The most overlooked trends in enterprise tech in 2019

What was the most overlooked trend in enterprise in 2019 (and what might happen next)?

The demise of systems of record from the rise of AI. Today, systems of record are maintained through manual data entry, which can lead to data integrity issues and time-consuming data cleansing and prep. As AI has become more powerful, it’s started to automate data capture. (For instance, AI can create data records in Salesforce from email exchanges.) But if AI can identify what data should be entered and how to structure that data, then we may not need a traditional system of record. Rather than starting with the data you happen to have and that determining the questions you can ask, in the future, people will start with the questions they want answered and the AI will gather and organize the necessary data to answer those questions.

The margin crisis. A number of macro trends – the move to cloud, AI/ML, the diseconomics of scale for data, a bottom-up go-to-market (GTM), and the shift from products to services – are creating a downward pressure on margins, as I explored in a recent tweetstorm. This isn’t to say that enterprise startups should sacrifice the right GTM motion or product trade-offs to protect margins. Rather, the first generation of SaaS may no longer be the right point of comparison for an attractive margin profile.

Design-first, code-second product development. In the past, companies would often write code first and then build the user interface as an afterthought. However, users now expect all applications (even enterprise) to deliver the same ease of use, and design and UX have become a competitive advantage. We increasingly see more cycles spent on design iteration. Products are prototyped as a nearly functional app, beta tested with users, and then code built (or possibly auto-generated) from the design. With this design-first, code-second approach, products are more dynamic and fluid, and teams can iterate on new ideas faster. While there is no substitute for product-market fit, at least now, that fit depends on design and speed to implementation.

The consumerization of security. A number of enterprise security tools are now coming in from the bottom and then being adopted at the top. In 2019, bottom up adoption led to security hardware (such as Yubico and Thinkst) emerging at scale, and the adoption of open source tools (such as HashiCorp Vault). This trend is a major win for both enterprises – better products – and for consumers, since things are safer.

Design & development tools spread to other business functions. Traditionally, tools made for developers and designers have only been used by those same developers and designers. But, more and more, those tools are being adopted and used by non-technical roles. For example, sales teams use Postman, an API development platform, to demo product APIs, and product and marketing teams are creating content and wireframes in Figma, a collaboration tool originally made for designers. While the trend has been happening for a while, what’s new is the intensity: adjacent roles are not just using design and development tools, but using them as an important part of core workflows in their own roles.

Ben Horowitz is a Cofounder and General Partner at the venture capital firm Andreessen Horowitz.

• Follow
• X
• Linkedin

Martin Casado is a general partner at Andreessen Horowitz, where he focuses on enterprise investing.

• Follow
• X
• Linkedin

Peter Levine is a General Partner at Andreessen Horowitz where he focuses on enterprise investing. 

Joel de la Garza is a partner focused on information security related investments and other chaos adjacent businesses.

• Follow
• Linkedin

Jennifer Li is a partner at Andreessen Horowitz, where she focuses on enterprise investments in data infrastructure and analytics, open source, developer tools, and collaboration applications.

• Follow
• X
• Linkedin

Company Building: The advantage startups have in machine learning

Some machine learning startup founders feel a bit of “impostor syndrome” around competing with big companies, because (the argument goes) those companies have all the data. Yet startups can, and do, successfully compete with big companies. You can actually achieve great results in a lot of areas with a relatively small data set, if you build the right product on top of it. Jensen Harris, CTO and co-founder of Textio, and AJ Shankar, CEO and co-founder of Everlaw, share lessons learned about building an ML startup in this 2017 episode of the a16z Podcast.

“There are huge valuable areas where Google or Microsoft isn’t building specific, domain-specific products. They’re fundamentally, by and large, generic B2B or B2C companies that address these broad areas.” –AJ Shanker, CEO and Cofounder of Everlaw

“All of the algorithmic stuff in machine learning is going to be commodity. There are like 20 places in the world where they’re inventing new algorithms… But that stuff more and more rapidly ends up in the public domain. And so it’s not so much about that as it is, can you craft the thing that blends machine learning with other techniques, with statistical techniques, with user experience techniques, to build a product that has actual value?” –Jensen Harris, CTO and co-founder of Textio

“To achieve product-market fit, there’s a whole bunch of stuff beyond a giant corpus of data, and the latest deep learning algorithm.” –Steven Sinofsky, board partner, a16z

Field Notes: Quick takeaways from AWS re:invent

Machine learning (ML) developer tooling is evolving quickly. It was the area where the most new AWS services were announced. It’s part of a broader trend that cloud native startups may want to take advantage of: ML dev tooling that is automating data science tasks and workflows, such as provisioning instances, creating ML models, selecting algorithms, training and tuning models, detecting drift, and code reviews.

Interest in compliance and cloud security is higher than ever, partly due to the CapitalOne hack in August. Cloud security and correct configurations are important; leaving room for startups focused on securing the cloud.

Performance-monitoring tools are shifting from passive to proactive monitoring. New startups in logging and monitoring are introducing greater sophistication in proactive monitoring that products should take advantage of. For instance, Gremlin, a chaos engineering platform, injects chaos into a system to test how it performs rather than passively monitoring performance. So what does this mean? If you build it, don’t just expect them to come – the market is still learning why proactive monitoring is important, so founders will need to educate the market on its value.

Sarah Wang is a general partner on the Growth team at Andreessen Horowitz, where she focuses on enterprise technology companies.

• Follow
• X
• Linkedin