Probably raises $9M to build a more reliable kind of AI

The recent $9 million raise for Probably speaks to a growing concern within the AI space: the reliability of generative AI. The promise of effortless data synthesis and insight generation is undeniably attractive, but the persistent issue of “hallucinations” – instances where AI confidently presents false information as fact – threatens to undermine user trust and limit practical application. As evidenced by recent consumer sentiment, with sixty percent of US consumers saying ‘AI’ in brand messaging is a turnoff Sixty percent of US consumers say ‘AI’ in brand messaging is a turnoff, survey finds, the ability to deliver accurate and verifiable results is becoming a critical differentiator. This isn't simply about aesthetics; it's about building a foundation for responsible AI adoption across industries, and it’s a challenge many are grappling with, as demonstrated by the recent layoffs at Robinhood, where attributing job cuts to AI restructuring appears to be falling flat Robinhood’s note on 10% layoffs shows blaming AI isn’t cutting it.

Probably’s focus on achieving accuracy comparable to deterministic systems—those built on rigid, pre-programmed logic—is a compelling and strategically important goal. While generative AI offers remarkable flexibility and the ability to handle nuanced queries, its inherent probabilistic nature makes it susceptible to error. The current landscape is filled with companies attempting to leverage AI for everything from meeting transcription Plaud says its software business topped $100M in ARR after shipping over 2M AI notetakers to complex data analysis. However, without a robust mechanism to ensure factual correctness, the potential for misuse or simply producing misleading results is significant. Probably’s approach suggests a belief that reliability shouldn't be an afterthought, but rather a core design principle. This resonates with the growing awareness that deploying AI without addressing these fundamental accuracy issues could ultimately prove counterproductive, hindering widespread adoption and fostering skepticism.

The challenge for Probably, and indeed for the entire field, lies in how to achieve this deterministic-level accuracy without sacrificing the inherent benefits of generative AI—its ability to extrapolate, synthesize, and generate creative solutions. Traditional deterministic systems are often brittle, struggling to adapt to new data or handle unexpected inputs. Simply reverting to rigid rule-based systems isn’t a viable long-term solution. The real innovation will be in developing techniques that allow generative models to self-verify their outputs, perhaps through mechanisms of cross-referencing, probabilistic reasoning, or integration with external knowledge bases. The $9 million investment signals a market validation of this need; users aren't just looking for AI that *sounds* intelligent, they require AI that *is* consistently reliable.

Ultimately, the success of companies like Probably will depend on their ability to translate this ambition into tangible results. The move towards more accountable and verifiable AI is inevitable, driven by both regulatory pressures and a growing user demand for accuracy. The coming months and years will be crucial in determining whether the industry can move beyond the hype and deliver on the promise of AI as a truly trustworthy and productive tool. It remains to be seen if probabilistic systems can genuinely bridge the gap to deterministic accuracy, but the pursuit of that goal is a critical step toward unlocking the full potential of AI-native data management.