Eric Pozolo, CFP® | Sep 02, 2026

Nearly everyone has an opinion on artificial intelligence these days, ranging from AI as a transformational innovation improving lives around the globe to AI being the end of mankind. As it stands now, we observe it presents a great opportunity when implemented thoughtfully but significant risk if used irresponsibly. We want to share how Provident uses artificial intelligence to benefit you without putting your data at risk.

We approach AI with both skepticism and curiosity. We respect the power of the technology and strive to understand completely what our tools can and cannot do, rather than taking the approach of "move fast and break things." Many of us use AI personally and pursue creative projects outside the office. This generates ideas for use cases and keeps us up to date with the latest developments. But in the office, safety is the primary benchmark. This is defined as zero data exposure and internal execution of AI-assisted workflows.

Artificial intelligence tools can help us do more. We use AI to automate tasks that do not require expert analysis. This includes aggregating data, summarizing facts, and manipulating spreadsheets. There is a human in the loop throughout the workflow, and we architect our systems with an eye toward scalability and integration with future technology.

Our AI strategy has evolved over the past year. First, we experimented personally with our own day-to-day research workflows. Then we explored various model providers as they improved to see which best met our needs. We now subscribe to a single firm-wide solution for our AI projects, administered centrally with appropriate security controls, governed by an enterprise data protection policy. We instruct our model that we are an SEC-regulated firm in possession of highly sensitive information, and no interactions should accept or process client-identifying information. The tools are ordered to ignore all instructions contrary to this. We have a remediation plan if client data is inadvertently exposed, but we have not encountered this, and we don't expect to. We do not give AI agents open access to our file systems or our network.

No sensitive information goes into any AI tool outside the systems that already hold your records. Your records live in our custodial platform (Schwab), portfolio reporting software (Tamarac), and in our enterprise Microsoft 365 systems, under strict confidentiality and data security agreements. Where these platforms include AI features, our interactions are protected under the same security measures as the underlying data.

The AI tool we use outside of that environment is treated differently. This is used for research, writing code, and for general problem solving. No client-identifying information is entered into this system, and no documents or emails containing such information are uploaded to it. Every new use of artificial intelligence here at Provident begins with this restriction and it is non-negotiable.

Here are a few ways we're leveraging this technology.

Research

Research is the bedrock of our work here. We employ tools to strengthen this process. Our investment committee continues to perform all fundamental analysis and make all investment decisions for our growth, dividend, and fixed income portfolios.

We may use AI tools for governance analysis and 10-K comparisons for public companies. All inputs are public SEC filings and licensed third-party data. We also build forward-looking models in Excel using our licensed data inputs. We integrate a subscription service that pulls together obscure information about companies, verifiable through hyperlinked sources. This supplements our broader data services and provides more granularity, saving time and giving us more robust information for our investment decisions.

We've built a tool to generate an overview of companies that we are starting to learn about. The summary will include narrative, descriptive, and numeric data from trusted sources. If we find the prospects appealing, we will do further research and might start following a new company. This improves upon the process of retrieving data manually from disparate sources, allowing us to review a larger pool of investment candidates. On a broader scale, we design and generate real-time reports on the market, the economy, and our portfolio. Taken together, this helps us be better investment managers.

Financial Planning

AI is proving a useful supplement in answering a range of client questions. We use it to extract key information from lengthy documents such as insurance policies. We also build federal and state tax estimates for scenarios such as Roth conversions, capital gains timing, and charitable giving. The inputs for these tools are income figures and tax assumptions only. Output is used as a guide to begin planning conversations. We continue to use licensed software for in-depth planning, and always recommend clients consult their accountant or estate attorney before acting.

Business Operations

We've streamlined our operations and improved consistency. We've used automation to improve our cash management process and to be more nimble assessing bond inventories from dealers. We've also refined our monthly client distributions and fee processing, resulting in more time studying the market for investments.

One of the most productive changes has been building Excel-based automation, including VBA scripts and Power Query workflows, that run within our secure environment. Tasks that require data manipulation and can take hours manually have been simplified to generate the source data and launch an automation that completes in minutes. We invest significant time creating and testing automations over multiple periods, running parallel with the manual process. Despite the large up-front investment, we will earn back that time within a matter of months. We're also reducing human error that can be introduced by working with the data manually.

The practical advantage is that the processing occurs inside Excel, the same place your data was being processed by hand previously. Models have advanced to the point that when designing the automation, we only need to provide descriptions of source data along with the format of account numbers and other naming conventions. If the project requires specific client data to be isolated or retrieved, the model will insert a placeholder in the code along with instructions for finding and replacing it by hand. We then make the edits ourselves. As we gain experience and as models continue to improve, we will be able to do even more while honoring our safety-first approach.

As artificial intelligence progresses, we will continue to harness it for your benefit. We are committed to finding safe ways to enhance our capabilities as a firm. Our focus remains providing high-quality investment analysis and planning for our clients, underpinned by our knowledge and experience. Artificial intelligence is a nice tool, but we're quite wary of outsourcing our thinking.