From Chatbot to Agent: The Next Step in How Professional Firms Use AI

Guest Writer: Adam Martocello, Founder - Lilac Automation

ROCHESTER, N.Y. — Most professional services firms already use artificial intelligence in some form. An accountant drafts a client email in ChatGPT. A law office summarizes a contract in a chat window. A marketing agency brainstorms campaign copy with a prompt. In nearly every case, the pattern is the same: a person types a question, the tool answers, and the person decides what to do next.

That pattern is starting to change. A growing number of firms are moving from AI they talk to toward AI that does the work on its own, a category the industry calls agents. The shift is already visible in professional services. Thomson Reuters' 2026 AI in Professional Services Report, a survey of more than 1,500 legal, tax, accounting, and corporate professionals across 27 countries, found that 15 percent of organizations have already adopted an agentic AI tool and another 53 percent are planning for or considering one.

The difference between a chatbot and an agent is not cosmetic. It determines whether AI saves an hour here and there or removes an entire task from a person's day.

What Separates a Chatbot From an Agent

A chatbot responds. An agent acts. When someone uses a chat interface, they stay in the loop for every step. They write the prompt, read the output, copy it somewhere useful, and repeat the cycle for the next task. The tool is capable, but it waits.

An agent is given a job and the permission to complete it. Instead of answering a single question, it works through a defined process: it can read an incoming message, decide what type of request it is, pull the relevant information, draft a response, and either send it or hand it to a person for approval. The work happens whether or not someone is sitting at the keyboard.

For a professional firm, that difference matters because the real drag on productivity is rarely the quality of any single answer. It is the number of small, repetitive steps that pull a skilled person away from billable work.

Customer Inquiry Triage as a First Use Case

One of the clearest places to start is customer inquiry triage, the steady stream of incoming questions that hits every firm's inbox.

A large share of those messages are routine in nature. A client asks for a document, a status update, an appointment, or a copy of an invoice. Each one is simple, but each one still requires a person to read it, understand it, and route it to the right place. Multiplied across a week, that work quietly consumes hours.

An agent built for triage reads each incoming message, classifies it, and handles the routine cases directly by drafting an answer or scheduling the request. Anything that requires professional judgment gets flagged and passed to the appropriate person with the context already attached. The staff no longer sort the inbox. They handle the cases that actually need them.

The reason triage works well as a starting point is that it is bounded. The task has a clear input, a clear set of outcomes, and a natural checkpoint where a human can review the agent's decision before anything goes out the door.

What Moving to an Agent Actually Requires

Moving from chat to an agent is less about buying a more powerful tool and more about defining the work precisely. An agent needs to know what counts as a routine request, what information it is allowed to access, what it can send without review, and what it must escalate.

That definition step is where most of the value is won or lost. A firm that can clearly describe how a task gets done today can usually hand that task to an agent. A firm that cannot is not ready to automate it, no matter which product it buys. In practice, the tool is one of the last decisions, not the first.

Guardrails matter for the same reason. Professional services carry obligations around confidentiality, accuracy, and client trust. An agent should operate with a limited, well-defined scope and a review step for anything consequential, particularly in the early weeks while the firm confirms it behaves as expected.

Where the Value Shows Up, and Where It Doesn't

The business case for agents is strong on paper and uneven in practice. In Writer's 2026 enterprise AI adoption survey of 2,400 executives and employees, 97 percent of executives said their company had deployed AI agents in the past year, and 52 percent of employees reported using them.

Yet the same research found that returns lag well behind adoption. Only 29 percent of organizations reported significant ROI from generative AI, and just 23 percent from AI agents. Professional services have the same blind spot: most firms are not measuring whether their AI is paying off. The Thomson Reuters report found that only 18 percent of firms track the ROI of their AI tools at all, and fewer still measure the effect on client satisfaction or revenue.

The gap between deployment and results usually traces back to the same cause. Firms buy the capability before they define the work. An agent pointed at a vague task produces vague value. An agent pointed at one clearly scoped process, such as inbox triage, produces a result a firm can measure in hours saved and response times shortened.

Why It Matters for Rochester

Small and mid-sized firms make up the majority of employers across the Rochester region, and professional services, including accounting, legal, consulting, and marketing practices, are a large part of that base. Most of these firms do not have an internal IT department to run an AI project, which makes the temptation to buy first and plan later especially costly.

The advantage for a regional firm is that the starting point does not require scale. It requires picking one high-volume, low-judgment task and defining it well enough to hand off. Customer inquiry triage is a common place to start, though the same approach fits other routine work: client intake, appointment scheduling, invoice follow-up, or assembling standard status updates. For a firm that gets the sequence right, the payoff is not a novelty chatbot. It is a recurring block of time returned to the people whose expertise clients are actually paying for.

---

Connected Know covers Rochester and Upstate New York business news, startups, technology, innovation, venture capital, manufacturing, and economic development—connecting the people, companies, and ideas shaping the region's future.

This article was written by Adam Martocello, Founder - Lilac Automation. Lilac Automation is a Rochester-based consultancy that helps teams become AI-native through training and building systems that handle time consuming workflows. Its work has saved clients thousands of hours a year on internal operations.