Our Biggest Game Changer
Three months ago, I had never heard of this company. Now it’s integrated in nearly everything we do.
n8n (n8n.io) connects your tools to each other so the manual steps in between just stop existing. Think of it as circuits wired between all your systems.
You receive an email from a customer. You open it. You read it. You interpret it. You decide what tool you need to help answer the email. You open that tool (eg, Excel). You slack an analyst to have them go get something. You open up the last PDF you sent them. You think about what has changed since then. You check notes about this customer’s preferences.
Claude has helped each individual step of that process, but we still had to execute each of them one at a time.
n8n connects those steps so we don’t have to intervene each time Claude finishes a step. It passes the baton from one step to the next.
If Claude is the brains, n8n is the muscle.
Two examples from the last few months.
Pensford - Cap Pricing Automation
We get over 900 cap pricing emails every day. Emails are inherently unstructured, which traditionally has meant there was no way to automate them. Before now, if we wanted to automate that you’d have to fill out a form and submit it. The form would create the structure we need (think about those online defeasance calculators), but piss you off because forms suck.
AI changes that. Instead of needing a form, Claude can interpret the email the second it lands. But then it stops and we have to manually intervene.
n8n changes that. It takes the baton from Claude and routes email interpretation to our pricing model. It passes the baton back to Claude, which gets the quotes. Then n8n takes the baton back and routes the cap pricing info to Gmail. It passes the baton back to Claude, which drafts the outbound reply. The deal manager reviews the email and presses send.
Claude was already helping with each step, but each step existed in a silo. n8n connected each step so we didn’t have to jump back in.
Here’s a subsection of how this workflow gets set up in n8n.
This took about two months to build, but now it saves each deal manager over two hours a day. That's 600+ hours a month. This is undoubtedly the single biggest efficiency gain we've ever had and I struggle to envision something ever topping it.
When you email Pensford asking for a cap quote, the reply email with your pricing is presented to the deal manager less than 60 seconds after hitting our inbox.
LoanBoss - Onboarding 100 loans in a Day
We spent the last two years building a proprietary AI abstract tool. First pass on the nastiest construction loans is 95% accuracy and 4 minutes. Agency loans are a layup.
But each step existed in a silo.
Organize the docs. Flag missing docs. Reach out to customer to request additional docs. Rename each doc for consistency. Create blank deal in LoanBoss.
Abstract the loan agreement, the note, gtys, JV Agreements, etc. The full closing binder.
Build out the property info. Address, sq footage/units. Occupancy. Upload appraisals and BOVs.
Set up property accounting integrations. Build out covenant tests.
QA all 400 fields.
Separately QA all 400 fields again.
Upload into LoanBoss. Mark as “live”. Repeat for every deal.
For a customer with 50 properties, that was about a two to three month project. Organizing the docs was the most underappreciated aspect of the whole process. That alone could take 2-4 weeks. Then another 4-6 weeks to abstract.
We just did a new 90 loan customer in one day.
Just as in the Pensford example, n8n connected all those steps. We don’t even see the deals until they pop up for QA when one of our in-house specialists reviews everything.
Here’s how a typical deal looks in n8n as it processes through each step. Note how several steps are measured in milliseconds. The longest step is less than 2 seconds.
These used to be measured in hours and days.
We abstract 100 loans a day without blinking. Throw more tokens at it and we'd do a thousand loans if we had to.
A 100 loan portfolio used to take 2 to 3 months to onboard. Now it's a few days, and most of that is a human expert QAing every field. We're close to letting customers onboard their entire portfolio in a single day. A three-month project is about to become an impulse buy.
Sounds Great, but How Much Technical Expertise Was Needed?
Neither of these needed software engineers or a sophisticated IT team. They needed someone willing to sit down and connect the boring pieces they already touch every day. That's where the leverage actually lives. Not in the flashy demo, in the plumbing.
Through its AWS hosting relationship, LoanBoss does have special access to enterprise Claude models (Pensford does not). But none of that was needed for these operational improvements.
A couple of teammates sat down, knocked it out, and flipped the organization on its head.
Cost
Oh, the cost? We got the fancy enterprise account because it comes with the SOC 2 security protocol, so we paid a bit more than the base cost. Buckle up…instead of $20/mo…
…it costs $60/mo…
Do I Need a Database?
Technically no, but it helps. I told you a year ago to build a database and you ignored me. I reminded you six months ago, and you ignored me again.
Invest in setting up a database and give your team the time to build it out.
The good news is that your tech vendors are your database. This goes back to why I don’t think software will die, it will just evolve.
You used to associate software with screens, but that experience is dying. Instead, you will rely on them to be a trusted data provider. Yardi et al will just be databases that feed your own AI hub.
You wouldn’t try to build out property accounting from scratch when you can outsource it to Yardi. Same goes for the other companies, including LoanBoss and Pensford.
Pay someone else to be your trusted data provider so AI and n8n can connect the pieces. You focus on the most impactful elements.
Next week, I’ll go over the database Pensford chose and how we went about setting it up.