How to win in this era of software

Building and backing startups has changed as software has evolved from a system of engagement to a system of intelligence. This changes what it takes to win for founders and investors.

It took a weekend like the one we just lived through for me to finally sit down and write about how (software) startup building and investing is evolving with the addition of an intelligence layer (thank you DeepSeek).

Note that I won’t comment on what DeepSeek might or might not mean for OpenAI, Nvidia, and others - plenty of people have done that already (this blog post is the best one I’ve read so far on this topic).

I haven’t written (publicly) in a bit as it’s been busy to say the least. If you’re new here - most of my posts are of more tactical nature vs this one - expect me to get back to those soon again.

That said, the release of DeepSeek’s R1 model is another clear data point supporting what we’ve believed for some time re software startup building and investing:

• Software’s TAM is exploding: The pace of innovation in software and AI is accelerating, and with it, the size of the opportunities software can target is growing exponentially. Software is no longer just a "system of engagement" focused on workflows - it’s become a "system of intelligence," meaning it’s now capable of tackling knowledge work at scale.

  

• Power of Open Source: Closed-source foundation model players are struggling to capture a significant portion of the value they create because open-source alternatives are highly competitive (thank you Zuck and now Deepseek). This dynamic keeps driving down the "cost of intelligence" (e.g., token prices), negatively impacting the economics of closed-source providers and positively impacting the economics of startups building on top of foundation models.

  

  Interesting coming from Mr. American Dynamism ;) - yet he’s not wrong.

• App layer wins: With intelligence becoming more powerful and cheaper at the same time, most value gets captured at the application layer. Coupled with the expanding TAM, this makes it a very exciting time to be a founder or investor in software startups.

  

  (Extinction-level event for those that heavily invested in closed source FM players)

Or as a founder we are fortunate to have backed put it (shoutout to Michael):

[DeepSeek’s latest release] is a game changer in terms of how much it's going to cost us and other companies to train on proprietary and contextual data. […] Building a company right now is extremely exciting in general, these progression make the internal economics insanely efficient.

20x TAM for systems of intelligence

Let’s first take a step back and discuss what it means that over the past year or two, software has moved from a system of engagement to a system of intelligence:

Until recently, software (in the B2B context) mainly enabled workflow efficiencies and incremental productivity gains (SaaS era). Which is roughly a $300B annual revenue opportunity.

During the SaaS era (2000 - 2024), software was mainly a system of engagement where code delivers workflow efficiency and incremental productivity gains. Solutions connect via deterministic APIs and break down with too much data. Success is about getting deeply entrenched in workflows / user habits with the goal of making users more productive.

We have now transitioned over to the “AI-native” era of software startups. Using code + intelligence, these startups can increasingly automate entire outcomes. With it, they can go after the far larger opportunity of disrupting knowledge work as a whole - a $5+ trillion annual revenue opportunity (~20x the size of the SaaS opportunity)

Software in this AI-native era (2024 onwards) is a system of intelligence that is about automating work. Solutions can connect much more broadly / fluid, and they get better with more data / usage. Success is about accessing (proprietary) customer data / workflows to train better (domain specific / vertical) models to deliver better customer outcomes. Increasingly, all of software has an intelligence layer to it.

Importantly, what we consider “AI-native” software startups aren’t research labs pushing the boundaries of AI innovation while burning through 100’s of $M (or $6M in DeepSeek’s case). Rather, they are the evolution of SaaS-era software companies built from the ground up leveraging AI at their core. They rethink old systems, reimagine customer value, and build software businesses that couldn’t have existed before.

A step change

Compared to their SaaS-era peers, AI-native startups have significant advantages:

• Higher customer value: Intelligent software doesn’t “just” improve productivity, it automates outcomes. This creates much greater value while increasingly shifting human labor from doing the work to managing the systems that do.

• Build better products much faster: What once required 10 engineers and six months can now be done by 2 engineers in days. Speed is everything in startups, and the gap between the fastest and the rest is widening - from 2x to 10x or more.

• Products get better with more data: SaaS tools often struggle with large amounts of data while AI-native software thrives on it - it improves exponentially with more data, becoming better for all users / more personalized at the individual level (I’m not surprised if a “warm up” period for new products where they provide significantly more value after a while become “a thing”).

• Much higher capital efficiency: Increased speed and reduced development costs mean the best AI-native startups need less capital to get much further - both in terms of product and revenue.

• Increased importance of Go-To-Market (GTM): Every intelligent software product needs training and real-world data. With everyone leveraging a similar underlying technology stack, only those that can access proprietary customer data and workflows to train better models can build differentiated products and win. With it, nailing GTM / distribution is now even more of a day 1 priority.

How AI-native startups win

Even more so than in the SaaS era, AI-native startups that want to win need to:

• Real, tangible customer value: The “value” bar for products keeps rising - to win, startups need to be able to fully automate clear and valuable outcomes from the start and increasingly in a better way than humans could (even if it means first selling a service that is then automated). Doing one thing exceptionally well first is even more important now. Thereby, focus on core a workflow with high impact, and make sure to capture relevant data that drives further improvement.

• Speed: Speed is everything and AI tools allow you to supercharge every part of the business. Every person in your startup needs to be on top of the latest and greatest in terms of AI tech stack for their functions given the speed of progress. In line with that, below’s tweet isn’t just an important interview question - understanding the AI stack of startups has become part of our diligence process as well for the same reason (speed).

  

  Along the same lines, having a model agnostic architecture that allows you to plug different AI models into your stack is critical to have the ability to get the efficiency gains from leaps forward by anyone (as we just saw with DeepSeek).

• Distribution: It’s worth reiterating - nailing distribution from the start is now more important than ever. Writing code is getting commoditized - accessing data to better automate outcomes is everything to differentiate.

  

• Speed of progress: This is the least tangible one yet it’s equally as important as the other points - assume that the underlying models will get exponentially better and their cost exponentially cheaper. Build towards a future that accounts for that.

How VC’s compete in the AI-native era

The rise of AI-native startups is also changing the game for early-stage investors.

---

As a quick reminder: At focal, we exclusively back strong technical AI-native software founders at the very start and lead / co-lead their pre-seed. That’s all we do.

---

While I’m biased - this is how we think early stage investors win at a tactical level (i.e. outside of betting on the right teams and ideas):

• Invest first: With increased capital efficiency from the start, the best AI-native startups burn less to get further. Meaning many of the best will only raise one “lift off” round at the start. Thereafter, revenue funds exponential growth - until they eventually raise a larger “expansion round” down the road (vs a pre-seed, seed, etc). With fewer early entry points, early stage investors who don’t enter in the initial “lift off” round increasingly risk missing out all together.

At focal, we work with founders as early as possible (often prior to incubation) to then catalyze and lead their “lift off” (pre-seed) round with up to $1M. That’s all we do.

• Match founder speed: The best move the fastest - similar to startup building, speed remains a big differentiator in early stage investing. A firm that wants to win needs to match the velocity of the founders they aim to back.

We designed our processes around getting to know many of the best founders at the start, and a highly selective process from there (the initial meeting) to have the bandwidth to pick and win the best opportunities the fastest, supported by a concentrated approach to our investing, and our sole focus on being “lift off” stage specialists.

Bridge the GTM gap: The best technical founders build the best products the fastest but often struggle with sales and marketing. Firms offering actually impactful GTM “lift off” support have a clear edge in getting to partner with the best technical teams.

With our concentrated approach, we can spend a lot of time with founders early on, thereby over indexing on helping them nail their early GTM - together with the The GTM Circle we run (a peer network of 175+ GTM executives at leading Seed through Series D software startups in the US. This is an area we'll continue to invest in as a firm).

---

This is an incredible time to be building and investing in software startups.

If you’re a technical founder building an AI-native software startup that is selling to businesses and this resonates, I’d love to chat.