Feasibly CEO Brian Connolly on Why Your Business AI Shouldn’t Be a Chatbot
A Note on Citation: The HBR article Using Gen AI for Early-Stage Market Research (July 18, 2025) is the authoritative anchor for this discussion.
Earlier this summer, Harvard Business Review sent waves through the innovation world with a key finding: Generative AI can be used as “synthetic customers” to speed up early-stage market research.
At Feasibly, the platform built to accelerate market and financial feasibility for real estate developers, this research wasn’t news, but it was validation. We sat down with CEO Brian Connolly to discuss how Feasibly takes HBR’s findings and applies them to the complex, multimillion-dollar decisions facing developers today.
The consensus is that the future of AI in business is not about chatbots but rather specialized expert systems.
1. The HBR Problem: The Speed-to-Insight Gap
The HBR authors highlighted that traditional research is “slow, expensive, and constrained in scope.” According to Connolly, this gap in speed was crippling many smaller and mid-sized real estate projects.
“When I was the managing principal of Victus Advisors, one of the most common complaints I heard, especially from architects and builders, is that their clients didn’t want to pay for or spend the time on a traditional feasibility report. We couldn’t serve them because our process took too long and we were too expensive.”
This inability to serve a large segment of the market—those who couldn’t wait 8-10 weeks or pay $40,000 to $80,000—was the tipping point for creating Feasibly.
The Feasibly Solution: Connolly noted that the platform runs a full, high-quality market and financial feasibility report in about 30 minutes.
“We’re literally delivering the product and the service 80-to-90 percent faster and 80-to-90 percent cheaper,” Connolly states. This shift from weeks to minutes fundamentally changes the risk equation for developers.
“From my experience, for developers and owners... time is risk.”
He explains that every week a project waits for validation, it faces interest rate risk, construction cost escalation, and the potential loss of lender or investor capital. Feasibly addresses the speed factor, helping clients “get the answers and the results and the viability reports that they need in days instead of months.”
2. Beyond Preference: The Difference Between Toothpaste and Towers
The HBR article focused on LLMs simulating consumer preferences (e.g., valuing fluoride in toothpaste). However, the authors also found that generic LLMs struggle with novelty and exaggerate interest. This is where Connolly draws a sharp line between consumer product testing and real estate validation.
“When it comes to market and financial validation of a real estate development project, it’s all about, What are the consumers and users in that space using now? What are they paying for now?” Connolly explains.
Feasibly’s core mission is to model fundability, not just preference.
“Nobody's going to lend you money as a commercial lender. Nobody’s going to invest in your real estate project... based upon some new unproven untested concept.”
Feasibly's models don’t chase after hypothetical concepts; they analyze the market for comparable, successful, newer projects to prove a concept’s viability. This means looking at what features, amenities, and pricing are driving value now to build a “fundable project concept.”
3. The Future of Business AI: Tasks, Not Chatbots
The HBR article implicitly suggests that LLMs need rigorous training to be useful. Connolly takes this further, explaining why Feasibly’s approach is superior to asking a generic AI bot for an analysis.
“The biggest mistake most people have about the way they think about the power of AI and LLMs for business is their heads immediately go to chatbots. That couldn't be further from the reality of how the power of AI and the use of LLMs is going to change the business world.”
Connolly emphasizes that Feasibly's success lies in its structure, developed by experts who know exactly how feasibility reports should be done.
“We know all of the minute steps, processes, procedures, and we built a software that even without pulling in the AI knows what those steps are,” he states. “We’re not asking (the AI) questions. We use the power of AI to perform tasks for us quickly and efficiently.”
The platform breaks the massive task of feasibility analysis into hundreds of microtasks, pulling in the appropriate LLMs, in-house databases, and trustworthy sources to complete each step.
4. Augmentation, Elevated: The Human Element is Indispensable
The HBR article concluded that AI must augment, not replace, human research. For Feasibly, this augmentation is managed through a “human in the loop” process, ensuring quality control and strategic input.
Connolly outlines two critical points of human intervention:
Client Input and Strategy: Feasibly’s team works with the client to define the right inputs. “You have to put in good inputs, you have to put in good data,” Connolly insists. This human-to-human consultation ensures the outcomes are high-quality because the inputs were sound.
Report Review and Quality Control: Every Feasibly customer is assigned a Customer Success Manager (who acts as an internal analyst). “Every time a report is generated, it’s our human analyst feeding the information in. It’s our human analyst reviewing the report, editing, making sure that everything looks good.”
This deliberate structure—using AI to perform microtasks under the guidance of a human analyst—ensures the data is accurate, the reports are high quality, and the nuanced, segment-specific insights that generic AI misses are applied to the final analysis.
In the future, the human element will also be key to addressing the HBR’s point about LLMs struggling with dynamic market conditions. Feasibly plans to use integrations to pull in forward-looking, real-time data to supplement its historical models, keeping the human analyst focused on interpreting the most current market shifts.
The Takeaway
Feasibly proves that when AI is transformed from a general-purpose chatbot into a specialized, expert system—trained by seasoned analysts to perform precise tasks—it doesn’t just cut costs; it creates a new level of confidence.
By blending human expertise with automated speed, Feasibly empowers developers to rigorously explore concepts that were previously too expensive or took too long to consider, ultimately leading to stronger, more viable projects.
