The Big Idea
Finance leaders are being presented with an increasingly broad range of AI opportunities:
- Analyze financial results.
- Research a transaction.
- Draft management commentary.
- Answer questions about company information.
- Review contracts and supporting documents.
- Forecast future performance.
- Communicate with shareholders and other stakeholders.
The potential value is real. Work that once took hours can sometimes be completed in minutes. Finance teams can analyze more information, respond more quickly, and spend less time assembling data manually.
But speed is only one dimension of value.
A finance process must also be accurate, controlled, explainable, and appropriate for the consequences of the decision being made. An AI-generated answer that arrives immediately is not valuable if it relies on the wrong information, exposes restricted data, overlooks a required approval, or cannot be supported during an audit.
This is why organizations should evaluate AI in finance through three connected questions:
Where does it create value?
Does it reduce manual work, improve responsiveness, expand analytical capacity, identify something that would otherwise be missed, or support a better decision?
What new risks does it introduce?
Could the system use incomplete information, disclose sensitive data, produce an unsupported conclusion, exceed its authority, or act without appropriate human review?
What controls should surround it?
Which sources may it access? Which rules must be deterministic? When is human approval required? What evidence must be retained? Who remains accountable for the outcome?
These questions should not be addressed separately. The value of an AI use case depends partly on whether its risks can be controlled without eliminating the efficiency it was intended to create.
Too little control creates unacceptable risk. Too much control can recreate the original manual process, leaving the organization with new technology but little practical improvement.
What This Looks Like in Practice
Consider a private company with dozens — or hundreds — of shareholders.
A shareholder emails the company with a question about a recent transaction, distribution, valuation, ownership position, tax document, or company development.
Answering the question may require someone to review the cap table, governing documents, financial reports, transaction records, previous correspondence, board-approved communications, and relevant public information.
The research may take hours. The response may take days. The quality of the answer can depend heavily on which employee receives the request and where that person knows to look.
AI can dramatically accelerate this work.
It can determine the shareholder’s intent, identify the relevant people and entities, search approved company records, gather supporting information, perform permitted external research, and prepare a draft response.
That is the value.
But the same process creates meaningful risks. The system could rely on an outdated ownership record, expose information belonging to another shareholder, confuse external research with verified company information, make an unauthorized statement, or provide an interpretation that should come from legal or tax counsel.
That is where control becomes essential.
A well-designed process separates the work into governed stages:
- The inquiry is captured and classified.
- The sender’s identity and relationship to the company are confirmed.
- Access is limited to information the sender is authorized to receive.
- Research is performed using approved internal and external sources.
- Material facts are linked to their supporting evidence.
- Legal, tax, valuation, or other sensitive questions are escalated.
- An authorized person reviews and approves the response.
- The inquiry, research, approval, and final communication are retained.
AI performs the time-consuming work of finding, organizing, and explaining the relevant information. The company retains authority over what is communicated.
The result is not merely a faster email. The company can respond more consistently, reduce the administrative burden on finance and management, preserve institutional knowledge, and provide better service to its shareholders — without surrendering control over sensitive information.
This is the balance finance leaders should expect from AI: greater speed without giving up evidence, authority, or accountability.
Three Things to Look For
Before introducing AI into a finance process, ask:
Can every important conclusion be traced to evidence?
The system should distinguish among verified company records, user-provided information, external research, and AI-generated interpretation. A polished answer is not a substitute for a supportable answer.
Does the system understand its authority?
AI may be permitted to research and draft without being permitted to approve, publish, transact, or make commitments on the company’s behalf. Those boundaries should be enforced by the workflow, not left to the model to infer.
Does the level of control reflect the consequence?
Drafting an internal summary does not require the same controls as communicating ownership information, changing a forecast, approving a payment, or making a representation to a shareholder. Controls should be proportionate to the data involved, the decision being supported, and the potential consequences of an error.
How We're Approaching It
We recently introduced shareholder management within Farsight DS.
The capability brings together cap tables, stakeholder records, governing documents, financial reports, prior communications, transaction history, and relevant company knowledge.
When a shareholder submits a question, Farsight DS can identify the intent, research the available information, assemble the supporting evidence, and queue a proposed response for review.
The workflow can restrict information according to the shareholder’s identity and relationship to the company. Sensitive questions can be escalated, authorized users retain approval authority, and the supporting research remains attached to the response.
The objective is not to automate the relationship between a company and its shareholders.
It is to give the people responsible for that relationship better information, dramatically reduce research time, and help them respond more consistently — with the controls the communication requires.
This illustrates a broader principle behind Farsight DS: AI should not operate as an uncontrolled layer placed on top of company information. It should participate in a defined business process with appropriate data access, rules, permissions, evidence, approvals, and accountability.
One Thing Worth Reading
COSO’s 2026 publication, Achieving Effective Internal Control Over Generative AI, applies the established Internal Control–Integrated Framework to generative AI.
Its importance for finance leaders is straightforward: AI governance does not need to exist apart from the company’s control environment. Organizations can extend familiar concepts — including responsibility, risk assessment, control activities, information quality, documentation, and monitoring — to processes in which AI participates.
That is a more useful starting point than treating every AI application as an isolated technology experiment.
Explore COSO’s guidance on internal control over generative AI
A Question for Your Team
Where could AI remove hours of research or preparation from a finance process — and what must remain under human control?
If you have an interesting example, email us. We read the responses.
See the shareholder inquiry workflow.
30 minutes: how a shareholder question is researched, checked against that shareholder’s information rights, and approved by a person before anything is sent.
Schedule a meeting →