Talk to Your Project – AEC Business

Back in the day, when mobile phones and Bluetooth earpieces first emerged, it felt weird to see people walking down the street talking to themselves. Now, nobody pays attention. Will the same happen with voice interaction with AI soon?

Could speech replace WhatsApp chats?

Voice is often the most efficient and convenient way to communicate. Still, many prefer written chat over just calling someone, even if they chat in real time.

In Finland, WhatsApp is probably the most popular communication method on construction sites. No wonder, since its penetration rate in Finns’ smartphones is 95.6%, and above 90% in many European countries.

Using voice in WhatsApp is possible by calling or by sending recorded audio messages. How about using AI on WhatsApp to communicate with your construction project data through speech? Technically, that is already doable, but I am not aware of any commercial applications for it.

The potential of voice on the jobsite

AEC environments are notoriously harsh for traditional UIs. Workers’ hands are busy, and teams often traverse large spaces, away from screens. You can use your smartphone, but switching between apps, BIM viewers, and field forms is a source of friction.

Voice user interface offers a way to reduce friction, keep workers in motion, and bring the data to life verbally.

Voice commands and wearables enable workers to query and update information without interrupting their physical tasks.  Voice can also deliver narrated summaries or “site podcasts”, turning dashboards and reports into audible briefings, which is especially powerful when visual attention is limited.

Conversational agents offer a more engaging interaction than simply broadcasting audio.

Perhaps the morning brief will evolve into a 3-minute narrated update on site status, issues, deliveries, and inspections. And when entering the site, a worker could receive personal, up-to-date spoken instructions on their tasks for the day.

Extending applications with voice

Voice can be helpful for BIM interaction for non-experts. Imagine interacting with a BIM system and asking it to display specific types of structures, identify upcoming installations, or pinpoint clashes caused by a recent design change.

Similarly, project managers could discuss the plan versus status and projections with the PM system. The interaction would be much quicker than navigating an app’s menus or conversing with an AI agent in writing.

Some companies, like Aiforsite with its AI assistant Wendy, have demonstrated how you can use voice to ask for information about your project’s live status within their apps.

The applications are many, but you must naturally decide where voice makes most sense.

Is the tech there?

Technologies for automatic speech recognition in noisy environments exist. Training or augmenting a language model with domain vocabulary is also possible. Still, understanding fast, natural speech can be difficult for AI systems.

One potential caveat is the impact of connectivity and latency. The AI-powered voice apps will likely require cloud services with fast connections to be effective. Users will expect the voice assistant to maintain a natural pace of conversation, which can be challenging.

On the positive side, AI voice interaction will show and store the discussion in written format, which can alleviate some of the challenges.

Smart glasses technology is advancing rapidly, and you can expect usable and affordable consumer-grade hardware to benefit the development of business applications significantly.

Jobsites are typically multilingual. Recent developments in real-time mobile speech translation make voice ever more useful in personal communication.

What’s next

The evolution of AI voice in AEC will likely begin with simple audio briefings using existing data. The next step is voice query, an audio-based data lookup. Further down the line, voice can be used for data entry and voice-triggered actions.

Whether workers will eventually feel comfortable talking to an AI agent remains to be seen. Many with smart home assistants and mobile ChatGPT already do it. I predict that it will become the norm.

PS.

What inspired me to write about AI voice interfaces? I recently learned about a new service, huxe.com, developed by former Google Notebook LM developers. It creates custom “radio channels” from internet content that interests you. You can make your own or subscribe to others’ channels. Additionally, you can grant Huxe access to your email and calendar. That way, you get a personal “daily brief” from your own data, spoken by two AI voices.