AI in Accounting · · 8 min read
How AI Is Reshaping Work: What Tampa Bay Business Owners Need to Know
AI is already changing how work gets done, from who checks what to how fast your books close. The research is clear about the upside. It's just as clear that the payoff depends on the people and the process you build around it.
In brief
AI is reshaping work by taking over repetitive tasks and speeding up analysis, which makes human judgment more important, not less. Field studies show real gains: 14% higher productivity for support agents and a monthly close 7.5 days faster for accountants using AI. The results depend on clear review points, named owners, and people trained to question the output.
On this page
- AI is a support layer, not a replacement
- What the research actually shows
- Three questions every owner should be able to answer
- Roles are changing, not disappearing
- Speed isn't the same as certainty
- Design the workflow before you turn the tool on
- The human skills that matter more now
- Train on real work, not in a classroom
- Guardrails matter most where mistakes are expensive
- The ethics are a leadership job
- What this looks like at Brown Business Advisors
- Frequently Asked Questions
- A note on scope
- The bottom line
Most talk about AI and jobs swings between two extremes. Either it's going to replace everyone, or it's a toy. Neither matches what we see inside real businesses. AI is already part of everyday work: drafting emails, summarizing documents, categorizing transactions, forecasting cash. What it changes most isn't headcount. It's how the work is organized, who checks what, and where judgment sits.
Stanford's Graduate School of Business recently laid out how AI is reshaping the workplace, drawing on research from its own faculty and others. We read it through the lens we use every day: owner-run businesses in St. Petersburg and across Tampa Bay that want the efficiency without the mess. Here's what holds up, and what it means for you.
AI is a support layer, not a replacement
Right now, AI works best as a layer underneath the work. Generative tools help with writing, research, and first drafts. Analytics tools find patterns in large sets of data, forecast demand, and flag risk. In both cases the tool speeds things up and surfaces things you might have missed. It doesn't decide what to do about them.
That distinction matters, because the output still needs someone with judgment, and a process, to use it responsibly. A forecast nobody questions is just a confident guess.
What the research actually shows
The biggest gains go to less experienced people
One of the largest real-world studies so far comes from Erik Brynjolfsson, Danielle Li, and Lindsey Raymond, who followed more than 5,000 customer support agents as they started using a generative AI assistant. Productivity, measured as issues resolved per hour, rose 14% on average. For newer and less-skilled agents it jumped 34%. The most experienced agents saw little change.
The lesson for an owner: AI spreads know-how. It can help a second-year staffer perform closer to a ten-year veteran, because the tool is effectively passing along what your best people already do. That's a real advantage for a growing team. It also means quality control matters more, because more of the work is being shaped by a system instead of by someone's own experience.
In accounting, the time shifts to higher-value work
Closer to home for us, Jung Ho Choi of Stanford and Chloe Xie of MIT Sloan studied accountants using AI in the field. The ones who adopted it supported 55% more clients per week, moved about 8.5% of their time away from routine data entry toward client communication and quality review, and closed the books 7.5 days faster each month. Their general ledgers got more detailed, too.
The part we find most telling: experienced accountants didn't hand everything to the software. They used automation selectively and stepped in more when the AI signaled it wasn't sure. That's the model. The machine handles the volume. The professional knows when to overrule it.
Three questions every owner should be able to answer
Once AI starts producing work in your business, accountability has to be spelled out. Before you roll out any tool, you should be able to answer these:
- Who reviews what the AI produces before it goes to a client or into the books?
- How do errors get caught, and who fixes them?
- What standard tells your team what responsible use looks like?
If those answers are fuzzy, the tool will eventually get trusted where it shouldn't be, or ignored where it would actually help.
Roles are changing, not disappearing
As routine work gets automated, job descriptions move. Analysts are expected to explain what the numbers mean, not just produce the report. Managers are asked to redesign the process, not just supervise it. Performance becomes less about how much someone produces and more about the quality of their judgment and how well they spot risk.
That changes how you should evaluate your people. The bookkeeper who catches the one odd vendor payment is worth more than the one who keyed in the most transactions. It also pulls departments together, because an AI rollout touches operations, HR, finance, and compliance all at once.
Speed isn't the same as certainty
AI can make decisions faster. It can also make them feel more certain than they are. Susan Athey, an economist at Stanford GSB who studies technology, warns against putting "blind faith" in AI. As she puts it, "Machine learning solves simple problems, but it is not sentient. It struggles when applied to many business problems."
That's why so many companies run a promising AI pilot and then see little change in the business. Dropping a new tool into an old process rarely works. The value comes from redesigning the workflow around it: deciding what the tool handles, what a person checks, and what happens when the two disagree. Good AI leadership is about better-built systems, not just faster answers.
Design the workflow before you turn the tool on
The businesses getting this right treat AI like a capable new hire who needs clear instructions and supervision. In practice, that means deciding up front:
- Which tasks can be automated outright
- Where a person has to review or sign off
- Who owns the outcome when something goes wrong
Skip this step and teams drift to one of two extremes. Some trust the output too much and miss errors or bias. Others don't trust it at all and quietly stop using it. A well-designed workflow is really a form of governance. It spells out which decisions need human sign-off and which ones get checked over time for errors or bias.
Regulators are heading the same way. The EU's AI Act requires certain high-risk AI systems to be built so that people can effectively oversee them. Most Florida businesses won't fall under EU rules directly, but the direction is clear: oversight and accountability are expected to be built into AI-driven processes from day one, not bolted on later.
The human skills that matter more now
As AI takes on more routine thinking work, the skills that set people apart are the ones software doesn't have:
- Creativity
- Empathy
- Ethical reasoning
- Clear communication
- Seeing how all the pieces of a business fit together
These aren't soft skills anymore. They're how results get explained to clients and turned into decisions. Jennifer Aaker, a behavioral scientist at Stanford GSB, says the real question isn't what AI can do, but what kind of human experience it amplifies. "We have choices," she says. "We can build technology that harnesses our humanity or settle for tools that diminish us."
Train on real work, not in a classroom
New skills stick when they're tied to actual problems. Teams learn AI tools faster by using them on real tasks, with someone checking the results, than by sitting through abstract training. The Brynjolfsson study points the same way: the assistant helped newer workers improve faster, but those gains came alongside feedback and coaching. Access to a tool doesn't build skill on its own.
Treat AI like any other capability in your business, something you keep refining. Being adaptable will matter more than mastering any one product, because the products will keep changing.
Guardrails matter most where mistakes are expensive
Mohsen Bayati, who studies operations and technology at Stanford GSB, stresses that AI-driven decision systems need strong guardrails because their errors can hurt real people right away. In finance, the equivalent is a misstated return, a missed deadline, or a wrong number in front of a lender or a buyer. In high-stakes work, planning for how things can fail matters as much as chasing efficiency.
The ethics are a leadership job
AI raises questions no tool can answer for you:
- How are biases caught and corrected?
- Who is accountable for a decision the AI influenced?
- How does automation affect opportunity and fairness inside the company?
- What safeguards protect your people and your data?
Amazon learned this in public when it scrapped an experimental recruiting algorithm that, trained on years of past hiring data, marked down résumés from women. When screening leans heavily on AI, the structure around it, meaning who reviews it and how it's monitored, decides whether it widens opportunity or narrows it. The fix is the same everywhere: be open about where AI is used, check the outputs, give people a way to raise concerns, and learn from mistakes instead of hiding them.
What this looks like at Brown Business Advisors
We use AI the way the research says it works best. Automation handles the volume in bookkeeping, reconciliations, and reporting, so the books close faster and carry more detail. Every number that matters to your taxes, your lender, or a future buyer is still reviewed and signed off by a CPA. And when we build AI automation or custom software for a client, we design the review points and ownership first, then the tool.
Frequently Asked Questions
Will AI replace my bookkeeper or accountant?
Not on the evidence so far. In a field study of accountants using AI, adopters supported 55% more clients per week and shifted time from data entry to review and client work. The role moves toward judgment, and experienced professionals were the ones who knew when to overrule the software.
Where should a small business start with AI?
Start with repetitive, high-volume work that's easy to check, like categorizing transactions, matching receipts, or drafting routine documents. Decide who reviews the output and who owns the result before you turn anything on, then expand once the process is reliable.
Do Florida businesses have to follow AI regulations?
Most Florida small businesses don't fall under laws like the EU AI Act, but rules on data privacy, employment, and financial reporting still apply to anything AI produces. Building human oversight into AI workflows now is the safest way to stay ahead of where regulation is heading.
A note on scope
This article is general information about how AI is changing the workplace, adapted from Stanford Graduate School of Business's overview and the research it cites. It isn't legal, tax, or technology advice for your situation, and the right approach depends on your business. For guidance on putting AI to work in your finances, talk to Brown Business Advisors.
The bottom line
AI is changing how work gets done, but the research keeps landing in the same place: the technology pays off when people design the process around it. It spreads expertise, speeds up routine work, and frees up time for judgment. It doesn't own outcomes, catch its own blind spots, or answer for its mistakes. Those stay with you and the professionals you trust. If you want the efficiency of AI with a CPA standing behind every number, schedule a consultation with Brown Business Advisors.
Sources
- Stanford GSB Executive Education, "How AI Is Reshaping the Future of Work" (April 13, 2026)
- Brynjolfsson, Li & Raymond, "Generative AI at Work," The Quarterly Journal of Economics (2025)
- Choi & Xie, "Human + AI in Accounting: Early Evidence from the Field" (working paper)
- Stanford GSB Insights, "Susan Athey: Why Business Leaders Shouldn't Have Blind Faith in AI"
- Stanford GSB Insights, "Human First: Designing Artificial Intelligence That Elevates Us" (Jennifer Aaker)
- Stanford GSB Insights, "Don't Let Artificial Intelligence Pick Your Employees"
- EU Artificial Intelligence Act, Article 14: Human oversight
Written by Pat Bennett, Applied AI Engineering · Reviewed by Hunter Brown, CPA