Summer turns Autumn, platform development continues

What we’ve added to the platform since I last updated.

Welcome to an update on ‘How to build a merchant bank with $10 or less’. Jokes aside, we are breaking down the elephant into bite sized chunks against capabilities, utilising the platform on the way.

Since we last provided an update:

1. The platform is live and being used by staff since late January over Hubspot.
2. 4 third-party API’s are now integrated into the Platform with various ones more active than others. 8 are on the horizon, which will push my budget past $10 unfortunately.
3. Platform is pricing credit for opportunities dynamically and analysing them.

As a showcase, I wanted to share some of the capabilities we’ve added to the platform.

1. Credit Pricing now working in seconds.

Pricing credit is usually done by cardigan wearers using an Excel. It’s not fundamentally difficult and operates in a similar way a Dive Chart does for a Scuba diver. Go down too far for too long and you’ll run out of oxygen (capital) and wont make it to reporting season.

Our credit pricing engine is built on analysing the borrower strength, the property fundamental data-sets (i.e. location, asset type, zoning), the loan economics (i.e. LVR, amounts), our portfolio (i.e. concentration on geography, product type) and macro-data (i.e. BBSW, property trends, demand) and bundles all this together to enable a customised interest rate suited to the borrower ask at this very time in literally a second.

While it doesn’t seem like much, instead of doing pricing points against a table that goes up in 20–50 basis points, the AI is learning how to enable dynamic pricing to 10 basis points immediately. At this point, technically this is a full verified borrowing entity being offered an actual pricing point, not a guideline or ball park. This API has been custom built and is its own module.

From a technical perspective, I’ve built a model which connects to a series of middleware when an opportunity is created. The code below shows my middleware modules. I could build more, but it’s a start.

module.exports = {
  creditLimitsCheck,
  calculateCreditLimitStatus,
  borrowerStrength,
  securityCheck,
  calculateLVRPercentile,
  treasuryCheck,
  productCheck,
};

2. Lending documentation enabled via Platform in seconds.

Most Lending Management Solutions (LMS) offer documentation storage and even some are enabling tables to be created using data inputs.

Few are enabling a suite of lending documents and credit committee approval documents being auto-generated on-demand, and then being stored into word and pdf version in a cloud storage solution ready for final review before being sent out to the borrower or legal firm (depending on the step).

We’ve consumed the legal documents from our legal firm and embedded these into our Platform so when a borrower is entered and the loan parameters are put into the system, the documentation is autonomously developed.

For this, we are not re-inventing the wheel, but using PandaDoc as its the only documentation platform with a custom json accepting API and security that meets our expectations. PandaDoc helps with sending data and managing document templates.

3. Automated Borrower Verification

We’ve integrated CreditorWatch and about to start with Illion around verification of borrower entities, trusts and SMSFs. As one of our Relationship Managers creates a borrower in the platform, they can select to run a creditor report, which we actually do often as it’s a great tool for reviewing directors and their business interests. You can then look those up as well to ascertain who we are dealing with.

The Borrower verification occurs either at the start of the data entry into the platform or at the time when the term sheet is generated. The information is recorded in the DB entry of the user and the KYC functionality handles ‘data aging’ so every 90 days the system runs another check to see if anything has changed.

4. Using ChatGPT for learning the credit committee functions

We have used ChatGPT to assist with actual coding and making modules. It’s saved a Biblical amount of time and effort. However there are other functions we can use including developer documentation and test scripts.

Additionally, we’ve added the Chat GPT to the API to learn. Each time we do a loan, it learns the parameters of the loan and why did it get approved, or why not. At the moment, it’s an exercise to see if this is function can automate initial components of the credit committee work and move to a single approval process with the pre-approval being automated, but I think we can push this forward.

To enable a trade on the Nasdaq or NYSE (official start to finish) used to be T+4 days, and now with a range of new companies and processes it’s closer to T+1 or T+2 depending on the timing.

Right now, most commercial loans are enabled (official start to finish) in around 2–3 weeks. Date back BC times, and while don’t quote me on this, I’d think the time would be at worst around the same or even faster to less than a week with travel being the biggest challenge. The point I’m making is why not use a combination of intelligent technology and robust processes to get the lending to T+3–4 days. For this, you need a strong platform, strong capital stack, and strong credit risk framework.

5. Slack integration — command centre

Many people see Slack as a comms platform. I see it as a command centre. For example, I’ve integrated the deployments and testing cycles to a slack channel. Therefore we know the deployments went well, or didn’t on slack, not logging into a platform to testing the site.

The same can be done for user creation, AML CTF Checks, Compliance checks etc etc. Therefore for most things we do, we use Slack. Yes, it can get noisy, but Slack has features for this.

A simple way to look at it, with a new feature or commit released, I’ve got ChatGPT writing notes, and this is populating on a Slack channel so the entire team knows about it. I don’t have to do a release notes to the team or schedule a session on what I’ve released — it happens autonomously which saves time on things that don’t push us forward.

Here is an example below of a new partner being created in our DB and here is the notification in the Slack API. (BTW — we’ve named some of the bots after James Bond actors, hence the Roger Moore reference).

BOX Storage enablement

We like using Box.

Box is great for what we have to do as it’s strong in moving between Google (Slides, Sheets) and Microsoft files (PPT, Excel, Word). Dropbox does not do this well and since we actively work on files, we needed something that is flexible.

The Box SDK is not bad, but the documentation when you want to use it is rigid so customisation took a while around direct connection with the developer team at Box.

However we managed to get it create a new folder array each time a borrower or partner is created and trigger specific events from this. This helps with automation and file controls as now everything we create is replicated in storage. The screenshot below is a basic set of folders created in our drive created by our API. While this doesn’t look like much of an integration feature, its a god-send for consistency and collaborating on documents.

Whats next?

We’ve each of the features mentioned above are working, but will need to be further developed in order to maximise their integrations.

Next on the list is:

• Price Finder — For property look ups from the API to then enable dynamic credit papers for the credit committee.
• Bloomberg — For macro trend data.
• Illion — For further AML CTF and KYC.
• Xero — For accounting integration.
• PPSR — For security registration for loans.

From an Epic feature level, the following are the next epics I’m building.

• Treasury & Portfolio Management
• Compliance Management
• Product Management (on platform rather than API)

The next update will be in and around these points above. Remember, I’m currently a one-man-and doing this and while we will get engineering resources into this eventually, the speed of which I can move with the tools available is helping trail-blaze. No scrums or agile meetings for me -> I do it on the fly.

Thanks for following.