Insights

A New Look for a New Era of Transformation

We’re proud to announce the launch of our PhoenixTeam new website. PhoenixTeam has always been about more than just delivering technology solutions. We exist to drive meaningful transformation in the industries that matter most: mortgage, financial services, and federal modernization. Our new digital home reflects exactly who we are today and where we’re taking our clients tomorrow.  

Built AI-First. Built for Impact.

Technology is evolving faster than ever. We don’t just adapt, we lead. Our new site highlights how we help organizations harness the power of AI to transform operations, decision-making, and customer outcomes. From AI strategy to hands-on implementation, we’re helping clients unlock faster, smarter, and more resilient ways of working.

Expertise Where It Counts: Who We Serve

We have deep roots in both federal and commercial sectors, especially in mortgage, financial services, and complex technology product delivery. The new website makes it easy to see how our experience spans mission-critical programs and innovation-driven growth initiatives and why we’re the trusted partner when the stakes are high.

Our Expertise: Specialized Service Hubs

We organize our expertise around the critical areas where transformation is happening fastest. Through specialized hubs, we deliver practical, AI-native solutions across industries: Our hubs include GenAI Services, Technology Delivery and Transformation, Technology Product Management, Mortgage Data Services, and Federal and Commercial Go-To-Market Support each built to accelerate results where they matter most. In partnership with the Mortgage Bankers Association, we also offer hands-on AI education programs to help teams lead with confidence in an AI-driven world.

Phoenix Burst: GenAI-Native Solutions for Business Process Fulfillment

On our new site, you’ll see how Phoenix Burst is changing what’s possible for business process fulfillment. Phoenix Burst is our genAI technology product designed to fulfill business processes, not just track them. Unlike traditional workflow tools, Burst delivers real outputs for human validation, accelerating operations across compliance, software development, and compliance decision-making.

Who We Are: People-First, Mission-Driven, Future-Building

Technology is just the start. The new PhoenixTeam website isn’t just about showcasing our services, it’s about reflecting the culture that drives everything we do. We’re a team of strategists, AI technologists, and builders who move fast, lead with purpose, and put people at the center of every transformation. We invest in growth, innovation, and solving the problems that move industries forward.

The industries we serve are changing...FAST. PhoenixTeam is here to help you move faster, smarter, and with more clarity. Whether you’re navigating AI disruption, federal mandates, mortgage modernization, or simply building the next phase of your tech strategy, we’re ready.

Connect with us to start building what’s next.

‍

Phoenix Burst, the industry’s first genAI-native mortgage business process fulfillment platform, has been named a winner of the MortgagePoint Tech Excellence Award, recognizing the most innovative technology providers transforming the mortgage and real estate industries.

For decades, efforts to modernize the mortgage industry have been stalled by the heavy burden of regulatory compliance. Phoenix Burst changes that. By making compliance seamless, the platform clears the way for real innovation. What once took months of legal, compliance, and operations coordination, Phoenix Burst now delivers in hours, turning complex regulatory updates into clear, actionable requirements, user stories, and test cases with a single click.

The MortgagePoint Tech Excellence Award honors companies that are not only pushing the boundaries of what’s possible through technology, but also meaningfully improving operational efficiency and the mortgage experience. Winners are selected based on nominations from industry professionals and judged on their impact, innovation, and ability to drive lasting change.

See the full list of award recipients: https://themortgagepoint.com/2025/03/31/mortgagepoint-announces-2025-tech-excellence-award-recipients/

‍

I’ve talked before about how managing regulatory change in mortgage is kind of like trolling the internet. Mortgage compliance requirements are primarily published in formats suitable for human readers, not machines. The requirements span hundreds of entities, including federal bodies, state attorneys general, investors, federal and local housing agencies, and the government sponsored enterprises (GSEs), with each providing requirements in diverse formats through multiple, inconsistent distribution channels. According to the National Mortgage News, this has created more than 1,000,000 pages of requirements overt time. Yikes.

How Mortgage Compliance Actually Works

We envision compliance as a wheel, with change at the center and evidence of process compliance as the enveloping outer ring. The change kicks off a cascading process that can take anywhere from 45 days to 18 months, depending on the size and complexity of the change.

There is never time to test everything, and compliance and line of business operators ship around copies of an excel spreadsheet to analyze and communicate the requirements. Engineering teams create Jira and ServiceNow tickets to track the development work, and a testing team somewhere might test a subset of the change before production. Mortgage companies then rely on first, second, and third line of defense to actually provide evidence of process compliance at the enterprise and loan levels. There is an alternative to this insanity.

Definition: AI-ready data is structured, consistent, high-quality information that has been cleaned, properly formatted, and labeled to be immediately usable for training or analysis by artificial intelligence systems without requiring significant preprocessing or transformation.

The Alternative to Insanity: AI-Ready Policy Data

The emergence of generative AI (GenAI) in the mortgage industry present previously unattainable opportunities for automating compliance, risk assessments, and underwriting processes. However, AI systems require structured, standardized data to operate effectively. Current unstructured formats severely limit AI’s accuracy and usability. Our tests for Phoenix Burst have shown that structured regulatory data dramatically improves AI accuracy. We can achieve up to 79% accuracy using image-based formats contrasted with nearly 100% accuracy with structured inputs.

Benefits to Regulators and the Mortgage Industry

The people who make the requirements have both the authority and motivation to lead the shift to AI-ready data. Standardized, open data directly aligns with the missions of affordable homeownership, clarity, transparency, and effective oversight. It facilitates real-time compliance monitoring, improved market surveillance, and reduces reliance on costly third-party interpretations.

For the mortgage industry, structured data reduces costs, enhances compliance accuracy, and accelerates adoption of regulatory changes. Lenders can integrate real-time regulatory updates directly into automated systems, significantly cutting manual compliance efforts and improving overall lending efficiency.

Hosting AI-Ready Data: Addressing Jurisdictional and Privatization Challenges

It is an interesting thought experiment to consider who might host such a centralized data service. It is this author’s opinion that the information should quite obviously be democratized, so the cost of access should be negligible (the Federal Register application programming interface for example costs nothing to access). Yet there is a cost to create and manage such a service. There is a responsibility to be impartial and independent. Who should create and host this service?

1. National Archives and Records Administration (NARA): NARA is the official record keeper of the United States federal government. They are currently the custodian of the Federal Register API. I love that they already much of the plumbing, but it doesn’t make sense for states to publish here. Furthermore, if Fannie Mae and Freddie Mac are privatized, it doesn’t make sense for them either. So that probably won’t work.
2. Mortgage Industry Standards Maintenance Organization (MISMO): MISMO is a non-profit organization that develops, promotes, and maintains voluntary electronic commerce standards for the mortgage industry. This option offers the potential for industry-wide acceptance and technical expertise. MISMO is a volunteer-based organization that undertakes a lengthy, consensus-based process, which would be the default process here. Diverging opinions will make this hard to do.
3. Consortium Approach: A dedicated consortium could effectively bridge federal-state-private sector dynamics, ensuring broad participation and standardized governance. This could be an alternative in the event MISMO chooses not to participate, the problem will be one of funding. I have no doubt the industry wants this, but the reality is standing up the API and supporting has a cost. Of interest here is that the Federal Register API is well documented and open source.
4. Open-Source Community: Kind of similar to a consortium approach, but really working from the ground up instead of top down. An open-source community is a collaborative group of individuals and organizations who voluntarily come together to develop, maintain, support, or improve openly accessible software or projects. Members contribute their time, knowledge, and skills, guided by shared goals, values, and openness.

Facing Headwinds: Overcoming Industry Resistance

Existing compliance data providers might resist open data, fearing revenue loss. However, AI-ready standards open new monetization avenues, such as advanced analytics, lower cost compliance validation services, and AI-driven tools built on top of structured data. Rather than competing on raw regulatory data, these companies can compete on sophisticated value-added offerings, growing their markets while improving industry compliance. Once the data is unlocked, an entirely new set of revenue opportunities will emerge.

States traditionally maintain independent mortgage regulations, posing a challenge to standardize. However, AI-ready data promises significant cost savings, streamlined compliance oversight, and improved accuracy in state-level enforcement. Demonstrating clear economic and operational benefits through pilot projects and leveraging existing cooperative models like the Nationwide Multistate Licensing System (NMLS) can effectively encourage state participation without compromising their regulatory autonomy.

The Call to Action

If you make or use mortgage compliance requirements, we really want to here from you. Please engage and tell us what you think. Transitioning mortgage policy data to an AI-ready standard is so much more than a technical problem, it’s a strategic solution that unlocks previously infeasible efficiencies.

‍

By Kenny Akridge

If you’ve ever struggled to get a large language model (LLM) to handle a complex task—only to end up with incomplete, confusing, or just plain wrong responses—you are not alone. The problem isn’t the model though; it’s the approach. Cramming too much into a single prompt will lead to disappointing results.

The solution? Prompt chaining. Instead of overwhelming the model, break your request into a series of prompts. This allows the LLM to tackle each step with precision and leads to more accurate and reliable results.

If you’ve been disappointed with LLMs, it’s time to rethink your strategy. The key isn’t asking for less, it’s structuring your requests more effectively.

When One Big Prompt Fails

Let’s look at a simple example.

Imagine we have two versions of a document, and we want to find changes and summarize the impact of the changes. We might take the following approach:

1. Compare the original text with the new text.
2. Describe the grammatical differences.
3. Summarize the changes.
4. Assess the impact.

It’s tempting to bundle everything into a single prompt, but LLMs can lose focus, misinterpret the goal, or hit text limits. The results might be:

• Vague or incomplete responses.
• Jumbled steps (e.g., skipping the summary).
• Incorrect or inconsistent impact ratings.

When so much is happening, it’s easy for the model to overlook important details.

LLMs perform best when given clear, focused instructions. Overloading a prompt with too many tasks forces the model to juggle too much at once, increasing the risk of errors.

Prompt chaining solves this by breaking the process into logical, manageable steps.

How Prompt Chaining Works

Think of prompt chaining as an assembly line, with each step refining and building upon the prior output. Here is a simple example based on our document comparison problem:

Step 1: Compare Text Versions

Feed the model the original and new text, asking it to generate a list of differences. Keep the prompt simple:

• Here’s the original text: [text A]. Here’s the new text: [text B]. Please list the differences.

Step 2: Summarize the Differences

Take the detailed comparison from Step 1 and ask the LLM to condense it:

• Summarize these differences in one or two paragraphs.

Step 3: Assess Impact

Pass the outputs from Steps 1 and 2 into another prompt to evaluate impact:

• Based on these differences, classify the overall impact as High, Medium, or Low. Briefly explain why.

By chaining these steps, each prompt has a clear focus, ensuring accuracy and consistency throughout the process.

Benefits of Prompt Chaining

1. Improved Accuracy

Each prompt focuses on a single task, reducing errors and ensuring precise responses. Instead of overwhelming the model, we guide it through a structured workflow.

2. Less Confusion

When prompts are too complex, the LLM can misinterpret them. By breaking tasks into separate prompts, each step is clear and easier to follow.

3. Easier Debugging

If a mistake occurs, identifying the issue is simple—we only need to adjust one part of the chain rather than rewriting an entire prompt.

4. Better Scalability

Whether we’re processing a few entries or thousands, prompt chaining ensures the LLM handles each one efficiently without running into input length limits.

Practical Tips for Effective Prompt Chaining

· Keep Prompts Clear and Concise: Direct, simple instructions yield better results.

· Use Consistent Formatting: Define how you want the output structured (e.g., bullet points, tables, plain text, JSON).

· Document Your Chain: Track each step to diagnose and refine your workflow.

· Validate Each Step: Before moving to the next prompt, verify that the output makes sense and aligns with expectations.

Wrapping Up

Prompt chaining transforms how you interact with LLMs, turning scattered, inconsistent responses into well-structured, accurate outputs. Next time you're working with an LLM, resist the urge to overload a single prompt. Instead, break it down, chain it together, and watch the results improve.

Try prompt chaining in your next project and see the difference. Have insights or success stories? Share them in the comments below and let’s learn together!

‍