NEW YORK -- Carta began as a company known for digitizing paper stock certificates, struggling to get investors' attention. Still, over the years, it has evolved into a platform for private capital, with annual revenue of nearly $600 million, using a playbook that the company recently revamped and is now powered by AI.
Carta’s original strategy was to supercharge its business: Start with a service business, transform it into software, dominate the market and repeat, said Vrushali Paunikar, the company's chief product officer, during a presentation at the AI Agent Conference on Monday. While the equity management firm has found success with this approach, the growth of AI technology and agents has helped it magnify results.
“The greatest business opportunity out there in the world is taking a dated service business and turning it into a product business powered with AI,” Paunikar said during the presentation. She said the new business strategy starts with a service business, transforms it into an AI-enabled product business and then scale.
Carta’s refined business methodology is an example of how businesses are having to shift to incorporate generative and agentic AI. While the equity firm has found success with AI, some enterprises are still figuring out where the technology fits within their organizations.
The Experimentation Phase
For enterprises still trying to understand how to use AI, Paunikar advised starting small.
“Pick a very finite problem and experiment,” she said in an interview. She added that one of the things that accelerated Carta’s learning process with agents is using Claude from Anthropic.
“We started building like CLIs and plugins and skills for Claude to use,” she said. “That actually helped us learn a lot about agent behavior.”
Experimenting with AI tools is the key for enterprises, especially C-level executives, according to David Treat, global CTO at Pearson, an education and academic assessment company.
“You have to be hands-on to really understand the power or potential,” Treat said during a fireside chat, adding that when C-suite executives know how to work with AI tools, it makes a difference.
While practicing with AI tools is essential, enterprises should not view AI automation as just another layer to add to their workload, said Ali Alkhafaji, CEO of Apply Digital, a digital transformation company.
“Reimagine that process, that workload,” he said. “I guarantee you more often than not, you’ll find places where AI can help you truly transform the way you work.”
However, businesses need to avoid rushing or thinking they need to experiment with all AI tools; they need to be nimble and able to prototype the next best thing, said Masha Sharma, vice president of merchant experience at Groupon, in an interview.
“Frankly, I see people getting overwhelmed because you’re trying to be on the cutting edge of all of that and you’re thinking that you’re missing out,” she said. “You kind of have to slow down.”
Mistakes Will Happen
However, AI transformation in business does not mean no failures and mistakes, and enterprises experimenting with generative and agentic AI should know that they will face challenges that sometimes cannot be predetermined.
“It is incredibly difficult to put all the rules and policies in place to protect enterprises with AI because you don’t know where a potential incident is going to come from,” Alkhafaji said in an interview.
Even Carta, with its refined business model, failed some experiments before succeeding.
“There were some experiments we did on direct manipulation, It was disastrous,” Paunikar said. She added that Carta does not give agents access to its data. Agents can access its product only through workflows that have data health checks and validation.
To deal with these unknown variables, Alkhafaji said Applied Digital has set up a set of principles and guidelines that can be used to make decisions in real time within the company if an agent makes a mistake.
“We’re probably going to face a challenger moment at some point,” he added, referring to the AI market. “An incident will come up, and it is going to be a major brand. It is going to make many people think twice about AI. I do not think it is going to stop it, but it is certainly going to put up like an additional level of rigor that is needed today but not really adopted everywhere.”
The Human Factor
One way to avoid a disastrous moment during the experimentation phase is to keep a human in the loop.
“Where do you place value in general as a human is trust,” said Deepak Shrivastava, CEO of Sunrise AI, in an interview. He added that, with agents being trusted to make financial transactions, make purchasing decisions or even shop, that level of trust is increasing. “The best way to build that trust and maintain that trust is people, human to human.”
More than maintaining trust, the human-in-the-loop, or people factor, is also a way for enterprises to differentiate, Sharma said.
“Everything is going to start to look very much the same,” she said, referring to the responses from AI agents. She added that businesses that want to stand out and personalize their products or services will need humans to stay involved.
