It's been a busy month for Mortgage Tech.
First of all huge congratulations to our fellow Mortgage Industry Tech Companies on launching their solutions this week. I am sure Product Owners, BA's and Engineers have worked tirelessly to bring this to market. It might sound strange to say, but at Instamo we're genuinely thrilled to see other technology companies launch their own mortgage application solutions, to embrace this technology shift rather than fight it, it means we're on the right track. Building and maintaining submission solutions for multiple companies at enterprise standards, while respecting lender portal policies, is an incredibly complex task. I believe healthy competition in this space drives innovation, and innovation drives better consumer outcomes. Hopefully we see more solutions emerge over the coming months.
When we started Instamo, we wanted to change how things were done. I've always worked on the basis that the simplest solution is often the best, which is why we took our solution back to fundamental principles. While we use AI to help us operationally and of course design innovative solutions, unlike other solutions in the industry, we don't pass client data within a million miles of LLM's. We made this decision from Day 0, for reasons primarily centred around Data Security and what these LLMs do with the huge quantity of client data you are providing.
The underlying reason was simple, we're building a solution where brokers must trust the workflow behind it, and trust that 100 times out of 100, Instamo will generate the same result. I've always felt this is the main reason why Lender API adoption has been so low, and we're determined to solve this problem.
Unlike our industry peers who may have explored API Platform solutions in the past, we're not turning our back on Lender API's, we firmly believe they are the future. However to drive real adoption means confidence and trust in the process behind it, unfortunately confidence and trust can't be built overnight.
We've all seen the meme where a user asks OpenAI to regenerate an image 100 times, and see how it's morphed. Now imagine a mortgage application.... Our concern from Day 1 was exactly this, data accuracy is the most important item when it comes to Mortgage Applications. That's why at Instamo we're building confidence that using Instamo to autofill a Mortgage Application will always produce the same result.
While it would have been easier (and much cheaper short term) to just build on top of Claude - I've learnt the hard way over my career, you don't build your house on borrowed land.
The AI Bubble
The next 12-24 months are going to be tough. It's not a question of if the Mortgage AI bubble is going to pop, it's a question of when. AI tokens might be cheap now, but when SaaS companies built as OpenAI/Claude wraparounds have to hike their prices to absorb the massive losses, that Claude are incurring now and soon look to recover, that's when the music is going to stop. Let's look at an example...
That's not what's concerning me most…
- Most solutions in the market today are not GDPR compliant, they're built as Claude wraparounds.
- Some have raised huge sums of funds or have been built using vibe coding tools. They will soon find the commercials they offered on 3 year pricing can't be honoured.
- More often than not, an advisor can achieve the same outcome using ChatGPT / Claude CoWork.
What happens next
Forgetting for a moment the number of GDPR breaches that occur daily with advisors pasting full bank statements, fact finds and other into Claude, my concern is that coworking tools will be the new norm. A few thoughts on how this plays out:
- Used responsibly, this is great - for companies that deploy effective IT policies to ensure that's the case. Running CoPilot on Azure, brilliant concept, tied to InTune, Defender and GPO's and you've got a great setup.
- However when the cost of tokens start to rise, and advisors are hooked on the drug, I believe we will see what we've always seen - people will start to move to Chinese based (significantly cheaper tokens) that will take GPDR breaches to another level. Think DeepSeek 2.0.
- Firms that have made huge operational cuts to leverage these LLM's will find they are rehiring and hiring quickly to fix the mess that these software companies have made.
- ICO fines will be issued, and the exercise will be a loss maker rather than a way to shore up the balance sheet.
Maybe I'm too doom and gloom but I believe this is a catastrophe waiting to happen.
My advice
So my advice to firms is simple at this stage:
- Understand your data, and make sure you choose a CRM that supports your data goal. AI on it's own is useless without a clean repository and well structured database.
- Understand what you're trying to achieve, and hire real technical advice to help you put together the right solution. More often than not, AI is overkill, there are plenty of open source / code based solutions that can deliver the exact same result.
- Work with firms who want to integrate with third parties. If your CRM is only building their own tools and is unwilling to work with third parties, that's a red flag. It means you will be heavily leveraged on a solution that (when the bubble pops) moving away is incredibly difficult and expensive, and getting data out is going to be much harder than getting data in to begin with.
- Finally - understand who's leading and who's copying. Research a company's past, understand what's changed from that system over the past 5-10 years, examine the leadership team and ethos of the company, what they're building and why. If they're a data business, understand what they're doing with your data and why. Where does their revenue come from, and what might they do with your data.
The future is bright for the mortgage industry, but we're slowly walking into a minefield, and it's important you have your map to hand...