An AI Utility Belt for Commercial Real Estate + 4 Ways to Use ChatGPT o3 for CRE
Welcome to the May 2025 Edition of Chat CRE
Welcome to the May 2025 edition of ChatCRE - Your monthly dose of updates on AI tools, and practical ways to put them to work in commercial real estate.
This month’s edition covers: A powerful AI “multi-agent”, a hilarious use of AI video marketing, and some great tips on how to use ChatGPT’s latest model from a retail commercial real estate broker actively using it in the field (including some uses I’ve never even thought of.)
Let’s dive in.
TL;DR
Genspark: Offers a utility belt of AI tools & features, combining access to LLMs like ChatGPT, Claude, Gemini, and tools for web scraping, generating presentations, websites, images and video all with AI. Check out the deep dive on CRE use cases to see some ways you can use it.
Chad Griffiths’ Baby-Faced Book Launch Videos: A hilarious (and effective) use of Hedra + ChatGPT to promote industrial broker Chad Griffith’s new book. Not just funny—scroll-stopping CRE marketing inspiration. See the actual process and tool he used to make them.
AI in the Field - Real-World ChatGPT o3 Use Cases: Learn how retail broker Andrew Poncher is using ChatGPT o3 to automate prospecting at conferences, tenant scoring, market analyses and more.
P.S. I’m booking public & private AI presentations and training sessions for Q3 and 4 of 2025. If you’re interested in having me present to your team, company, or association, you can find more information here, or just send me an email and let’s connect.
The Rise of the Multi-Purpose AI “Agent”:
(I put “agent” in quotes because the term is thrown around so much it seems meaningless, but it feels warranted here..)
Over the past few months I’ve started to see more and more “multi-agents” popping up - AI tools that combine some of the best features of stand-alone AI tools into one platform. I’ve battle-tested a bunch of them, and many have fallen flat, but I’d gotten enough positive feedback from CRE folks about this particular platform that I had to block out some time to try it out:
Enter Genspark:
Genspark is billed as an “all-in-one AI agent”. Not only can you use Genspark to chat with most of the latest models (ChatGPT o3, Claude 4, Google Gemini 2.5, etc.), under the hood it combines a slew of AI features that were previously only available in stand-alone tools. It has separate “agents” that can scrape the internet for market, property or tenant data, analyze the information, create spreadsheets, presentations, videos and websites, conduct Deep Research, or fact-check information. Finally, it offers a “Super Agent”, that can do a lot of the above with one prompt.
Think: “Take this PDF of property data and map it to a spreadsheet, then scan the internet for similar properties in the same area and add them to the spreadsheet, search for businesses located in that building and add them to the spreadsheet then provide an analyses of the data based on the building information, then create a presentation and short video summarizing the intel.”
Here’s a quick rundown of the most celebrated features in Genspark:
Deep Research – Synthesizes information across the internet and public documents with live citation, bias checks, and screenshot-backed claims
AI Sheets – Import documents with data, map it to a sheet, analyze it, or just have Genspark SEARCH the internet for additional data (think web scraping, enrichment, row-by-row validation/fact-checking)
Sparkpage Publisher – Publishes live, web pages with images, graphs, videos, etc. these become the “hub” of information that other agents can access
AI Slides – Generates slide decks or reports automatically from information in the Sparkpage, with visuals and styles based on your prompt
Video Generator – Produces short videos using text prompts and images (you can use a variety of AI video generators just like you can with the chatbots, it’s not Genspark’s proprietary video creator)
Super Agent – Chains multiple agents into a sequenced workflow (like: research, analyze, draft summary, build slide deck)
Phone Call Agent - Genspark DOES have a feature that can make calls for you when you’re searching for or uploading contact info (full transparency, I don’t love AI phone calls and haven’t used it)
So why would you actually want to use Genspark? To me, the pitch is pretty clear: If you want to do everything listed above, but you don’t want to subscribe to 10+ standalone tools for chatbots, research/data analyses, video, presentations, etc., Genspark represents a real opportunity to do them all out of one system starting at $25/mo.
But don’t just take my word for it. I tested Genspark across a few commercial real estate use cases. Check out the breakdown below.
Spoiler Alert: The standout feature of the platform for me turned out to be AI Sheets.
1. Prospecting: Search Google Maps for Businesses + Contact Info
Agent: AI Sheets
The Prompt:
Please search Google Maps for all the doctors offices you can find in Magnolia, Texas, and the surrounding areas, and put them in a sheet with columns for the following: - Name of the doctor’s office - Street address - City or town - Phone number - Source link - Confidence score At the end, review the sheet and fill in any missing cells.
Watch it Work:
How it did:
Really well. It returned 19 accurate businesses with contact information and website links, confidence scores, no blanks in information. When I asked it to “continue searching,” it expanded to nearby towns. Is it exhaustive? Maybe not—but it’s fast, and complete enough to start hitting the phones if you’re calling on mom and pop businesses.
2. Web Scraping, Analyses & Presentations: Research & Build a Competitive Set Report from Crexi Listings
Agent: AI Sheets (But I probably could have done it with one prompt in Super Agent)
The Prompt:
Please make a list of all the Industrial Properties for Lease, located in Portland, Maine listed on Crexi.com. Make sure they are listed as properties FOR LEASE. You can find them at this link: [INSERT LINK]
Please make sure you include a column for: Property street address, Available square footage, Lease rate, Type of lease rate, Description listed on Crexi, Link to the source, Confidence score of the information, Photo (Add a photo of each property)
Once the spreadsheet is compiled, please review to make sure there are no blank cells, if there are any blanks listed, find the information and add it.
Next, fact check the information, double check that all of these spaces are actually LISTED FOR LEASE on crexi.com
Watch it Work: This one took a while to collect the data, so I cut out a good chunk in the middle to keep the video short.
How it did:
Much than better than expected. It scraped listings from Crexi and compiled the data.
After collecting the intel, I asked for an analysis and a presentation, which turned out excellent. You can see the final presentation here. Clean layout, listing photos embedded into site, arguably a much better comp set presentation than I often see from brokerages — And honestly better than what you can make in something like Gamma for this particular purpose.
Note: Gamma is still an amazing AI presentation tool. You can learn more about using Gamma to create commercial real estate marketing collateral here.
Bonus: This is just ONE of the web scraping tests I ran in AI sheets. I was also able to get it to build an inventory database of industrial properties, and one for local industrial tenants over 50,000 SF. I know what you’re thinking - No, it can’t get this information from Costar. But you’d be shocked at how much property and tenant information is published in the news. Genspark can find and access all of that information.
3. Marketing: Research a Property & Design the Marketing Materials
Agent: Super Agent
The Prompt:
Create a Sparkpage for 750 Warren Avenue, Portland, Maine including: - Parcel data (lot size, year built, zoning) - Suite mix (unit sizes, current availability)- Loading specs and clear height - Recent industrial sales comps nearby - Nearby logistics-related tenants - Drone or map images - A short summary of what matters about this property for potential industrial users
Then, generate: - A 10-slide marketing brochure using AI Slides - Make sure it includes actual photos of the building, A drone style video orbiting the building - Upload the original photo to RunwayML to generate the video, A listing microsite that can be password-protected and shared
Watch it Work:
How it did:
In two words - Mostly awesome.
The research was overall very accurate (I know this property). The brochure came out fantastic, with almost zero creative direction (and in case you didn’t catch it in the video, the presentations it creates are fully editable). The only downside is video didn’t work out via the Super Agent, instead of creating a video of the subject property, it just imagined a similar property. So I just went directly to the video creation agent and made it there directly: Uploaded an image of the property, gave it a simple prompt, and had an aerial video in under a minute. I selected Runway ML as my video model, but there are a lot of video tools to choose from. Here’s how it turned out:
The final output: Property research, marketing document, and a password protected property website, all executed with 1 prompt.
The Downsides to Genspark:
A lot of the web scraping tasks took a while, the videos are obviously sped up and had some segments clipped out. But you can queue up the agent and have it run in the background while you’re working on something else.
The Super Agent is cool, but not all that it’s advertised to be - The video generation obviously didn’t work when I tried it out of Super Agent.
The web scraping DOES NOT all happen in one shot. For some of the tasks I tried, like building property or tenant databases, it would find information for 5-10 properties/ tenants, and I’d need to hit “Continue Search”. If you try this with some other AI platforms, they’ll basically search again and return the same information. Genspark’s AI Sheets keeps going and only adds new information to your database, because it references the sheet as it searches for new information.
Credits: Genspark isn’t an all-you-can-do buffet. $25/mo buys you 10,000 credits. Most of the examples I mentioned above seemed to take 100-300 credits. But if you’re giving it giant documents to extract data from, that’s one task in particular that could absolutely MOW through credits. What Genspark can do is super impressive, but you’ll need to try it out to see if 10,000 credits/mo will give you what you need. You can try it for free with 200 credits.
Final Thoughts:
You probably noticed a theme in the use cases above: I REALLY liked the AI sheets. Genspark can do a lot of things - But the ability of AI Sheets to let you upload YOUR OWN intel in any format you want, have it mapped out into structured data, and then use AI to search and fill in the gaps, or even fact-check the information, is something I haven’t effectively seen done in any other platform. At absolute minimum, it’s impressive to see an AI tool that can build a database of tenants, properties, availabilities, analyze the data, and then create analyses and presentations from the intel all from one prompt.
SO: If you’re looking for an AI tool that can execute A LOT of CRE-related tasks with one subscription, Genspark is worth at least trying out. However if you’re already drowning in AI subscriptions and are only interested in tools that add a new arrow to your quiver, I think you’ll be impressed by the AI sheets. You can sign up for a free trial and give it a whirl here.
Chad Griffiths' Hilarious Book Marketing Strategy: AI-Videos in Action
There’ve been a lot of updates in AI video generation over the past month or so - I’ve been deep in an AI automation rabbit hole and haven’t experimented with these new features as much as I’d like. Lucky for all of us, industrial CRE expert Chad Griffith has been taking this bull by the horns, creating genuinely hilarious AI-generated videos to market his new book "Industrialize." See Exhibit A below:
(Chad’s recent podcast with Bob Knakal and Rod Santamassimo.. Obviously transformed as babies.)
Perhaps you’re thinking “Topher, why is this in the newsletter? Should I start making AI-generated videos of myself as a baby?” Maybe, maybe not, but it’s a FANTASTIC example of leveraging AI to create content that cuts through the noise. Chad’s videos are funny, impressive, and they’re easy to make.
If you’d like to try it yourself, Chad is using a tool called Hedra to create the videos.
Here’s how he did it:
1. Chad generated images of himself as a baby using ChatGPT, by uploading a photo of himself, and using the following prompt (provided with Chad’s blessing):
Generate a photo realistic image of a baby version of the person in the reference photo. Preserve key features including eye colour, skin tone, hair colour, eye shape, eyebrows, nose structure and overall facial proportions. Style the babies hair in a child-appropriate version of the adults hairstyle. The baby should have a calm, neutral expression with a closed mouth (no smile). Dress the baby in clothing that mirrors the adult style - such as a formal, white-collared shirt. Use soft, flattering lighting and place the baby against a colourful modern background that harmonises with the original photos aesthetic.
2. Upload the image to Hedra
3. Add the audio of you (or any other babies in the shot) actually talking
4. Let the AI work its magic
You might not have a book launch coming up that you need to promote, but I’m a huge believer in using video to get people’s attention. Think about everything you send out that potential clients hit the snooze button on because they look just like.. everything else. Market updates, new listing or transaction announcements, deal sourcing requirements - How can you use tools like Hedra to create entertaining videos that tell a story and get your message heard? To be clear, the videos don’t HAVE to have baby’s in them, they just need to stop the scroll. You can learn more about creating AI-generated videos for CRE properties in this recent newsletter edition.
AI in The Field: Using ChatGPT o3 for Commercial Real Estate
Tips provided by Andrew Poncher: Senior VP, Director of Leasing at Renaud Consulting
Over the past month or so, you’ve likely seen a new model available in your ChatGPT account (if you have a Plus or Teams account) - The o3 model became available to all paid users on April 16th, 2025. The models rolling out across all major LLMs are so impressive at this point, it can be confusing to keep track of what’s actually improved. But o3 update is a significant leap forward.
I recently spoke on a live AI workshop with my friend Andrew Poncher, a retail specialist, and he mentioned he’s absolutely living out of o3 to run his retail brokerage business. So much so that he upgraded to the $200/mo ChatGPT Pro so he could have unlimited access (you get a limited amount of o3 credits with the $20/mo ChatGPT Plus).
Andrew was kind enough to send me some specific examples of how he’s using ChatGPT o3 to save time in his brokerage business so he can lean into the activities that actually generate more deals, armed with better data. I hope it will inspire you to do the same! Warning: Some of these use cases are advanced, but you won’t know what this thing can do until you try.
Conference ROI Maximization:
Using o3's Advanced Data Analysis, Andrew's team cross-referenced the ICSC Las Vegas attendee list with their team’s LinkedIn profile-viewer data to identify overlaps between attendees and people engaging with their team online. The model scored prospects by engagement and brokerage opportunities based on their portfolio, then drafted tailored intro emails to each prospect, all in one chat. They uncovered 37 "warm" prospects that would have been missed manually. Meetings were booked!
Lease Abstraction That Finds What Matters:
For a grocery-anchored center, Andrew’s team used o3 to process a 92-page lease packet: key dates, expense pass‑through language, and escalation schedules. The model even surfaced a mismatch between renewal‑option rent and an outdated CPI clause before their attorneys caught it! These are the kinds of details that can save real money.
Tenant‑Risk Scoring via Custom GPT:
Andrew’s team built a private “Tenant‑Risk Scorer” Custom GPT that they can feed P&L snapshots, foot-traffic data, and public credit scores, then returns a 0‑to‑100 stability rating plus red‑flag commentary for vetting potential tenants. The scores are fed directly into their LOI approval memos. If you’re not sure what a Custom GPT is - Check out this newsletter edition on building custom AI tools for CRE.
Automated Market Inventory Reporting:
Andrew’s team relies on up-to-date lease availability reports to service their clients. Believe it or not, NOT every space available for lease is listed on the major listing platforms. Instead of manually searching every regional brokerage portal, Andrew’s team had o3 build a script that crawls every brokerage website in their market and adds listing updates to their inventory report. Monday‑morning inventory reports now arrive before coffee.
This is actually just a handful of the use cases Andrew shared with me, but this newsletter is already pretty long. If you’d like to hear more from Andrew, you can watch the LinkedIn Live AI workshop we did together, or just reach out to him on LinkedIn!
That’s It, That’s All.
That’s it for the May 2025 edition of ChatCRE. I’d love to hear your thoughts on this edition—what you found valuable, what you could do without, or any topics you’d like me to cover in future newsletters. Feel free to comment below, send me an email, or reach out on X and LinkedIn.
If you found this newsletter helpful, please consider sharing it with a colleague or friend who could benefit from enhancing their CRE operations, marketing, or pipeline with AI.
P.S.
I'm currently booking private and public presentations on how to leverage AI tools for commercial real estate for Q3 and Q4 of 2025. If you’re interested in helping your team, company, or association learn about powerful AI tools for CRE, please send me an email, and we can set up a time to connect.