The Engagement

We work in your building, on your problem.

A lean team. A scoped SOW. Working software before the engagement ends. Here is exactly what happens, week by week.

Who shows up

A lean team, and nobody shows up with a template.

The subject-matter expert

Someone who has run a bank's balance sheet, sat through the exam, and knows what a two-day account-opening process actually costs. They lead the diagnosis, because the hard part isn't the technology — it's knowing which problems are real.

The forward-deployed engineer

Someone who ships. They're in the room during discovery, not briefed afterward, which is why the first working artifact lands in weeks instead of quarters.

Domain expert first, engineer second. That's the whole model.

The four phases

From scoping call to shipped software, in twelve weeks.

01

We deploy · Week 1

A lean team arrives: banking subject-matter experts and forward-deployed engineers. The same week, your team goes live on Perlucem — TBPI™ scores, peer benchmarks, and the MCP connector running inside Claude.

Your team is asking your bank questions in plain English on day five.

02

We diagnose · Weeks 2–6

The Intelligence Gap Assessment. We assess three domains — strategy, data layer, applications — and overlay your actual operations. Then we triangulate what you have against what you need, and against what you think you have.

A current-state map, a named list of gaps, and cycle-time findings with hours attached.

03

We build · Weeks 6–12

We ship the quick win first — the bottleneck you feel every week. Then the bespoke skills that sit on top of it.

Working software your people use. Not a recommendation to build it.

04

You compound · Ongoing

The intelligence layer stays. Seats for your team. Every skill built for your bank keeps working, and your peer benchmarks keep sharpening.

An institution that gets smarter every quarter without adding headcount.

The contrast

The two things a bank usually chooses between. And the third option.

Hire a consultancy

  • A team that’s never run a bank
  • 200-page deck at the end
  • Recommendations to build
  • The knowledge leaves with them
  • Priced by the hour, scoped by the month

Buy a platform

  • A login and a training webinar
  • Six-month implementation
  • Generic dashboards you build the queries for
  • You rent the intelligence indefinitely
  • Annual contract before you see value

Deploy Perlucem

  • Bankers and engineers, in your building
  • First artifact in weeks
  • Working software your people already use
  • The findings and the fix are yours to keep
  • Scoped SOW, named deliverables
What you keep

Five artifacts. Yours to keep.

1

Current-state map

Across strategy, data layer, and applications.

2

The gap analysis

What you have, what you need, and what you think you have. The third one is where the money is.

3

Cycle-time findings

Named operational bottlenecks, with the hours they cost attached.

4

Intelligence-layer recommendation

The one thing missing over all of it.

5

Prioritized roadmap

Led by a quick win you’ll feel in weeks, not a five-year plan.

What's yours stays yours. What's ours stays ours.

You own the findings, the roadmap, and everything we configure for your bank. We keep the framework and the platform underneath it — the same way your core provider keeps their software while your data stays your data. It's in the SOW in plain language, before you sign anything.

What you're agreeing to

No surprises. Ever.

We never connect to your core banking system.

No customer PII, no NPI, no account-level data ever enters Perlucem.

Your questions and your data never train our models.

Fixed scope, named deliverables, signed SOW. You know what you’re getting before we start.

You own your deliverables — the assessment, the roadmap, and the skills built for your bank. We keep our platform and frameworks. It’s written down before we start.

Platform seats start in week one. You see what we see, at the same time we see it.

If the first engagement doesn’t produce a win you can measure, there is no second one.