Broka Lender Monitor
One person was chasing hundreds of lenders by hand. Now the system does the chasing, and she just watches the calendar fill up.
- Services
- AI Systems, Workflow Automation, Internal Tools, UX/UI Design, WhatsApp Business API, Gmail Integration
- Client
- Brokerbrain Ltd T/A Broka
- Year
- 2026 - Ongoing
- Website
- broka.uk
Challenge
Brokers can only find the right loan if lender data is current. Keeping it current was one person's job, and 124 lenders were overdue.
Broka connects brokers with over 200 lenders, and every search depends on those lenders keeping their products up to date. Rates move and criteria change, and a listing nobody has touched in months quietly misleads every broker who finds it. When we looked, 541 of 976 products had gone past 90 days.
Chasing them fell to Broka's lender liaison, one lender at a time: find the contact, write the email, pick up the phone, follow up, log it, repeat. It never ended. Automating it was the obvious answer, but it had to be done without damaging relationships Broka had spent years building. Hundreds of clumsy messages to senior lenders is how a brand becomes spam.
And it was harder than it looks. Teaching an AI to write to a lender the way a good account manager would took a lot of programming and a lot of rewrites. WhatsApp was a nightmare in its own right: you can't automate it from the normal app, so everything had to run through Meta's own API, where the first message to anyone must be a pre-approved template. Getting those templates enabled meant a proper fight with Meta.
Goal
Automate the chase end to end, without it ever sounding automated or running away with itself.
We designed and built Broka a custom monitoring and outreach platform. It watches every lender's product data, works out who needs a nudge, and runs the conversation across email and WhatsApp, with a human in charge of anything that goes out.
Stale products, spotted automatically.
Every product is checked against a 90-day threshold, and lenders are ranked by their most overdue listing so the worst offenders are contacted first.
Messages written for each lender.
Each email names the actual products and how long they've sat untouched, then offers a call through Broka's booking link. No mail merge, no Dear Sir or Madam.
WhatsApp, done the official way.
Built on Meta's WhatsApp Business API with an approved template and a Book button. Once a lender replies, ChatGPT carries the conversation in Broka's voice.
Email from a real inbox.
Everything sends through Broka's own Gmail, so replies land where the team already works and nothing reads like a marketing platform.
Learning mode before trust.
Every AI draft waited for approval at first. The team could edit, send or bin it, and lenders move to auto-send only once the tone has proved itself.
Safety rails built in.
A master live switch, a kill switch, hourly and daily send caps per channel, and watchdogs that pause everything if Meta's quality rating dips or deliveries start failing.
One inbox for every conversation.
Email and WhatsApp threads side by side, filtered by replied, booked and stalled, with a take-over switch the moment a human needs to step in.
Cleaner data as a by-product.
Duplicate lenders are caught from shared email domains and merged, and a contact health view flags anyone who has gone quiet at 60 days and six months.
Result
A job that used to fill the week now runs itself. The only thing left to do is take the calls.
Broka's lender liaison doesn't ring anyone any more. The system finds the stale products, writes to the lender, follows up, and hands over a booked call. Her day now starts with checking the calendar.
In its first weeks the platform wrote 687 personalised messages across 124 lenders and opened 841 conversations over email and WhatsApp, with every early message reviewed by a person and sent on a steady drip, so it never read like a blast.
And it doesn't stop. The moment a product ages past 90 days, the lender behind it is picked up automatically, so Broka's data stays current without anyone having to remember to check.
Receipts
From one-by-one to 0 conversations
A manual chasing job turned into a monitored, AI-run outreach system across email and WhatsApp.
0
Products monitored
0
Lenders
0
Conversations
Milestones
JUL 2026
0 products audited
Every listing checked against a 90-day threshold.
JUL 2026
0 stale products found
Spread across 124 lenders.
JUL 2026
Meta templates approved
After a proper fight with Meta's review process.
AUG 2026
Learning mode live
Every AI reply approved by a person before it went out.
AUG 2026
0 drafts written
Personalised per lender, per product, per channel.
SEP 2026
0 conversations opened
Email and WhatsApp running side by side.
TODAY
Always on
Lenders picked up automatically the day a product goes stale.
BUILT FOR BROKA · FIGURES TO SEPTEMBER 2026
