Actively AI Raises $22.5M for 'Superintelligent' Sales Approach, Challenging Traditional AI SDRs

Actively AI raises $22.5M to offer sales 'superintelligence,' says AI SDRs failed | TechCrunch

AI sales rep startups are a very crowded market these days. If you're driving into San Francisco from the airport, you'll probably spot billboards promising that you can "Stop Hiring Humans" (Artisan) or urging you to "Hire Piper, the AI SDR" (Qualified). While some of these startups are certainly growing fast, the field has its challenges and some VCs are wary. Anshul Gupta, co-founder of Actively AI, admits the early versions of these AI sales tools don't live up to their own hype. Gupta claims classic AI sales reps aren't the right approach, telling TechCrunch they've "failed" by focusing too much "pure volume" - that means contacting as many potential customers as possible. Founded in 2022, Actively AI argues it has a different approach. The startup builds custom 'reasoning' models for companies to sift through their data and find the highest-value prospects to sell to, mirroring the work that top human sales reps do. It's a new way of leveraging reasoning tech, a technique that's taken the AI world by storm by forcing AI models to flesh out their logic and double-check their work. Actively claims this method is working, touting that it has helped clients like fintech Ramp get tens of millions of dollars in extra revenue. The New York-based startup has now raised $17.5 million in Series A funding from Bain Capital Ventures, it exclusively told TechCrunch. That follows a previously unannounced $5 million seed round from First Round Capital, bringing total funding to $22.5 million. "We call it 'GTM Superintelligence' -- a reasoning-driven approach that doesn't just automate or assist, but actively makes the best possible decisions to drive growth," Actively's CEO (and other co-founder) Mihir Garimella said in a statement. The startup says it uses a combination of in-house models and popular reasoning models from OpenAI and Anthropic to power its tech. Both founders previously studied AI at Stanford, with Garimella focusing on a field closely related to reasoning called active learning, giving Actively its name. Actively's fundraise is the latest evidence that the boom in reasoning models could be spreading beyond foundational AI companies like OpenAI or DeepSeek to startups. Just last week, for example, a YC-backed startup raised $5 million claiming it had built a 'reasoning engine' for slashing paperwork in healthcare. That startup, Taxo, said it had passed $1 million ARR in six months. (Actively declined to share its exact ARR, but said it has grown tenfold in 9 months.) It's still a bit early to tell whether Actively's reasoning-powered approach will work as advertised, or if this will become just a new spin on AI sales tools. After all, reasoning only really took off late last year with the rise of DeepSeek. For now, though, some investors are certainly buying the pitch.

Actively AI Raises $22 Million to Improve on 'Failed' AI Sales Reps | PYMNTS.com

Actively AI has reportedly raised $22.5 million to develop its AI-powered sales reps. Speaking with TechCrunch Wednesday (April 2), company Co-founder Anshul Gupta argued that classic versions of artificial intelligence (AI) sales reps had "failed" by concentrating too heavily on "pure volume," or contacting as many leads as possible. Actively AI, he said, takes a different tact, building custom "reasoning" models that let them comb through their data for highest-value prospects, similar to what a human sales rep would do. The report says this is a new way of leveraging reasoning tech. As covered here earlier this week, reasoning models in AI are designed to mimic the way humans engage in logical thinking, decision-making and problem-solving. "While generative AI models like ChatGPT use pattern recognition -- looking at sequences of content to make a calculated guess as to what comes next -- reasoning models take it one step further," PYMNTS wrote. "They follow steps of logic, make inferences and explain how they reach conclusions. Their primary purpose is not just to generate content, but to reason through a problem space, whether that involves analyzing facts, answering complex questions or solving multi-step tasks." It's why a reasoning model like OpenAI's o1 tends to take longer before answering. For instance, load a 1,000-word document to ChatGPT and o1 will take a minute, letting users view the stages: Reasoning, example reasoning, tracing AI reasoning, harnessing hybrid techniques, advancing reasoning strategies, creating and enhancing, pinpointing differences, enhancing precision and other steps. "We call it 'GTM Superintelligence' -- a reasoning-driven approach that doesn't just automate or assist, but actively makes the best possible decisions to drive growth," Actively AI Co-founder and CEO Mihir Garimella said in a statement to TechCrunch. According to the report, the company says it powers its tech with a mix combination of in-house models and popular reasoning models from OpenAI and Anthropic. The two founders studied AI at Stanford, with Garimella focusing on a field related to reasoning known as active learning, which is where Actively gets its name. Looking at the difference between generative and reasoning models this week, PYMNTS wrote that while both "fall under the umbrella of AI, their goals, architectures and applications differ." Research from PYMNTS Intelligence shows that businesses are finding more use cases for generative AI and seeing positive results from their investments. One recent survey of 60 chief financial officers at American companies that made at least $1 billion in revenue last year found that 90% of these firms reported positive returns on investment. "Understanding the distinction between reasoning and generative AI models is crucial for selecting the right tool for a task," PYMNTS wrote. "A marketing copywriter may benefit from generative models that can come up with creative ideas for an ad campaign, but a financial analyst needs an AI that can reason through facts, evaluate trade-offs and mitigate errors."