The last time Pramod Sharma was raising capital for an artificial intelligence startup, the industry was in a very different place. Generative AI products such as ChatGPT and Google Gemini were not yet household names, and while startups in the space could still attract venture capital, investors were years away from the fever-pitch AI fervor they've settled into since then.

Still, after launching in 2013, Sharma and co-founder Jerome Scholler were able to get enough backing for their game company Osmo--which used computer vision to connect tablet apps with physical toys--that the company was worth $120 million by the time it got acquired in 2019.

Now Sharma and Scholler are back for more. On Wednesday, the former Google engineers took their latest startup, Napkin, out of stealth, announcing that they'd pulled together $10 million in seed funding from Accel and CRV. Napkin's core tool is meant to make it easier to add visuals to workplace documents: It looks like a standard text editor, but can auto-generate editable diagrams, graphs, and charts based on pre-existing writing.

The idea was born in part from the duo's first startup experience. Sharma says he spent a lot of time at Osmo trying to communicate effectively but found all the corporate slide decks and documents to be a drag.

"Communication is so draining at a workplace," he tells Inc., adding that, while diagrams and graphics seem to keep people more engaged, employees often lack the skills to make nice ones. "That's when we had this aha! moment. This is ripe for automation."

Generative AI is pulling in a lot of cash right now, but it's not always small startups that stand to benefit. Crunchbase recently found that although AI raked in a record $24.1 billion in venture funding during the second quarter of 2024, the number of AI deals has actually been falling since Q3 2023--down from 1,825 to 1,001--and it was the size of deals that accounted for the sector's growth. According to PitchBook, investors have adopted a quality-over-quantity mindset and are raising their standards for pre-seed and seed deals. There's even been speculation that the AI sector is in a bubble.

"Raising funding is always hard," says Sharma, Napkin's CEO, who has nevertheless pulled it off twice now. "What really has worked for us is going after a really hard problem, and [having] a lot of technical expertise in that problem space."

A common criticism of many smaller AI firms is that they're built atop preexisting AI models, such as ChatGPT, but add little innovation beyond superficial tweaks. (Developers frequently build new apps around preexisting ones using something called an API.) That dynamic puts these "wrapper" AI firms in a vulnerable position.

But Napkin has a strong value proposition "because we are so deep into [the] visual space, and this itself is extremely complex," Sharma says. "The level of depth has definitely been very helpful for us as we raise capital."

Sharma raised his first round of capital for Osmo over a decade ago. Since then, he says, it's gotten easier to raise an initial round of funding but harder to move on to a subsequent one.

"Now there are more people who can give you the first angel round or friends and family round, but ... it's harder to graduate," he explains. "The end of the funnel is about the same, but the starting of the funnel has been getting much, much larger."

Still, some things have stayed the same. In addition to funding from CRV, Napkin is being backed in its seed round by Accel, a fund which previously invested in Osmo. It's no coincidence: Sharma said he wanted to balance one new investor (that could bring a novel perspective to the team) with one old one (that he already had a rapport with).

Rich Wong, a partner at Accel, says he is optimistic about Napkin's strategy of depth over breadth in the crowded AI field. (It's a common refrain from VCs.)

"Napkin is not trying to be all things to all people," Wong tells Inc. "It's clearly a narrower, specific use case. But we think there's a better chance, a likelihood, that we can become the best of breed in that specific thing."

Asked what advice he would give other AI firms that want to attract venture capital, he offers: "Pick a lane. Pick a use case. Pick a vertical. Try to solve a real problem."