect Practice.
Secondly, no information harvesting.
Lastly, the AI role will be more targeted at the company's requirements.
Allowing the use of mini PC’s like “Raspberry PI’s” or similar and reduces the hardware and running costs.
All AI models are released as free open-source models, which require some understanding of Python coding, networking, and cluster computing.
The Advantage of hosting these free AI models yourself is that they don’t track cookies, save data, or compromise a business's cybersecurity.
AI Network Structure
This is a basic formation of an AI using a cluster system
The (Red) identifies Network Switches can be formatted to best suit size and power requirements.
The (Green) identifies Storage servers, AI host, Cluster units which are all mini PC’s.
Proposed AI Roles
For an initial application of AI within an Architect Practice. It should start with simple applications which benefit the company.
To make initial time savings, support staff, fast searching and staff monitoring (possibly).
Assistant
This role is a simply being able to locate information online and manage staff calendar appointments and monitoring urgent emails. Recording a meeting, taking meeting minutes and then converting this to a text document and making key point summaries.
It is a big time-consuming task, most practices are starting to generally ignore meeting minutes in favour of summarised meeting minutes. If minutes were properly recorded, no backtracking can be done by any other parties; it could provide more fees.
This removes time spent for someone to type up the meeting as they listen to a recording.
The AI application for calendars and meetings, can point out with reminders, that the meeting is coming up, which can be from an alert to an email, a text message or vocally, or all of these.
If any information for the meeting is missing, or required last minute, the AI can be tasked with collating this or locating this information at the meeting, or in last-minute situations.
Location information on a server can take more time than we would like to admit too, this is because every time we access server files, and they display on screen, the server is processing, and due to the number of previews or file details loading up takes time, we have folders in sub folders in more sub folders.
AI can bypass this process and locate file/folder names far quickly, and remotely if needed.
Time which can be lost while waiting for an urgent email from a contractor or structural engineer, or planner, can be better used.
AI can work as a watchdog for an Urgent email, this will inform you when the email arrives and keep alerting if required, until the email is read.
These are some of the basic applications for an Architectural Assistant, which could be expanded as practice requirements identify future applications they wish to apply to the AI assistant role.
BIM Library
Similar to folder and file searches, AI can be used to manage BIM component libraries. Something which would be more useful which AI could be used for, is to create a locally hosted website directory to manage and view BIM information and preview the components. This would then allow the AI to search for the required components quickly and give the exact preview and file location so it can be applied to a model.
Revit
AI can be added as an add-in app to Revit, which allows it to be taught how to revise drawings as simply as applying a revision.
Example “revision AA to sheets 1000-1300” all sheets which have been created between the sheet number ranges, multiple ranges can be added.
The same style of ranging can be done to get the AI to publish in PDF and DWG using a sheet range, with one simple text command.
This style of input is only text at the moment; I am working on the development of making it work as a vocal command.
There is a possibility of AI working with DYNAMO on Revit, this would possibly work where AI can input commands and even create DYNAMO scripts.
AI would need to be taught this.
Web Information
Just like the commercial AI’s available online, a locally hosted AI can search the internet for information, provided it is given permission to search for this information.
The advantage of this is that, unlike a commercial AI, this will not record the information or use tracking cookies.
This protects the company in the very way that cybersecurity requires.
Hosting an AI model in-house allows you to limit your information vulnerabilities, Which all commercial 3rd party AI’s have.
Wellbeing
This would be a staff permission-use-only.
AI can be programmed to assess an individual's stress levels and flag potential problems.
For some people, a conversation is all that is needed to stop them from slipping down the slippery slope. Stress in the workplace can cause absences, lower morale generally in an office, and cause more stress on others. If identified, a simple conversation, “how are you?” is all that is required.
Think of it as giving a person who is silently screaming a voice which can be heard. As busy individuals, we don’t always notice others around us.
AI should never work as an office spy to assess productivity.
For health and wellbeing, and picking up on individuals struggling due to stress for whatever reason, is a tool which helps productivity in the company as much as the individual.
Future Progression
Potential future applications:
1. Generative design: Use AI algorithms to generate multiple design options based on specific parameters, such as sustainability, cost, and aesthetics.
2. Data analysis: Leverage the cluster's processing power to analyse large datasets related to building performance, energy efficiency, and occupant behaviour.
3. Visualisation: Utilise the cluster's graphics processing capabilities to create detailed, interactive 3D models and simulations.
4. Machine learning: Train AI models on architectural datasets to predict design outcomes, identify trends, and optimise building performance.
Some potential benefits include:
· Increased design efficiency
· Improved sustainability and energy efficiency
· Enhanced collaboration through data-driven decision-making
· New opportunities for creative exploration and innovation
To get started, consider the following steps:
1. Define project goals: Determine specific use cases and objectives for the AI cluster.
2. Develop a workflow: Integrate the cluster into existing design workflows and software tools.
3. Train AI models: Collect and preprocess datasets, then train AI models to perform specific tasks.
4. Test and refine: Evaluate the effectiveness of the AI cluster and refine its applications.
Architects could explore various AI frameworks and tools, such as TensorFlow, PyTorch, or Keras, to develop custom solutions. Collaborating with AI experts or researchers might also be beneficial.
What specific aspects of integrating a Raspberry Pi 5 AI cluster into an architect's practice.
Pro's and Con's
Pro’s
· Supports existing staff
· Increases productivity
· Potential to reduce staff stress
· Cyber secure
· Trainable to suit business direction
· Scalable
· Endless potential
· Capable of long-term and short-term memory
Cons
· Requires someone with the skills to manage
· Will require expansion the more you add to it
· Training time to learn can be extreme
· During downtime, staff will be required to know how to do tasks manually
· Potential to increase staff stress
Conclusion
AI has a lot of potential, presently and into the future, in Architectural Practices.
As an Assistant, as an office wellbeing monitor for staff, efficient searcher and a model tester, clash detector and many other potential possibilities.
Should an Architect Practice invest in AI?
Well, that would be down to the practice, not only is it about the advantages listed in the report and whether it is worth applying to the Architects Practice, but can this be commercially beneficial to the company as its own revenue stream.
Early adopters of AI will have the advantage of selling their progress to other practices and even offering the IT support in updating and maintaining their AI.
Providing a potential extension to any Architectural Practices which take on AI in the near future. This will also provide opportunities for shaping the future of AI.
Will this be easy?
Depends on a clear direction of progression laid out. Possible additional options not covered in this integration report, and learning time to train AI. Nothing is easy, and would require a set plan to get training completed, one area at a time. Trying to train more than one task at a time will only cause difficulties for the AI.
Key things to remember are to back up the AI often so as to not lose the progression and trained skills it has learned. Assign someone as an AI manager to ensure training, development and updates are completed.
An AI manager could have an AI running at home, which could do the training for the office AI. This could form possibilities for a structure of AI development program.
Possible commercial issues
The advancements in mini PC’s might create a dilemma, as the improvements are on a large scale in every new model which is released.
This might create a problem in AI services for other practices in the future. This can be identified in every new release of the updated mini PC’s as they come out.