Berkadia - Elias Sipin

Context

The Landscape in 2021

Berkadia is a joint venture between Berkshire Hathaway and Jefferies Financial Group. The company originates over $30 billion in annual commercial real estate loans and operates across Mortgage Banking, Investment Sales, Servicing, and multiple technology platforms serving both internal teams and external clients.

When I arrived in 2021, the design function was an 8-person team with no operating model. Designers got assigned to projects reactively. There were no shared standards and no research practice. Nobody had visibility into the product roadmap, and each business vertical operated on its own, which meant fragmented user experiences, duplicated effort, and growing design debt across every platform.

The Chief Product Officer hired me to fix this. Not to redesign screens. To build an organization that could drive experience quality at enterprise scale while the business was growing and the product landscape was shifting underneath us.

Executive Decisions

Three Decisions That Defined the Transformation

Decision 01

Build the Organization While Shipping the Product

There was no luxury of stopping work to restructure. The business needed experience design, product design, and OKR delivery across a growing portfolio from day one. So we built the organization in parallel with the product work.

Over the first six months I restructured the team around business verticals, establishing dedicated pods for Mortgage Banking, Investment Sales, Servicing, and Innovation. I hired researchers, design ops leads, and senior designers who could own their domains. We built a shared component library and created a design review process that tied design decisions back to business outcomes. All while continuing to ship across every active product.

The tradeoff was real: we were doing two jobs at once, building the machine and running it at the same time. But it meant the organizational foundation grew alongside the work instead of arriving too late to matter.

Rejected: Pausing product delivery to restructure. The business couldn't wait and neither could we.

Decision 02

Bet on AI Before the Industry Did

In mid-2023, before most commercial real estate firms had an AI strategy, I was asked to lead the creation of BerkadiaGPT, an enterprise AI assistant for mortgage banking and investment sales workflows. The team was small: the VP of Product, a group of data scientists we had gained through an acquisition, and on the design side, myself and one of my researchers.

We didn't build a demo. We ran a 4-phase discovery program across Mortgage Banking and Investment Sales. We interviewed 13 stakeholders from VPs to analysts, synthesized findings into concrete use cases (data assistant, guideline extraction, document analysis, BOV generation), and scored each one on desirability, feasibility, and viability to figure out what to build first.

The research told us something that shaped the entire product direction: accuracy and data access were the top concerns across every business unit, not speed or novelty. So we prioritized the Data/Research Assistant use case (10.0 desirability, 9.0 viability) over flashier features like automated BOV generation.

Rejected: Building AI features based on executive assumptions instead of evidence.

Decision 03

Fix the Platform, Don't Redesign It

RedIQ was the highest-revenue product in the portfolio, Berkadia's client-facing deal management platform. Stakeholders wanted a full visual redesign. I pushed back.

The real problems were structural. The data architecture was brittle, workflows were inefficient, and missing functionality forced users into manual workarounds. A visual refresh would have consumed 6+ months and delivered cosmetic improvement without solving any of that.

We prioritized a platform rewrite focused on data integrity, customizable views, and workflow automation instead. We stabilized the Salesforce integration for ADM/DAS at the same time. The result was measurable business value delivered incrementally rather than a big-bang launch that risked regression.

Rejected: Full visual redesign of RedIQ. The problems were in the data layer, not the interface.

Scope

What I Owned

I was responsible for total experience quality across every Berkadia product, both consumer-facing and enterprise. That meant influencing roadmap priorities, driving cross-functional alignment, and making tradeoff decisions across competing business unit needs. I had no direct budget authority, so every initiative required building the case and earning buy-in.

Products Owned experience quality across the full product portfolio: RedIQ (client-facing deal management), ADM/DAS (account and deal analysis), Servicing applications (loan servicing workflows), Cash App (transaction management), Location Data (property intelligence), and BerkadiaGPT (enterprise AI).
Organization Scaled the team from 8 to 47+ people across UX design, UX research, design operations, and visual design. Structured into vertical pods aligned to business units with embedded research capability.
Strategy Owned design strategy fully and co-owned product strategy with the CPO. Drove adoption of research-informed decision making across the organization through workshops, prioritization sessions, and the company's first dedicated user research practice.
Stakeholders Partnered with C-suite leadership (CPO, CDO, business unit heads) to align design investment with revenue priorities. Presented product design strategy at the annual technology summit alongside the Chief Data Officer.
Innovation Created and led the Innovation Design Pod that researched, prototyped, and launched BerkadiaGPT. Ran discovery across 4 business units with cross-functional teams of designers, PMs, and engineers.

Products

Surfaces Under My Direction

RedIQ deal management dashboard showing pinned deals and deal list

RedIQ - Client-Facing Deal Management 4.5k leads / $1B production volume

We rebuilt the platform around data integrity, customizable views, and workflow automation. The goal was to eliminate the manual workarounds that were costing deal teams hours every week.

BerkadiaGPT conversational AI interface for commercial real estate

BerkadiaGPT First enterprise AI product at Berkadia

An AI assistant built for mortgage banking and investment sales. Research pointed us toward data access and guideline extraction as the highest-value use cases, so that's where we started.

Outcomes

What We Delivered

Here's what the work added up to over three years:

$12M+

Revenue impact across product portfolio

8 to 47+

Team scaled across 6 verticals

35%

Reduction in time-to-market

$1B

Production volume via RedIQ leads

1.5k

CAs signed through ADM/DAS (2022)

6

Business verticals supported simultaneously

Reflection

What I Learned

Three years inside a regulated financial services enterprise will teach you things you can't pick up anywhere else. Some of them were hard lessons.

What Broke

I underestimated how long it takes to shift a culture from treating design as a service desk to treating it as a strategic function. Even with executive sponsorship, some business units took over a year to fully integrate design into their planning cycles. Looking back, I should have embedded designers directly in business unit meetings from day one instead of waiting for the formal pod structure to be ready.

What I'd Do Differently

I would have started a lightweight research practice in month one, before hiring a dedicated researcher. The insights from BerkadiaGPT discovery proved how much credibility you earn with stakeholders when you show up with evidence. If I had been generating those kinds of findings earlier, the organizational transformation would have moved faster because the data would have made the case for me.

Password

Scaling Design as an Operating System for a $30B Enterprise