Nivo has been building AI workflows for lenders as projects for a while. Why launch an out of the box service now?
Because projects are a great way to get exactly what you want, but a slow way to get the service into the market.
A project means scoping, workshops and bespoke build before anyone sees a case go through. That suits a lender with a very specific journey and a budget for it. It doesn’t suit the lender that wants to find out, this month, whether this technology is for them. We wanted to put what we had learned into more lenders' hands much faster than one project at a time allows.
The other reason is what we learned from those projects. Nothing beats the learning you get from running real cases. Every lender we’ve worked with has learned more from their first three cases than from any demo, requirements document or workshop. Once they’ve seen the assistant chase a broker, read a bank statement and write the underwriter report on their own deal, proving value gets much easier, because you’re measuring something real against your own baseline.
So, we took the standard journey that kept emerging across those projects and packaged it. The set up is free, your first three cases are free within two weeks of signing up, and after that it’s a minimum of £1,000 a month spent on AI credits, rolling monthly with 30 days' notice. The intent is that a lender can be live within five working days of sending us their essentials, with nothing to commit to beyond that.
What does Lender Collect actually do?
It’s built to help a lender grow volumes without hiring more people. The simplest way to think about it is hiring a team of admin staff whose job is to get a fully packaged case in front of your underwriters, right first time, and then to get everything signed afterwards.
It works over email, in your brand, as part of your team, and it runs in five stages.
One, broker application and product check. A case arrives by email from a broker or your team, or by API from a portal or web form. The assistant reads everything, checks the deal against the products and criteria in your knowledge base, and quotes from your pricing calculators if you load them. The output is a scoped deal.
Two, gather evidence. It builds the list of evidence the case needs from your standards, runs the checks in scope, such as credit searches, company reports and ID, asks the right people for the rest, reads each item as it lands, chases on a rhythm and keeps going until every task is complete.
Three, pre underwriting gap check. It reads your knowledge base again against everything gathered and fills any gap a first pass can miss. The output is a case that is underwriter ready.
Four, the underwriter report. It drafts the report, in our format or your own template, with its findings and flags, and emails it to your underwriter with the full evidence pack. The underwriter completes it, records the decision and emails it back. The assistant never makes the credit decision; that stays with your people by design.
Five, post underwriting actions. It acts on the decision: the loan agreement, guarantees and conditions for signature, each checked as it lands, until the case is ready for your fulfilment.
So what value does this add for a lender?
Our lenders have told us that improving the packaging quality of cases coming in from brokers would significantly increase the amount of deals that could be done and mean faster deal times.
There’s a gap between a broker sending a case and an underwriter being able to decide it and our research with more than 70 lenders and brokers found most firms achieving right first time rates of 50 percent or below. A typical case involves 15 to 20 message rounds and more than five hours of administration. Every one of those rounds is a person opening a file, finding a gap, writing an email, waiting, and opening the file again.
The assistant does that gathering, checking and chasing work, responding in minutes on as many cases as you send it. We recommend lenders target at least a five times return on what they spend and measure it against their own baseline. This could be the proportion of right first time cases, number of days from enquiry to a case being underwriter ready, chases per case, hours of administration per case, and, crucially, completed deals per month per underwriter. That last one is the key point – it’s how you grow without adding headcount.
Specialist lending is messy and no two cases are the same. How does a standard service cope with that?
This is the part I’m most excited about, because it’s what we have changed most in the last few months. We can now configure the service quickly for a wide range of markets and lending workflows, where variability is high and no two cases are alike: bridging, development, commercial mortgages, asset finance, agricultural lending.
The service handles that variability the way a good admin person does. It reads whatever arrives, works out what it is and what it means for the case, and asks for what’s missing. A case with four borrowers gets four sets of identity checks. A document that doesn’t add up, like an address on an ID document that doesn’t match the bank statement, gets flagged for the underwriter with the detail, not silently accepted.
You tell it what to do in your own words. Your product guide, criteria, evidence standards, forms and templates go into a knowledge base. Your house rules are plain sentences and a one-off instruction on a case is as simple as an email. As a lender you can also choose how much of the journey it takes on. Some lenders run all five stages from day one. Others point it at their own team first, send cases in and get a complete, checked pack back, before opening it up to brokers and borrowers when they are ready.
Have lenders had any uncertainties about this approach and how has the service changed because of it?
I can be open about this, because the service is shaped by it. Many of the things lenders were unsure about have been addressed and those changes are now built in.
The first question was simply about what it does. Lenders expected a tool that helps their team with tasks and were surprised that it takes on the whole job and produces a complete output. So, we wrote the five stages down and we use those stages with the lenders we talk to. If any of those five stages would be valuable to you, this is worth trying.
The second was what happens between signing up and going live, what it costs, and what ‘s included versus what becomes a project. That’s why the standard set up is free and defined, why a case costs between £1 and £40 depending on how much work the assistant does, and why there is a published list of what is included and what’s a project. Most asks are a rule or a document, not a build.
The third was overlap with systems that lenders already have or are buying. Lender Collect is the front door and the engine that gets a case to underwriter ready and on to signed documents. Your CRM or loan management system stays the system of record. The overlap is resolved by a handover at the point you choose.
And the fourth was trusting an AI assistant in front of brokers. The answer is control. Which messages are sent automatically and which wait for a named person is a rule that you set, in your words. Your BDM can be copied on everything that goes to a broker. Anything that doesn’t fit comes to your team first. And you can start with internal only approach.
Lenders worry about compliance and data security with AI. How do you answer that?
We hear it on almost every call, and we treat it as a selling point rather than a hurdle, because this is where a bank grade platform earns its keep.
Each lender gets their own segregated instance, hosted on AWS in Ireland with backups in London. The lender is the data controller and we are the processor, working on their instructions. AI models are used through EU endpoints under terms that mean a lender's data is never used to train them. We hold ISO 27001 and ISO 42001, the management standard for AI, as well as Cyber Essentials Plus. We have come through data protection and resilience reviews with several banks.
Just as important for a regulated firm is the audit trail. Every message, document, check and the assistant's reasoning sits on the case in the lender's own console, in an immutable record, and the underwriter sees every source behind every finding. The assistant flags rather than decides. When the FCA talks about a firm retaining full responsibility for outsourced activity, that is what it looks like in practice.
How does Lender Collect fit with a lender's existing systems and third party checks?
Two ways. First, the service includes an API gateway. A lender adds their system's API details on a secure screen and the assistant pushes case data and documents into their loan management system or CRM at the handover point. If they would rather connect from their side, we give them our API and webhook documentation. Either way, nothing is retyped.
Second, it comes with a full set of connections built in: biometric identity checks, open banking, e-signatures, and options for credit searches. There are no separate charges for most of these, nor agent licences, the only thing a lender pays for is the AI processing the assistant does.
And finally, what would you say to a lender who is interested but cautious?
Look at the five stages and ask whether they would be valuable to your business. If the answer is yes, run three cases. It costs nothing, it takes two weeks, and you will learn more from them than from six months of planning.
Send us your products and criteria, the evidence each case type needs, and who is involved in the process. A list in an email is enough. A named onboarding lead has you live within five working days, runs the first case with you, and tunes the service on your feedback through your knowledge base and rules at no extra cost. Agree your measures before the first case, so the comparison with today is fair.
Then decide with evidence. If it’s working, it is £1,000 a month and you cancel on 30 days' notice. If it’s not, you have lost nothing but a fortnight, and you will know far more about what AI can do in your business than you did before. The technology is at a point where the learning comes from trying it, and we have built the service around that.
