 |
Every time January rolls around, venture capitalists bring
out their crystal balls on where they believe things are headed in the new
year. As an early-stage investor at Mayfield, I suppose I'm no different. But as a product guy
at heart, I try to stay out of the ivory tower, keep my ear to the ground, and
constantly meet with talented entrepreneurs, engineers, product managers, IT
executives, open-source maintainers, and other industry leaders within the
enterprise infrastructure community. It's these conversations that have driven
my areas of startup investment for 2020. If you're building or thinking of
building something in these areas, I'd love to talk with you.
Cloud Native Day 2
The world of cloud-native has changed. Six years ago, I was
a fiery (possibly misguided) product manager trying to tell anyone who would
listen how this whole container thing was going to revolutionize the way
applications are built. Now, Kubernetes has effectively been accepted as the
next-generation standard across enterprises and startups for hybrid-cloud
applications and infrastructure. 48% of enterprises adopted Kubernetes in 2019, almost
doubling from the year before. But as Lyft engineering manager Vicki Cheung stated in her
KubeCon keynote, "It's not all unicorns and rainbows." If you
spend any amount of time talking to DevOps and IT leaders, you will consistently
hear three things:
- Operationalizing
Kubernetes in production is hard.
- Teaching
novice users to work with Kubernetes is hard.
- Digital transformation, cost reduction, and security are the real top of mind priorities. Kubernetes is just part of the implementation.
Beginning the journey to cloud-native and selecting
Kubernetes was Day 1. We need to solve problems for Cloud Native Day 2--operations
that it takes to actually keep modernized applications and infrastructure up
and running. Examples of potential Day 2 problems include
monitoring/observability, troubleshooting and incident management,
multi/hybrid-cloud ops automation, security and policy management, and making
it easier for developers to build and manage microservices applications. These
are complex and difficult issues, but ones that IT and DevOps leaders have
significant budget to have solved.
Developer Augmentation
As a former product leader, I spent a lot of time talking
with leaders in engineering, product management, and design across both enterprises
and the startup world. There are several key evolving issues hitting these
product development teams over the next few years:
The cost of talent for engineers, product managers,
designers, and similar team members continues to rise. As a result, companies
are desperately looking to augment their existing headcount with force
multiplier tooling. In addition, product development leadership wants more
actionable insights into the effectiveness of their teams. Or, as one CTO put
it to me: "Why is it that my marketing and sales counterparts
have their pipelines fully instrumented, but I can't even tell if my product is
going to ship on time?"
In addition, product development personnel today are being
forced to use a dizzying array of tools to get their job done--tracking issues
on Jira,
code reviews on Github,
iterating on designs in Figma and Invision, not to
mention having to integrate microservices across a truly vast array of tooling
and frameworks. At the time of this writing there were 1,287 development tools
listed on the Cloud
Native Computing Foundation Landscape. Whatever happened to focusing on
building business logic?
Finally, the future of work is remote. Companies like Hashicorp (disclaimer:
Mayfield portfolio company), Gitlab, and Automattic have shown that fully distributed companies
at the scale of hundreds of employees are manageable, and tools like Zoom, Slack, and Github have led the
charge. But there are gaps across these workflows. The simple truth is that is
still not as easy to work remotely as it is to bring in a product development
team in an office and iterate rapidly together. That needs to change.
There are huge opportunities for startups that augment the
effectiveness of product development personnel (engineers, PMs, designers,
etc.) by automating repetitive workflows, stitching together disparate tooling,
facilitating collaboration of remote teams, and providing actionable insights
and analytics for product development managers and executives. I call this
area Developer Augmentation, and it's only just getting started.
Machine Learning for the Masses
There's a lot of buzz around Machine Learning these days.
Most software companies today focus on ML in one of two ways: either leveraging
ML capabilities to solve a particular use case (think Netflix's recommendation
engine) or providing software services to users of ML (e.g. labeling, training,
etc.). One of the unifying points across these companies is the need for data
scientists in order to leverage ML at scale. But there is a known shortage of
data scientists in the market; as far back as 2018, LinkedIn posited an unfilled demand of 150,000 data science jobs, and that
talent shortage continued well into 2019 and beyond. One solution is to train
more data scientists (which is happening), but perhaps this is a not quite
enough. Thus the real question is, which startups will truly unlock the power
of Machine Learning for the Masses?
I would divide this into two categories--unlocking ML for
technical folks (e.g. engineers, developers), and ML for non-technical folks
(e.g. business analysts). The first technical category opens up a smaller pool
of users but with incredible programming flexibility through the use of APIs
and integration points to existing tooling and frameworks. The second
non-technical category opens up a substantially larger pool of users, but
require tightly focused use cases and personas. Think about what Airtable did with
spreadsheets--provide people an interface they understand, and then put the
power of ML behind it.
In addition to opening up the pool of ML developers, once ML
really takes off in the enterprise, there will be a strong need for
infrastructure tooling for ML operators--that is, the software that manages,
monitors, and secures ML workflows in production environments. This is
effectively what DevOps and IT teams do for traditional applications today. The
world of ML is no different, and will require strong platforms for managing the
infrastructure across multi and hybrid cloud environments with integrations
into Kubernetes and other "conventional" cloud-native tooling.
Wrapping It Up
There are a lot of problems out there today in the world of
enterprise infrastructure. The good news is, where there are problems, there are
opportunities for startups. If you're working on or thinking about these areas
I'd love to chat with you. Happy hunting.
Originally published on LinkedIn
