The Great Wealth Transfer is coming: How AI is shaping the future of legacy planning

When it comes to personal finances, estate planning is often overlooked. In fact, 2 in 3 Americans don’t even have a basic will in place. Many assume they lack sufficient assets, while others are deterred by the complexity and cost.

But a generational time bomb is ticking.

With Baby Boomers currently holding 72% of U.S. household assets and $84 trillion set to be passed down to their heirs by 2045, we are facing what is commonly known in the tax and estate field as “The Great Wealth Transfer.”

To meet this great moment, we will need great estate planning like never before.

It’s a challenge Rei Carvalho experienced first hand, not long after he sold his fraud protection platform, Emailage, to LexisNexis in 2020. Carvalho wanted to secure his family’s future with a comprehensive estate plan. What he found instead was a complex, outdated process with no tools to visualize his plan or evolve it over time.

It’s this experience that inspired him to create Wealth.com.

Carvalho’s vision was clear: a platform that simplifies estate planning while equipping financial advisors with powerful tools to visualize and manage complex plans. It began with building document creation tools, so financial advisors could offer clients a simpler way to create an estate plan without needing to consult an attorney.

But what about clients who already had estates and trusts, some of which can run to hundreds of pages and be quite complex in their structure? The Wealth.com team wanted to provide advisors with the tools to more quickly and seamlessly unpack existing estates and provide clients with new options and better planning.

And so Ester was born. This AI-powered agent is Wealth.com’s answer to modernizing an antiquated industry.

Creating the first AI-powered estate planning co-pilot

Estate planning documents — like revocable and irrevocable trusts, last will & testaments, advance health care directives and financial powers of attorney — are notoriously long and filled with legal complexities.

Just think of a 100-page trust packed with nuanced contingencies, details of assets will be distributed after death, and many more important relationships. It can take financial advisors hours, if not days, to manually review just one of these to extract key details, like beneficiaries and asset distributions.

Ester changes that, automating the most relative and low value tasks, so advisors can focus on exactly that: building better plans and providing better customer relations.

Yet building Ester itself was no easy task. The complexity of estate planning documents required a careful balance of automation and human oversight. The latter, in the form of our legal experts, lays a critical role in annotating datasets and validating outputs.

Also, a false positive in estate planning — such as misidentifying a beneficiary — can have major consequences for an advisor working with a client. To mitigate this, Ester incorporates user review by clearly showing where in the document each piece of information was found with a simple click.

Along the way, we’ve learned a lot of lessons, some big and some small. As we enter the AI revolution, these are insights any company or AI developers could follow:

• Domain experience is just as crucial as engineering excellence.
• High-quality data is the foundation of effective AI.
• AI should empower humans, not replace them.
• Adaptability keeps AI systems future proof.

How Ester AI works

Ester analyzes critical legal documents to locate the most important pieces of information, such as who the beneficiaries are, how assets are distributed upon death and any other key provisions or contingencies. It then provides a clear overview of this information, displaying it alongside the original uploaded PDF. Advisors can easily review, edit, and confirm the extracted data directly from the wealth.com platform.

For advisors that want a more concise view, the Executive Summary feature delivers a snapshot of the document’s key details, all quickly compiled by AI. This summary can be downloaded and shared with clients or colleagues, streamlining collaboration and decision-making.

This structured information from Ester also flows directly into visualizations, allowing advisors to see how the disposition of a trust works, for example, and to present it visually to clients. By making these complex documents easier to understand, Ester empowers advisors to spot areas for updates, identify strategic opportunities and have more meaningful conversations.

At its core, Ester uses a combination of advanced AI layers to transform how estate planning documents are processed, analyzed and understood:

1. Optical character recognition, or OCR, layer: Using Google Cloud’s Vision AI, Ester extracts document layout structure and text from scanned PDFs, removing noise like stamps and notary marks. This ensures that every critical detail is captured cleanly and ready for analysis.
2. Large language model, or LLM, layer: Powered by Google Gemini along with other proprietary and open-source models via Vertex AI, Ester excels at understanding key relationships, pertinent details and identifying contingencies so advisors can suggest optimizations.
3. Post-processing: Our proprietary agentic workflow refines and structures outputs into actionable insights via visualization reports.

Behind all of this, there is another critical layer: the human factor.

Our AI suite is trained on annotated datasets curated by wealth.com’s legal team, ensuring that it understands the complexities of estate planning. For example, different attorneys may use varying terms for the same concept — like calling a bypass trust a family trust or a credit shelter trust. Ester learns to identify these nuances, providing a consistent and accurate review regardless of how the documents are drafted.

Why Generative AI is the key

Traditional machine learning models simply can’t handle the scale or complexity of estate planning documents. They often falter when tasked with interpreting nuanced legal language across lengthy documents.

This is where we found LLMs, particularly Gemini, shines. Google’s industry-leading long context window has the ability to process over two million tokens, allowing Ester to achieve a level of understanding previously unattainable. In fact, Ester has identified critical information even wealth.com’s experienced legal team overlooked — an impressive example of how AI can augment human expertise.

How we’re rethinking estate planning with Ester

When building Ester, we tested leading large language models (LLMs) from across the industry. Ultimately, we decided to leverage Vertex AI, which unlocked our ability to leverage an ensemble of state-of-the-art LLMs to optimize our performance in three crucial areas: accuracy, speed, and scalability.

The capabilities of Gemini, and other AI foundation models via Vertex AI, are augmented by wealth.com’s proprietary breakthroughs, including:

• Annotation data: Our internal legal experts have meticulously curated and annotated thousands of actual estate planning documents so far, and continuously add more to our database, all of which are used towards training and evaluating Ester. This ensures Ester’s outputs are both reliable and contextually accurate.
• Tailored prompts and post-processing: Noise-reduction techniques eliminate irrelevant details, like notary stamps, and structure data so it can be fed directly into visualization reports the advisor and client can easily understand. We also work with our attorneys on prompt engineering to ensure Ester can navigate the nuances of legal texts.
• Distributed execution: Ester is engineered to parallelize document processing, reducing end-to-end analysis to under two minutes on average — faster than any other estate planning extraction AI on the market.
• Interactive citation: With a single click on the results, advisors can jump to the exact spot in the uploaded PDF that serves as the source of truth for a given generated content. .
• Advanced RAG pipeline: We incorporated a robust retrieval-augmented generation (RAG) pipeline to mimic how an estate planning attorney would review a legal document.
• Knowledge graphs and contextual analysis: Ester extracts a comprehensive understanding of entity relationships within the documents, ensuring that critical data is extracted accurately and placed into context for strategic decision-making.
• Human-like fact-checking: We developed advanced fact-checking mechanisms to cross-reference extracted information before delivering results back to the end user, providing an added layer of confidence.
• Responsible AI: Ester undergoes a thorough red team process for each major release to ensure that our AI copilot is trustworthy, fair, transparent and interpretable.

AI lessons for other industries

We’re in a golden age of AI right now. Amazing advancements are seemingly being made every day but there’s also a lot of expectations about what AI is and what it should be able to do. Throughout our own journey with Ester, these are the biggest lessons we’ve learned:

1. The value of domain expertise: Building AI for a specific industry requires more than engineering breakthroughs — it demands deep collaboration with domain experts. In our case, estate planning is a field where precision is critical. Mistakes in legal documents, like incorrect beneficiaries, can have significant consequences. That’s why Ester is trained with extensive input from our legal team, ensuring it understands the complexities of trusts and wills. Without this expertise, AI tools can misfire — much like instances where lawyers have received inaccurate, or even non-existent, information from generic AI tools.
2. Data is the foundation: Annotated datasets are the bedrock of effective AI, regardless of industry. For Ester, our legal experts meticulously curate datasets from real estate planning documents to fine-tune the model. This ensures the AI delivers outputs that are not only accurate but also contextually relevant to advisors and clients.
3. AI augments, not replaces: Ester doesn’t replace financial advisors — it empowers them. By automating time-consuming tasks like document review, AI frees advisors to focus on more strategic, impactful work. The power of AI lies not in replacing humans, but in enabling them to move faster, work smarter and spend more time building relationships.
4. Flexibility in AI development: AI is evolving rapidly, and adaptability is key to staying ahead. We designed Ester to be model-agnostic, allowing us to integrate different LLMs as technologies advance. Today, Gemini plays a significant role in Ester’s success, but we also leverage Vertex AI to ensure scalability and future adaptability. Companies that design their systems to be flexible avoid being tied down by legacy technology and can pivot seamlessly to new innovations.

What’s next for Ester AI and wealth.com

At wealth.com, we’re committed to building on our vision of integrating financial and estate planning for advisors and their clients. Ester makes complex estate plans both manageable and actionable, transforming how advisors approach their day-to-day work. While Ester excels at extracting critical information from estate documents, our focus now is on making it even more actionable.

Next up, wealth.com is committing resources to harnessing the power of agentic AI workflows. Whether that be automating the often tedious trust administration process, quickly modeling hundreds of scenarios to optimize for taxes, or simply helping users navigate to the next best action, we believe agentic AI can vastly elevate the status quo.

It’s an exciting time to be innovating in any industry. For the team at wealth.com, it’s our sincere hope that we’re not just solving inefficiencies in estate planning — we’re setting a precedent for how others can integrate AI effectively into their lives and work.