management consulting

Who this applies to: Read this if you are a state or local government finance professional and want to use generative AI more effectively while maintaining appropriate review, documentation, and data safeguards.

Government finance professionals are increasingly exploring generative AI to save time, organize information, and improve efficiency. However, the usefulness of an AI-generated response depends largely on the instructions provided. A vague prompt can produce incomplete, irrelevant, or misleading results, while a well-structured prompt can help organize complex accounting information, surface important questions, and identify areas requiring further analysis and professional judgement. To help government accountants get better AI results, this article outlines five elements for effective accounting prompts and shares practical guidance for using AI responsibly.

Important considerations before using AI

Organizations should establish an approved AI-use policy before employees use AI for government finance work. Before entering organizational, confidential, personal, or third-party information into an AI tool, coordinate with IT, information security, legal counsel, and other appropriate stakeholders. It’s important to remember that AI-generated output may be incomplete, inaccurate, or unsupported, and should supplement rather than replace professional judgement, independent verification, established internal controls, or required approvals.

Recognizing when to use AI and what questions to ask are two important considerations. Imagine it is the middle of the year-end close process. You have a bank reconciliation with several old reconciling items, a list of prior-year audit adjustments, and a meeting with the finance director later that afternoon. An organization-approved AI tool may help you organize the issues and prepare questions, but only if it receives useful instructions. “Review this reconciliation” leaves too much undefined. Ask yourself key questions to help determine what is needed to create an effective prompt:

• Which balances should agree?
• Which reporting period is involved?
• How old must an outstanding item be before it receives attention?
• Does the user want possible errors, follow-up questions, or proposed entries?

Writing a useful prompt is like assigning work to someone who is new to your finance department. You need to:

• Explain the background
• Define the task
• Identify the available evidence
• Describe the expected deliverable
• Set boundaries around the analysis

Five elements of an effective accounting prompt

A practical prompt generally includes:

• Context: The government, fund, reporting period, accounting basis, and purpose of the work
• Task: The specific review, comparison, summary, or analysis requested
• Evidence: The records and authoritative sources the AI should use
• Output: The format and level of detail expected
• Guardrails: The assumptions the AI should avoid and the matters that require professional review

The amount of direction should reflect the complexity and risk of the assignment. Rewriting a routine email requires less context than reviewing a year-end reconciliation or federal grant expenditure.

Context: Explain the governmental accounting environment

Context is particularly important in governmental accounting because the same transaction may be treated differently depending on the financial statement being prepared. For example, governmental fund statements are prepared using the current financial resources measurement focus and modified accrual basis of accounting. Government-wide statements use the economic resources measurement focus and accrual basis. A capital purchase recorded as an expenditure in a governmental fund may need to be reported as a capital asset addition at the government-wide level.

A prompt asking AI to review a government-wide conversion entry should therefore identify:

• The applicable governmental fund or activity
• Whether the analysis relates to fund-level or government-wide reporting
• The accounting basis and measurement focus
• The reporting period
• The purpose of the entry
• The supporting schedules provided

Without this context, AI may apply a generally correct accounting concept to the wrong reporting level.

Task: Be specific about the work requested

Broad instructions such as “review,” “analyze,” or “summarize” do not explain what a successful result should contain.

Weak prompt: Review this bank reconciliation and tell me whether anything is wrong.

Stronger prompt: Review the attached June 30 bank reconciliation for the pooled cash operating account. Compare the bank balance, general ledger balance, and listed reconciling items. Identify mathematical inconsistencies, unexplained differences, duplicate items, deposits or checks outstanding for more than 90 days, and items without adequate descriptions. Present the results in a table showing the issue, amount, source reference, and recommended follow-up question. Do not propose a journal entry or conclude that an item is an error without sufficient support.

The stronger prompt defines the scope and focuses the review on specific risk areas. The 90-day threshold is an example for the exercise, not a recommended standard. Each government should use criteria appropriate for its policies and circumstances.

Even with well-crafted prompts, AI tools can still make calculation errors, overlook spreadsheet formulas, and produce inaccurate page, tab, or cell references. Always recalculate totals independently and trace every significant observation back to the original record.

Capabilities also vary by tool and configuration: some tools cannot reliably read attachments, complex spreadsheets, or scanned PDFs, and the same prompt may produce different responses across multiple runs. Long-outstanding checks may also raise unclaimed property or escheatment considerations, which should be evaluated under applicable policies.

Evidence: Identify the records and authorities to use

Government finance conclusions should be traceable to reliable evidence. Tell the AI which files it may use and request a reference for each significant observation. This can be especially useful when developing an inventory of agreements that may require additional accounting analysis.

Agreement-review prompt: Review the attached agreement and extract the commencement date, end date, renewal and termination provisions, payment terms, underlying assets, rights to use or operate an asset, software or hosting components, implementation services, and other provisions that may be relevant to an accountant’s evaluation under GASB Statements 87 or 96. Cite the page and section supporting each item. Identify missing or unclear information as a question. Do not determine whether the agreement meets the definition of a lease or SBITA; the government’s accounting team will make that determination after reviewing the agreement and applicable authoritative guidance.

This separates document review from the accounting conclusion. AI can help locate and organize relevant terms, but the accountant remains responsible for determining the applicable standard and treatment. The same approach applies to technical research. Instruct the AI to use specified authoritative sources, such as GASB pronouncements or federal award documents, and then verify quotations, citations, and paragraph references directly.

Output: Describe the deliverable

A useful analysis can become difficult to use if it arrives in the wrong format. Tell the AI whether you need a table, checklist, memorandum, list of questions, or plain-language summary.

For audit preparation, a government might use an organization-approved tool, consistent with any confidentiality terms in the audit engagement, to review prior-year adjustments and prepared-by-client (PBC) lists:

Prompt: Using the attached prior-year audit adjustment summary and PBC list, prepare a current-year planning table with the following columns: financial statement area, prior-year issue, support previously requested, current-year owner, current-year support needed, due date, and follow-up question. Consolidate duplicate requests. Do not assume that a prior-year adjustment is required again this year.

In my own work, I have found that asking for an approved AI tool to organize prior-year adjustments and open requests into a structured current-year tracker produces a far more useful starting point than asking for a general summary. The structure makes recurring issues easier to spot, reduces duplicate requests, and helps focus follow-up where it matters most.

Other helpful output instructions include:

• Separate confirmed observations from open questions.
• Rank items by urgency or potential financial reporting impact.
• Limit the executive summary to five bullets.
• Identify the source supporting each observation.
• Place unresolved matters in a separate follow-up section.

Guardrails: Explain how uncertainty should be handled

Generative AI may respond confidently even when information is incomplete. A prompt should address uncertainty directly.

Consider including instructions such as:

• Use only the records and authoritative sources provided.
• Do not invent missing balances, transaction details, policies, or citations.
• Separate facts shown in the records from interpretations.
• Label unsupported items “Requires additional support.”
• Draft a question when the cause of a difference cannot be determined.
• Do not make final decisions about standard applicability, allowability, materiality, accounting treatment, or journal-entry approval.
• Identify conclusions that require higher-level review.

These guardrails improve the reviewability of the response, but they do not guarantee accuracy. A finance professional should validate the source references, calculations, and conclusions before using the work.

Break complex assignments into stages

Large assignments are often more manageable when divided into sequential steps. Consider the review of expenditures charged to a federal award:

• Ask AI to summarize the award’s purpose, period of performance, allowable-cost provisions, and any specific or special conditions.
• Review and correct that summary.
• Ask AI to compare selected expenditure descriptions with the summarized criteria.
• Have it identify items requiring invoices, payroll records, procurement support, or program clarification.
• Evaluate the flagged items against the award terms and the Uniform Guidance.

AI can help organize the review, but it should not make the final allowability determination. That decision depends on the complete transaction record, award terms, applicable federal requirements, government policies, and professional judgment. A staged approach is also helpful when reconciling governmental fund statements to government-wide statements. Begin with identifying conversion categories, then review the supporting schedules, and finally develop questions about unresolved differences. Do not ask AI to solve the entire reporting conversion in a single step.

Protect sensitive government information

Use only organization-approved AI tools and follow applicable policies governing confidential information, acceptable use, and records retention.

Before entering information, consider whether it contains:

• Employee or payroll data
• Banking or payment information
• Taxpayer or utility-customer information
• Protected health or personally identifiable information
• Confidential legal matters
• Procurement-sensitive information
• System configurations or cybersecurity information
• Information belonging to auditors, consultants, or other third parties

Remove, mask, aggregate, or replace sensitive information with representative data when appropriate. Approval to access information does not automatically mean approval to enter it into an AI system. Governments should also consider their public records and retention obligations. Depending on applicable state law and how the material is created, used, or retained, prompts and AI-generated outputs related to public business may be subject to disclosure and retention requirements.

Examples of AI prompts

Examples of strong vs. weak prompts

Keep professional responsibility with the finance team

AI can identify an unusual balance, extract terms from an agreement, or organize questions from prior-year audit files. It cannot accept responsibility for an accounting conclusion. Management remains responsible for the government’s records, internal controls, financial statements, and compliance decisions. An AI-generated observation should therefore be treated as a starting point for review, not by itself as support for an accounting entry, authorization, or approval. A strong prompt will not turn AI into an accountant. It can produce a more focused first review that helps the finance team spend less time organizing information and more time investigating issues and applying judgment.

Optimizing AI as a support tool

Effective prompting can help government finance professionals use AI as a practical support tool while keeping judgment and accountability with people. By providing clear context, reliable source material, specific instructions, and appropriate guardrails, finance teams can obtain more useful results without compromising verification, security, or internal controls. AI may accelerate the work, but the accountant remains responsible for the conclusion.

Key takeaways

• Structure prompts with clear context, a specific task, supporting evidence, the desired output, and appropriate guardrails.
• Define the government, fund, reporting period, accounting basis, and purpose to help AI apply accounting concepts correctly.
• Verify AI-generated calculations, citations, source references, and conclusions against original records and authoritative guidance.
• Protect sensitive information by using approved AI tools and following organizational policies for confidentiality, security, and records retention.
• Keep accounting decisions, approvals, and professional judgment with the finance team rather than relying on AI-generated conclusions.

About BerryDunn

BerryDunn’s governmental accounting team approaches each new engagement with an open mind, assessing client needs objectively to inform clear, powerful, and actionable recommendations. We partner with clients to establish compliance practices, optimize processes, and enhance the overall financial reporting process. Learn more about our team and services.

Who this applies to: Healthcare leaders across the industry, including C-suite executives, executive directors, owners, finance directors, compliance, revenue cycle, and administrators.

As BerryDunn’s Healthcare Practice Group lead, Lisa Trundy-Whitten is closely attuned to the healthcare industry. From challenges faced by healthcare organizations to the solutions BerryDunn’s experts can provide, Lisa shares thoughtful insights for healthcare leaders. 

Healthcare is moving from a volume-and-growth mindset toward one centered on value, productivity, and sustainability. As your organization prepares for 2027, this means focusing on financial discipline, operational redesign, and proving value. Rather than considering if you can expand a service, it’s time to ask if you can financially sustain delivery of the service. It’s time for organizations to scrutinize service-line profitability, labor productivity, payer contracts, and capital investments more closely.

Strengthen financial performance and 2027 readiness

As your leadership team plans for 2027, pause to assess where you are today. Before year-end, leaders should:

• Identify the factors driving margins and areas where performance is deteriorating. 
• Review trends in labor, supply, pharmacy, and purchased-service costs. 
• Evaluate cash flow, liquidity, and revenue cycle performance.
• Test whether 2027 budget assumptions are realistic. 
• Develop best case, expected case, and downside scenarios. 

Financial improvement should not occur in isolation. The shift toward accountable and value-based care makes the connection between quality, utilization, and financial performance increasingly important.

Address revenue cycle leakage

At year-end, focus on the revenue that your organization has earned but isn't collecting efficiently. That means taking a close look at:

• Denials and their root cause
• Authorization failures
• Coding and documentation issues
• Underpayments
• Accounts Receivable aging
• Clean-claim rates
• Charge capture
• Contract reimbursement compared with actual reimbursement 

Rather than simply asking whether revenue increased, ask whether you’re collecting all the revenue you’ve earned and where in the cycle you are losing it.

Prepare for reimbursement and regulatory changes 

The last quarter is an important time to determine how 2027 reimbursement changes could affect financial and operational planning. Review:

• Medicare and Medicaid payment changes
• Medicare Advantage changes
• Commercial payer contracts and upcoming renewals
• Value-based payment arrangements
• Prior authorization and denial trends
• Physician reimbursement changes
• Programs or legislative provisions scheduled to expire at year-end

This is especially important now because several healthcare programs and payment provisions are scheduled for year-end action, while CMS has also proposed significant changes to Medicare physician payment and accountable care. For organizations with significant Medicare Advantage exposure, 2027 deserves particular scrutiny. Major insurers have announced narrower networks and market changes in response to utilization, cost, and reimbursement pressures. These developments underscore the strategic importance of payer mix, as shifts in payer participation or enrollment can have a significant impact on revenue and long-term financial sustainability. The payer mix is OKbecoming a strategic issue, not just a finance department metric.

Look at workforce and productivity

Workforce remains one of the major strategic priorities identified by healthcare organizations. It’s time to move beyond concerns of whether you are adequately staffed and instead take an in-depth look at how effectively you are leveraging the staff that you have. Examine:

• Vacancy and turnover rates
• Overtime and agency utilization
• Compensation competitiveness
• Productivity by department and service line
• Leadership succession
• Recruitment pipelines
• Retention of critical clinical positions

Consider how jobs and workflows might be redesigned to reduce administrative burden. This might create a shift to more team-based care, different skill mixes, expanded use of technology, centralized administrative functions, and revising clinical workflows to better achieve operational goals.

Evaluate service-line and program profitability

Year-end is an excellent time to challenge the assumption that every existing service should continue operating exactly as it does today. This can lead to three different decisions: invest, redesign, or reconsider.

For each major service line, examine:

• Volume and contribution margin
• Staffing requirements and capital requirements
• Payer mix
• Referral patterns
• Market demand
• Strategic importance
• Capacity constraints

As healthcare organizations face continued pressure to make sharper choices about where to allocate scarce capital and management attention, taking a hard look at these areas is especially relevant.

Shift from AI experimentation to measurable value

Healthcare organizations have spent the last couple of years experimenting with generative AI. Now it’s progressing to implementation of AI, and increasingly to accountability. By year-end, you’ll want a clear answer to the question: What have our technology and AI investments accomplished?

For every significant AI initiative, measure:

• Hours saved
• Labor costs avoided
• Revenue generated
• Denials reduced
• Documentation improved
• Patient access improved
• Employee experience
• Clinical outcomes
• Implementation and maintenance costs

It’s time to evaluate your use of AI to determine whether you’re using these tools appropriately and generating a measurable return on the investment.

Adapt as care moves outside traditional settings

There is a broader shift toward more convenient, lower-cost, and digitally enabled care. While patients are embracing virtual health visits, organizations are continuing to invest in digital engagement and monitoring. The strategic question becomes: Which care truly needs to occur on site?

This has implications for:

• Outpatient services  
• Urgent care  
• Home-based care  
• Telehealth  
• Remote monitoring  
• Ambulatory surgery  
• Post-acute care  
• Hospital-at-home models 

Ask the right questions as you prepare for 2027

As your leadership team reviews capital allocation and strategic priorities for 2027, consider answering a difficult question: What are we going to stop doing in 2027?

Organizations often create long lists of strategic initiatives without eliminating lower-value work. Spend time evaluating strategic initiatives, stop work that no longer adds value, protect essential programs, invest in high-impact opportunities, and fix what isn't performing as expected.

I encourage you to replace your broad planning discussions about strategic initiatives for 2027 with more targeted questions that get to the heart of financial and operational sustainability:

1. Where are we generating the strongest financial returns?
2. Where are we experiencing financial losses?
3. Which payer relationships warrant scrutiny?
4. Where can technology genuinely reduce costs or improve capacity?
5. What work should we redesign rather than continue to staff traditionally?
6. Which services should we invest in, redesign, or reconsider?
7. What assumptions in our 2027 budget are most vulnerable?

Your answers should serve as a strategic guide to 2027 for your organization.

BerryDunn can help 

If you need support, reach out to us to discuss ways we can guide you and help you improve outcomes. I encourage you to explore our comprehensive breadth of services and learn about our team of experts across healthcare practices.  

Best, 

Lisa Trundy-Whitten

For a more in-depth discussion of these topics, download the comprehensive guide at the bottom of the article. 

Who this applies to: Professionals responsible for credit/risk, accounting, loan operations, and the C-suite at financial institutions.

Financial institutions have devoted significant time and resources to implementing the Current Expected Credit Losses (CECL) methodology. As CECL becomes part of the recurring financial reporting process, attention has shifted from implementation to ongoing governance, monitoring, and independent review. This article briefly summarizes our full guide on the differences between CECL model validation and CECL internal audit.  

What is a CECL model validation?

A CECL model validation is an independent assessment of the methodology and model used to estimate expected credit losses. Its principal objective is to determine whether the model is appropriate for its intended use, supported by sound reasoning, implemented correctly, and producing reasonable and reliable results. For a CECL estimate, validation may address conceptual soundness, model implementation and technical accuracy, and ongoing model performance.

What is a CECL internal audit?

A CECL internal audit evaluates the broader governance, risk management, compliance, and control framework surrounding the allowance estimate. While model validation asks whether the model is sound, internal audit asks whether the institution has an effective process for operating, reviewing, approving, and reporting the estimate. The review may include governance, data and reconciliation controls, access rights, management review controls, reporting, documentation, and remediation.

The most important difference: The questions being answered

The clearest way to distinguish between the two reviews is this: The model validator evaluates the substance of the estimate, while the internal auditor evaluates the process and controls by which the estimate is developed and reported.

Areas of overlap

Model validation and internal audit are distinct, but they are not completely separate. Both may review areas such as data quality and reconciliation, documentation, oversight, model changes, third-party service providers, model monitoring, findings and corrective actions, and policies and procedures. However, the two functions examine these areas through different lenses.

Why one review does not automatically replace the other

Relying only on model validation may leave governance and control risks unaddressed, while relying only on internal audit may miss flaws in the methodology, assumptions, calculations, or model performance. A sound oversight framework considers both model risk and control risk.

A practical illustration

Consider a community bank that applies a qualitative adjustment for worsening economic conditions. A model validation may identify that the adjustment is based on an economic indicator with little demonstrated relationship to the bank’s historical credit losses or creates a risk of double-counting. An internal audit may find that the allowance committee did not review and approve the adjustment as required. The example illustrates how a technically sound model can still have a deficient process, or a well-controlled process can still rely on an unsound model.

Independence and qualifications matter

Both CECL model validation and CECL internal audit should provide objective challenge to the CECL process. Reviewers should be independent from developing, operating, and approving the model, and they should have the appropriate expertise for the review. Institutions may use outside specialists when the necessary expertise or independence is unavailable internally, but management remains responsible for understanding the scope, reviewing the results, and addressing identified issues.

How frequently should each review occur?

Review frequency should be risk-based and reflect the institution’s size, complexity, portfolio composition, methodology, model risk, and prior findings. More frequent or extensive validation may be needed after significant model changes, data-quality issues, recurring findings, portfolio changes, or increased use of management judgment. Internal audit frequency should also reflect risk, and the validation cycle and audit plan should be coordinated to avoid duplication while addressing both model risk and control risk.

Questions management and directors should ask

Management, directors, and audit committee members should evaluate the institution’s CECL oversight framework by asking the right questions across three main areas: model validation, internal audit, and coordination between the two functions. Together, these categories help assess whether model risk and control risk are being addressed without unnecessary duplication.

Download the full guide to CECL model validation and internal audit: Understanding key differences

About the team

For institutions looking for the best value, with the right understanding of scope and combination of skills and expertise, BerryDunn’s dedicated financial services professionals can customize an approach to cover both. Learn more about our services or reach out to us through our Ask the Advisor form. For more insights and expertise, follow and connect with Susan Weber and David Stone on LinkedIn.

Change is inevitable for local governments. Leadership transitions, rapid technological advances, workforce challenges, economic pressures, and shifting community needs can all affect how a locality operates. Yet a strong strategic plan should be able to withstand those changes.

Most local government strategic plans are designed to guide decision-making for three to five years. That time frame provides enough stability to establish meaningful direction while allowing flexibility as conditions change.

The challenge is knowing when change requires an adjustment to the plan, when it calls for a thoughtful refresh, and when it signals the need for an entirely new planning process. The answer depends less on any single event and more on whether the plan continues to reflect the community's priorities and provides a realistic framework for achieving them. Most organizations don't need to choose between rigidly following an outdated plan and starting from scratch. In many cases, the right path lies somewhere in between. To determine whether an organization will refresh an existing plan or create a new one, ongoing evaluation is critical.

Effective strategic planning is an ongoing, multiyear process supported by performance measurement and reporting. The most successful organizations regularly review progress, communicate results, and assess whether their objectives continue to reflect community needs. Effective metrics help leaders understand where progress is being made, where challenges remain, and whether actions are producing meaningful outcomes.

Organizations that regularly discuss progress, celebrate successes, and connect daily work to strategic priorities are more likely to keep their plans relevant and actionable over time. Many strategic plans fail not because they were poorly designed, but because they gradually disappear from regular conversations. With a strategic plan in place, implementation, and regular progress reporting, the question becomes how an organization should respond as circumstances evolve: stay the course, refresh the existing plan, or begin a new planning process.

When to stay the course

Not every change requires a strategic planning reset.

A well-developed strategic plan should be resilient enough to weather organizational and leadership changes. When the vision remains relevant and the community's priorities remain aligned with the plan, staying the course may be the best decision.

Leadership transitions are often among the first events that prompt questions about strategic direction as they may bring fresh perspectives. However, leadership turnover alone does not necessarily mean the community's vision has changed. In many cases, incoming leaders build on foundation of priorities that were developed through strong community and organizational engagement.

The same principle applies to election cycles. Changes within a governing body may introduce new ideas and areas of emphasis, but strategic plans should remain grounded in organization and community priorities. When the long-term aspirations that are actionable, remain unchanged, the existing plan continues to provide valuable guidance.

When an existing plan needs a refresh

While the vision may stay the same, the path to achieving it sometimes needs to change.

Communities experiencing changes in economic conditions, demographic shifts, development patterns, or changing resident expectations often find themselves facing challenges that were not fully anticipated when their plan was created. Leaders should evaluate the community's most pressing needs and determine what remains realistic and achievable. In many of these situations, a refresh of the plan is enough. By updating strategic priorities, performance measures, implementation strategies, and timelines, organizations can help ensure their plans remain relevant while preserving the broader vision that continues to resonate with the community.

When to start over

When leaders find themselves questioning not just how to achieve the vision but whether the vision itself is still relevant, it may be time to start fresh.

A new planning process may be appropriate when community priorities have fundamentally shifted, when major external changes have altered the community's future direction, or when the original vision no longer reflects the realities facing the organization. When the public, leadership, or staff expectations have created an entirely different environment from the one that existed when the plan was first developed, then it can affect an organization's ability to execute its strategy.

Rapid growth, major development, workforce shortages, retention issues, and cultural challenges can limit an organization's ability to carry out its strategy. Increasingly, local governments recognize that organizational health plays a direct role in community outcomes.

External factors can create similar challenges. State mandates, funding uncertainty, economic fluctuations, and policy changes can significantly affect available resources and organizational priorities. These pressures may not warrant a complete restart, but they often require leaders to revisit assumptions and identify new objectives for achieving established goals.

A new strategic planning process allows organizations to revisit foundational questions, gather fresh perspectives from stakeholders, and establish a new roadmap for the future. As the destination changes, revising the map and route may be necessary.

Key takeaways

• A strong strategic plan should provide continuity and direction even when circumstances change.
• Leadership transitions and election cycles do not automatically require a new strategic plan.
• Refresh the plan when the vision remains relevant, but the approach needs to change.
• Consider a new planning process when community priorities or the long-term vision has profoundly changed.
• Regular evaluation, communication, and stakeholder engagement help keep plans relevant and actionable.
• Strategic planning works best when it is treated as ongoing practice rather than a one-time event.

BerryDunn can help

Ultimately, strategic planning is not about producing a document. It is about creating a framework that helps organizations make informed decisions, adapt to change, and move consistently toward a shared vision. The strongest plans provide enough structure to guide decision-making while remaining flexible enough to evolve alongside the communities they serve.

A neutral facilitator can help organizations navigate these discussions. By gathering input, guiding conversations, and building consensus without advancing a particular agenda, a facilitator can help communities create strategic plans that are actionable, reflective of shared priorities, and achievable.

BerryDunn's local government consulting team helps communities develop strategic plans that align stakeholder priorities, create organizational focus, and support long-term success. Whether your organization needs leadership and staff development guidance, a wide-ranging planning process, a strategic plan refresh, organizational development support, or change management assistance, our experienced consultants can help you build a roadmap for the future and turn vision into action. Learn more about our services and team.

Who this applies to: Leaders responsible for finance/accounting, risk management, compliance, operations, IT, and the C-suite at financial institutions.

The FMS East Coast Regional Conference recently brought financial professionals together in Annapolis, Maryland, for three days of learning, discussion, and networking. During the conference, we presented “The Future of Finance: How AI and Tech Can Transform,” a session focused on how artificial intelligence and automation are changing finance, accounting, risk, and operations within financial institutions.  

Our central message was straightforward: Financial institutions will not gain a lasting advantage simply by using AI. As these tools become more widely adopted, differentiation will come from how effectively leaders redesign processes, develop their people, strengthen governance, and improve decision-making.

Finance is moving from reporting to anticipating

Historically, many finance functions have concentrated on periodic reporting, manual reconciliations, historical analysis, and responding to issues after they occur. The emerging model is more continuous and forward-looking, with real-time visibility, automated monitoring, predictive insight, and proactive decision support.  

AI may help financial institutions move routine processing into the background while allowing finance professionals to spend more time interpreting information, advising leadership, and applying judgment. Potential applications include:

• Reconciliations
• Regulatory change monitoring
• Board-package preparation
• Vendor due diligence
• Internal audit planning
• Fraud detection
• Policy summarization
• Credit and liquidity analytics

Importantly, AI is not the same as traditional automation, which generally follows predetermined rules and structured inputs. AI-enabled solutions can work with structured and unstructured information and may generate, predict, summarize, or recommend. Those capabilities create meaningful opportunities, but they also introduce new risks.

Innovation must be matched with governance

Institutions considering an AI use case should ask several questions at the outset:

• Is the use case material?
• Who owns it?
• What data will the tool use?
• Could the output affect customers or members?
• What controls and monitoring will be necessary?
• What vendor risks are involved?

AI may introduce risks involving data quality, inaccurate or fabricated outputs, bias, cybersecurity, privacy, model drift, third-party vendors, overreliance, and governance. Addressing these risks begins with clearly defined ownership, appropriate board visibility, acceptable-use policies, approved tools, data restrictions, training expectations, and a documented risk assessment for each use case.

One of the most important principles from the session was to treat AI-generated outputs as assertions requiring evidence, not as facts. Depending on the use case, controls may include human review, performance testing, segregation of duties, source citations, third-party validation, exception management, and evidence retention.

What the case studies revealed

During the session, attendees worked through scenarios involving AI-assisted reconciliations, AI-generated board reporting, and regulatory change monitoring.

In the reconciliation scenario, an overly broad matching threshold caused the system to clear items incorrectly. In the board-reporting scenario, AI-generated commentary relied on an outdated ALCO assumption. In the regulatory-monitoring scenario, an incomplete source library caused the institution to miss an important update.

Although the scenarios involved different processes, the lessons about what matters were consistent, including:

• Data sources
• Human review
• Evidence
• Ongoing monitoring

The objective is not to slow innovation. It is to make innovation reliable.

Start with the process, not the technology

Institutions do not need to transform everything at once. A practical approach is to:

• Establish governance
• Inventory current use cases
• Assess risks
• Design controls
• Implement and communicate expectations
• Monitor and adapt

Before selecting a tool, leaders should map the existing workflow, identify bottlenecks and manual workarounds, locate sources of error or delay, and clarify ownership. Opportunities can then be prioritized based on business impact, frequency, risk reduction, feasibility, and data availability. A controlled pilot allows the institution to measure results, refine controls, and scale intentionally. 

For many institutions, practical starting points may include reconciliations and exception management, board-package summarization, internal audit risk assessment, regulatory change monitoring, and vendor contract review. More complex applications, including fraud detection, credit risk analytics, CECL forecasting support, liquidity forecasting, and continuous auditing, may provide significant value but require more deliberate governance. Autonomous approvals or high-impact customer decisions without meaningful human review should be approached with particular caution.

The future finance function will not remove people from the process. Instead, technology will perform more routine tasks while people remain responsible for judgment, ownership, challenge, and decision-making. The institutions best positioned for that future will be those that start thoughtfully, build governance into their initiatives, and scale only after they understand both the benefits and the risks.

Key takeaways

• Focus on process improvement, not just technology adoption, because long-term value from AI comes from better workflows, stronger governance, and more informed decision-making.
• Shift finance teams from historical reporting toward real-time insights, predictive analysis, and proactive decision support.
• Establish governance before scaling AI initiatives, including clear ownership, risk assessments, usage policies, data controls, and monitoring.
• Verify AI-generated outputs through human review and supporting evidence, treating AI results as assertions that require validation rather than fact.

About BerryDunn

Have a question about how your financial institution can thoughtfully assess its processes, train team members, create effective internal controls, and improve decision-making to maximize the value of AI? Please don’t hesitate to reach out. Learn more about our team and services.