Shreyas Sampath has spent more than a decade guiding large enterprises through some of the most demanding finance projects they undertake. As a Manager in the Finance Transformation practice at a Big 4 consulting firm, he has led ERP implementations, M&A finance integrations, and chart-of-accounts migrations for Fortune 500 clients across the technology, consumer, and life sciences sectors. His recent work includes standing up a full ERP implementation for a Fortune 50 client in under five months, an engagement that shaped much of his thinking on how finance teams should approach speed, governance, and risk. In this conversation, Shreyas shares his perspective on where AI is beginning to reshape financial reporting and compliance, why regulated environments demand a different adoption playbook, and what CFOs should be asking before they bring automation into the close process.
You’ve spent over a decade in finance transformation at a Big Four accounting firm, most recently leading ERP implementations and M&A finance integrations for Fortune 500 clients. How has the conversation around AI in finance functions shifted in the last 18 to 24 months among the CFOs and controllers you work with?
In the last 18-24 months, I’ve observed that CFOs and Controllers are being more cognizant about the implementation and usage of AI, due to the rising access and token costs. Finance executives are investing time and effort in identifying and targeting pain points that can be solved using AI, rather than procuring additional technology applications. I have been working for a few clients over the last couple of months and helping them identify these pain points by conducting workshops and client interviews and analyzing whether these issues can be solved by either a process change, training or if it truly requires the use of AI. Executives are realizing that investing this initial time and effort will benefit in the long term with respect to maintenance and scalability of their technology stack.
There’s a lot of enthusiasm about AI in finance, but adoption inside the close and reporting cycle has been slower than in areas like FP&A or decision support. From your client work, what explains that gap?
The close and reporting space is a highly regulated environment with checks and balances maintained in every layer. I believe that areas such as FP&A or decision support fall under low-risk categories, where AI models can read data, interpret, analyze and provide recommendations to finance leaders. But in this case, there’s no external reporting or no financial/accounting impacts to the overall ledger. Finance leaders can choose to use this information to make decisions or simply ignore them, thus making them relatively low risk. On the other hand, when it comes to close/reporting, if the AI produces anything that’s incorrect, there could be severe real world consequences as it pertains to quarterly or annual reporting which could directly impact stock prices.
Regulated reporting environments come with SOX requirements, internal controls, and audit trails. Where are companies getting AI integration right in those environments, and where are you seeing the most common missteps?
Auditability and controls are core requirements when it comes to accounting. Companies that get it right are the ones that select applications that have been certified to be SOC compliant. One of the common missteps I’ve seen is companies trying to “vibe code” their own application rather than bringing in external consultants or procuring a SOC certified tool. At a first glance, it appears that some of these business problems can be solved in-house by coding a brand new application, but companies soon begin to realize the complexities around permission violations, audit trails, etc. By the time they come to realize these issues it’s usually too late with several hours and dollars invested and no viable solution visible in the near future.
You recently led a full NetSuite ERP implementation in under five months for a Fortune 50 client. When a transformation moves at that pace, how does AI factor into the design decisions, and what trade-offs come up between speed and audit readiness?
When it comes to implementing with speed, factoring AI into design decisions becomes a key lever to ensure timelines are met, as long as there are no risks from an audit standpoint. From a project implementation standpoint, I ensured that there was sufficient time in the design phase to align on the path forward on any requirements that require customizations that can be solved via AI. For example, one of the requirements was a completely custom dashboard that provided views for specific team members to view Purchase Orders, Invoices, Bills, Credits, and Returns, dynamically filtered by dates, with actionable To-Do statuses. The dashboards that are produced natively do not allow for these specific views. I had my team utilize Claude to create a NetSuite SuiteScript code that could be deployed to NetSuite that would accomplish this task. The team was able to do a POC in the initial stages and validate that this was a solution that could be possible. Without creating a POC, a use case like this would have normally been discovered and tested 50% of the way through the project, by which point it would be too late to come up with an alternative if the solution does not meet the needs.
A human sign-off operating model seems to be emerging as the default for AI in finance functions. Is that a transitional state, or do you think it’s the durable model for the next several years? What would have to change for that to evolve?
I believe that having a “human sign-off operating model” is the model that will sustain into the future. It is vital to have a human-in-the-loop model when it comes to accounting and reporting decisions. It will be an audit risk to replace the humans with AI agents when it comes to final approvals. The risk of incorrect payments to vendors or audit violations due to incorrect journal or invoice approvals, will prove to be extremely expensive to unwind, while potentially opening up clients to fines from the SEC. I believe that the most successful companies will be the ones that ensure a human review in every layer of the approval matrix.
Governance of AI in finance often defaults to IT ownership. Based on what you’ve seen at Fortune 500 clients, what does effective cross-functional governance actually look like in practice, and who needs to be at the table?
Governance of AI in finance definitely needs a cross-functional ownership between business teams and IT. The IT teams should own the deployment and management of the application, but the results produced by the application cannot be verified by the IT teams, which could lead to incorrect reporting metrics. In my experience, it is always the finance and accounting business operations that defines the requirements, and validates the results produced by the tool. At the end of the day, the purpose of the AI tool is to aid in reducing the workload of the team members working across business functions. Ensuring all business stakeholders and IT team members are at the table early on, will help establish distinct roles and responsibilities.
For finance leaders evaluating AI tools right now, what are the questions they should be asking vendors that they typically aren’t? And conversely, what vendor claims warrant the most skepticism?
For finance leaders evaluating AI tools right now, the primary question that needs to be asked is “what pain-point is being solved by AI that could not previously be solved”. Just because an application claims to be AI native, it does not mean that it outperforms tools that do not have in-built AI capabilities. Additionally, finance leaders need to understand how AI is deployed within the application and what this means to token usage and costs. For what vendor claims warrant the most skepticism, vendors need to be able to address scalability and customizability of the application. Several AI native ERPs, and other technologies, cannot be customized per the needs of a client or industry and any change to the application needs to be made by the vendor’s engineering team, which prevents certain client specific use cases from being satisfied. Ensuring that the application can be customized per client’s requirements will ensure long term success.
Looking ahead two to three years, where do you see AI creating the most meaningful change in finance functions, and where do you think the hype is outpacing reality?
Looking ahead two to three years, I believe companies that invest time and resources in identifying and solving targeted pain-points, using AI, will achieve meaningful gains. Based on my experience, I see financial reporting as one of the most impactful areas where AI will thrive. I believe companies that try and re-invent the wheel by building their own ERP solutions, or AI solutions, will be cautionary tales, and areas where I believe hype is outpacing reality.
