Ten Questions With The Founders of Semanta.ai

On the PPW Podcast, we talk a lot about the future of real estate search and more than once, we've mentioned a future in which house hunters will have an "always on" search agent working for them, which they can interact with conversationally.

Semanta.ai is the first product I've seen that does this. It takes user queries via a conversational interface, and then it does its work in the background, sifting through listings from all the portals to find matches. It emails you when it finds good matches and is upfront about when a property is a full match or a soft match without pushing paid listings on you.

I was impressed by how Semanta is able to match so many more criteria than all the big-name portals in its native Switzerland, so I reached out to the company's founders (who wish to remain anonymous for now) to learn more about a product which just may be a vision of the future...

What is the problem that Semanta.ai solves, and what's the theory behind it?

Semanta is a specialized AI real estate assistant that provides comprehensive property data, market analytics, and automated search features. It helps all home seekers, whether professionals or non-professionals, achieve better results while spending less time scrolling on various portals.

Online marketplaces are the primary method for property search, standardizing listings into a single database. Their business model relies on paid advertising from owners and agents and free services for mass consumers. To maximize user traffic and engagement, they must remain neutral regarding advertised properties. As a result, their interfaces use simple, generic filters, forcing users to manually sift through listings daily for months or even over a year. Marketplaces have no inherent incentive to reduce the consumers' screen time.

In contrast to classified portals that aim to maximise user scroll time, Semanta approaches the problem from the seeker's viewpoint, aiming to minimize screen time.

What made you want to build an ‘always on’ search agent?

The idea of a private and trustworthy search assistant is not new. Generic search engines like Google and nowadays ChatGPT/Gemini can handle "one-time" search questions well. However, certain tasks require repeated searching over time, such as finding a home, an investment property, a vintage design chair, or specific company news.

While simple search alerts exist, and ChatGPT offers internet search functionality, we hypothesize that these tasks benefit from access to domain-specific, curated data streams, which is why specialized portals coexist with Google. For complex decisions like property searching, basic alerts are insufficient. Recent AI advancements now enable personalized search assistants that continuously search on your behalf, a capability that was technically challenging even five years ago.

What is the business model, and who are your customers?

Ideal customers of Semanta are individuals, professional and non-professional alike, who want an AI search companion that truly works for them, spending less time scrolling and constantly checking listings.

Our vision is that these AI search assistants will be a paid service, similar to a utility. We hypothesize that data and analytics will be provided as add-ons to empower these search agents, creating a transparent data ecosystem. For example, the AI assistant could access listed properties, valuation APIs (MCPs), personalized mortgage offers, property history, and other relevant services.

For now, I imagine it’s mainly professionals or property investors who use Semanta.ai. How long do you think it will be before ordinary home hunters turn to conversational, always-on search?

Semanta can help both professional and ordinary home hunters. The classified portals' 25-year-old free information search model (attention economy) is changing. ChatGPT has dramatically shifted user expectations toward having a natural, paid AI search assistant. Given that buyers spend 6 to 12 months on average searching for a house, they value this support.

Additionally, Semanta can be used by professionals like agents, relocation advisors, or investors for daily search and due diligence.

For individuals who search infrequently (every 4-5 years for renters, 10-15 years for buyers), one adoption strategy is to offer similar services across several verticals (cars, jobs, travel). This horizontal dimension can build trust between the individual and the personalized agent.

Most Swiss portals don’t have that many fields of information. How do you make sure that if a user asks for features like ‘south-facing garden’ or ‘off-street parking’, you don’t deliver false positives?

The challenge of false positives or negatives is a common machine learning hurdle that requires additional AI agent evaluations to prevent hallucinations. We address this with three strategies: Context engineering, statistical evaluations, and, most importantly, enriching the core listing data.

Semanta's key idea is that listing data alone is insufficient. Listings are enriched with additional analytics, utilizing multi-modal AI to answer specific feature questions by analyzing property texts and images.

While some existing portals offer shallow search inputs, Semanta provides every user with a uniquely assessed, individual search profile.

What is Semanta.ai’s biggest challenge?

If we pursue a B2C model, the biggest challenge is scaling adoption against the prevalent expectation of free end-user services. While we acknowledge that ChatGPT and Gemini have significantly shifted the willingness to pay for AI services, it will still take time for consumers to grasp that Semanta provides AI-driven advisory capabilities that transcend current online marketplace offerings.

How have the portals and agents reacted to what you’re doing?

We are at a very early stage and have just begun our marketing campaigns. Discussions with team members at SMG in Switzerland confirmed the value of our service for home seekers. However, our hypothesis is that, given their strong market position and recent IPO, SMG will currently overlook the seeker side. They appear focused on revenue growth through price increases and new analytical services for advertisers. 

Which products or companies have you taken inspiration from when building Semanta.ai?

More than anything, Spike Jonze's movie Her, specifically the simplicity of the main character's pocket device. 

An interesting recent development is the San Francisco-based company Yutori and its product, Scout. While Scout offers broad search capabilities, our approach at Semanta is a vertical focus on real estate, aimed at building a valuable, data-driven advisory service within that sector.

What’s the priority for the business now? Is it fundraising, building out more features or marketing?

In terms of product, our priority is to ensure the high quality and advantageous nature of our two core services: Property Check and Search Profiles. We are prioritizing customer satisfaction and validation that Semanta outperforms existing online marketplaces before expanding with new features like agentic workflows.

We are also maximizing the use of AI coding assistants to have a lean development team and accelerate delivery, especially for rapid prototyping. The cost of software development is now measured by consumed tokens, not the number of human FTEs.

As for fundraising, we do need additional funds for large-scale marketing to introduce Semanta to the public. Early investors who believe in Semanta’s vision are critical for this growth phase.

We're also looking to expand geographically. While based in Switzerland, we are eager to test Semanta in other countries, contingent upon securing additional resources or partnerships.

What is one thing you think the real estate marketplace industry should be talking about more?

Online marketplaces are locked into the paid advertising and user traffic management model. AI search assistants are emerging and will replace the manual search process, fundamentally changing the concept of the user interface. Marketplaces must embrace this shift, as the traditional measure of traffic count may become meaningless. They will remain as essential data aggregators and maintainers of listings, but their focus must shift to enriching the data to be valuable for AI search assistants.