At one end, a few elite engineers with exceptional architectural and business-function skills will command fat pay-cheques to troubleshoot code generated by artificial intelligence (AI). On the other, leaner teams of low-level engineers who generate code using AI assistants will replace the vast teams once required to execute large-scale software projects.
Also Read: The good news is that AI is hiring: The bad news is that it’s not us
Insofar as the technology startup space represents the leading edge of what eventually transpires in the software industry, this phenomenon is already playing out. Before the rise of Generative AI’s capability to write syntactically perfect base-level code in any programming language, software startups looking for investment from venture capital (VC) firms hunted for large rounds of finance to pay small armies of programmers to write software code. This is now changing.
The previous startup funding model for software-oriented companies came with received wisdom that went like this: first, create a minimally viable product (MVP); second, if possible, get a proof of concept (PoC) going that can satisfy the needs of potential customers; and third, hit the business-funding trail by making rounds of VC firms with a slick presentation that talks about the total addressable market, offering an entirely hypothetical revenue curve which rises slowly at first and then, beyond some future point, transforms into an exponential-growth curve that looks like a hockey stick.
Also Read: DeepSeek’s R1 may be the first of many AI super-apps to come
The only thing real in this sort of presentation was the ask for money—to put together an army of software programmers and fund sales-and-marketing efforts to promote the software among a raft of users (individual consumers or enterprises).
Now, while MVPs and PoCs remain the same, the approach to VC funding is beginning to change. Startups are betting that they do not need an army of programmers and can instead accomplish most of the task of building their product by using GenAI to spew out viable code—which, of course, would next need to be vetted and fine-tuned for business-function capability by a smaller team of experts who are intimately familiar with the startup’s business model.
This phenomenon has even led Sam Altman, the head of OpenAI, to comment that it is probably possible for a single-person startup to reach a billion-dollar valuation simply by using AI.
Also Read: Arun Maira: Don’t let techo-optimism over AI crowd out concerns of equity
In a previous column, I had written that DeepSeek may be the first of such ‘killer apps’ created by leveraging larger ‘foundational models’ built at great cost by pioneering firms like OpenAI and Google.
What I am describing is a variation on this theme, where startups use foundational models for code generation aimed at a particular business function or business case, and are not trying to replace the foundational model itself, which is what DeepSeek sought to do (pardon the pun).
It appears that this view of leveraging foundational AI for killer apps was not far off the mark. The New York Times reports that an entirely new set of funding asks has emerged in Silicon Valley with AI-enabled startups now veering away from asking VCs for money to pay hundreds of programmers to create a new product. (bit.ly/3D7BQQT).
The NYT report points to several examples of newer age startups that are leveraging foundational models to get to high annual revenue rates within a short period of time (bit.ly/43iqxzQ).
Also Read: Indian VCs join a global race to back the next big AI disruptor
The advantage of this, of course, is that these startups need significantly less funds to reach revenue levels that would allow their businesses to sustain themselves without the need for much additional external funding after the first round or two. Any funding requirements after the initial small rounds will only go into expansion of marketing and sales efforts for the actual product, and not into building the product itself.
This turns the traditional VC model in the technology startup space on its head. The need for patient capital in such scenarios reduces significantly.
If a startup can reach financial self-sufficiency early, then its ability to either bootstrap its growth without outside funding or get up and running in the first place, and do so profitably with relatively small amounts of funding, calls into question the need for VC money in this sort of investment class.
Also Read: BeerBiceps fiasco: How did influencers get to flex artificial muscles anyway?
To me, this is akin to VC firms looking to fund a podcast or some other content creator on YouTube or some other such online platform. In most cases, the content creator already has access to enough funding to create his or her content; VC funding is only needed for the struggle to market the content widely so that it reaches the holy grail of ‘virality,’ which would lead to an explosion in the content’s reach and thus also its revenue-generation capacity either through paid advertising or subscriptions.
In cases such as these, the content’s quality and appeal are taken for granted, or at least the fact that this aspect doesn’t need to be funded is abundantly clear.
Maybe it is time for venture capitalists such as me to pause funding for software-as-a-service (SaaS) related startups, unless they have a quick path to delivering a fully functional software product within a few short months.
The author is co-founder of Siana Capital, a venture fund manager.
#transforming #role #venture #capital #startups
