The world is still buzzing from the excitement of OpenAI's Dev Day--and X continues to go off on the latest and greatest consumer AI applications. Yet as an investor who began investing in b2b software in 2017, I'm far more excited about AI's impact on enterprise & SMB productivity and believe this technological revolution will play out differently.
First I will discuss my upbringing as a software investor, why I gravitated towards this category, and how to navigate software investing in 2024.
The 2.0 tech wave was based around network effect driven businesses that benefitted from scale and user data as a flywheel. From 2017-2021, we saw a flourishing of fragmentation within the software landscape, as emergent businesses, each specializing in market niches, were able to successfully challenge large software behemoths, stretched too thin across to many categories. Whether it was DataDog & Splunk in APM, Crowdstrike or ZScaler, Zendesk and Five9 in customer service, Atlassian in development tools, Hubspot in Marketing, the software revolution demonstrated the viability of smaller players to challenge large markets where incumbents were asleep at the wheel, and in doing so accrued tremendous value for the venture firms that were early backers.
I gravitated towards this area as many of my peers entered the tech world, and joined or built business productivity and future of work applications. As an investor, I'm always thinking about a way to develop edge, how to be their first, and more importantly, how to have a predictive advantage, even if miniscule. The first step is to bet on the right market: it is an inevitability that certain developments will occur within the future of software, but it is not an inevitability whether certain consumer trends will take off. As such, I focus my aperture on b2b software businesses, because they are easier to predict future outcomes for since they are less discretionary. Furthermore, as an investor, I have developed systematic methods of surveying CTOs, CISOs, and other tech leaders in order to see where spend is moving, and understand how different tools are added or cut out of the stack across functions.
Some of these theses include:
- Software development will continue to grow in value, and as such will require more robust tooling and infrastructure to power the stack
- Software is becoming increasingly more mission critical, as such observability and security tools will increase in value
- Security testing needs to be embedded in the code-writing process, rather than done after the fact via inefficient review processes, thus frictionless, low-touch tools will need to emerge
Generative AI native capabilities fundamentally augment workflows, and this revolution in Generative AI or SaaS 4.0 will both accrue value to massive incumbents (big tech) as the past ML wave in 2018 did, but also open the frontier for nimble, early stage AI native businesses to flourish (leaving overvalued "Growth Equity" businesses in the dust). B2B Gen AI businesses will have the best potential for success, as the consumer internet is a hollowed out, paywalled, disinformation-and advertisement filled mess with comparatively less actionable data value. Companies on the other hand, have rich databases of customer interactions, security threats, marketing touchpoints, all of which can be unlocked
Paradigm shifts result in a reimagining of the tech stack
Trends I see playing out - the implications are where theses begin, from there you need to find N of 1 founders in each space:
- Increasing number and frequency of AI startups: as such, bets on tooling, DevOps, LLMOps, observability (langfuse, helicone.ai, context.ai), infrastructure decoupled from big iron, and all of the components for building businesses can be performant bets.
- Security testing will be shifted further into the developer function via low-touch software solutions and AI is the unlock that can embed this cheaply and quickly nullify.ai
- Companies that were not positioned as security will become security tools. In the world of AI there will be increasing social engineering scams, and humans are the most vulnerable part of the loop as employees and consumers have the potential to become breach targets. Voice recognition software from AI models like resemble.ai can be used to detect deepfake audio and previously have been used by the music and film industry for anti piracy tools, however, these tools may be able to expand into cyber markets
- AI will diffuse into the enterprise via a new crop of no code solutions, and thus dev tools will unlock a new, much larger, but less sticky TAM, the non-technical knowledge worker. Graphical and text base interfaces to quickly spin out LLM apps and link them with vectorDBs will emerge. DevTools no longer will assume users can code, and business apps will now treat their end-users like developers. (Hex already pivoting to non-technical user interface powered by natural language, large platforms like Salesforce will embrace extensibility as coding becomes easier). www.stack-ai.com
- Data lakes and data warehouses being operated simultaneous become challenging for companies by 2015 and 5 years ago databricks built the data lake house to solve this to query and govern all data sources and workloads together. 2024 will be the rise of the data intelligence platform. BI tools only represent one narrow (although important) slice of the overall data workloads, and as a result do not have visibility into the vast majority of the workloads happening, or the data's lineage and uses before it reaches the BI layer. Without visibility into these workloads, they cannot develop the deep semantic understanding necessary. As a result, these natural language Q&A capabilities have yet to see widespread adoption. With data intelligence platforms, BI tools will be able to leverage the underlying AI models for much richer functionality. Example: all customer data, what products they bought, customer support interactions all in a data lakehouse. They use AI to make that richer, so say someone complains about a feature in a customer support data, that feature can be automatically tagged. AI helps you get structure from unstructured data.
- Cost of compute is increasing, and an emergent category of software and hardware solutions will attempt to tackle this centml.ai
- Value will accrue to data platforms in this IQaaS era
What do you think are emergent dev tools?
ML Ops / AI Ops as new category tools for AI engineers
- Evaluation of AI is one of the many ‘last mile’ tools needed for AI in production that can prevent hallucinations in critical sectors
Where is the value from the software coming from. For a lot of the value of software in the past, it was about owning a workflow for people (CSM software like Vitally, they let you know when to reach out and send the email, so the value is storing that workflow and manage it). In AI, the wrappers was interesting because the value was the model, for next iteration, is owning the workflow or model the winning. When can we use AI to do all components of a workflow