I think an ideal team should complement each other on a personal as well as professional level. In terms of professional level, it’s usually very clear; the team should have all the hard skills for success: Sales, Tech, Product, Finance, and Operations. Equally important is the personal bond (some might argue that it’s even more important).
Running a startup involves a lot of uncertainty, which frequently places pressure on the founding team’s structure, roles and expectations. If your team doesn’t bond well and doesn’t match on a personal level, every change, conflict, or challenge is greatly amplified from a local one to a global company issue. This in turn makes the team slower and less optimized in its ability to make decisions or execute processes. Moreover, having solid discussions and plans is difficult if there are emotional residues in the background. Personal bonds influence completing soft professional skills. For example, if one member is strong at seeing the “big picture,” they should be complemented and balanced by a member who can see the “small picture” of tomorrow’s details.
1. Identify market patterns and know how to adapt. Whatever you think you’ll do will change as time goes by. Founders should constantly listen to their prospects, customers, and technology trends to find the golden path built from market size, product, competitive landscape, and business model.
2. A common sellable vision. Founders are constantly selling their vision, which eventually turns into a product. The initial buyers are always new team members that should believe the vision, business, and the team they’re buying from. Next will come the first customers, usually buying more vision than product, and finally, the investors that look at the complete story and buy shares of a combination of vision, team, and product.
3. Focused execution. Whatever your plans are, execution will be the deciding factor between planning and reality. The team should execute plans on time, ensure quality and provide deliverables that can be adapted to an incomplete execution. A good execution team always has solid and dynamic prioritization and planning. Execution is not getting things done on time, but rather getting the minimal deliverables required to achieve a very clear goal at the right time. Failing to set realistic goals or delivering on time is where execution starts to fail. Balanced execution will use minimal resources to build the minimal features required to sell. It’s easy to ask for more features in order to improve a product, have more resources for faster development, and more cash to generate more leads. But ultimately, execution in a constrained environment (fewer developers, features, sales budget) is key.
My biggest fear was that users/customers will find what we do useless and that no one would care about the problem we solve; that it would be regarded as insignificant. This is naturally translated to many aspects: the future of the AI technology stack (will data labeling become the core activity?), the competition (will we be able to compete against better-funded competitors?), the business model (will the data labeling activity transform to training data pipelines?) and the market adoption (will every business need to label data in years to come?).
Constant learning: articles, books, lectures, and courses. Naturally, learning without experimenting has a lower impact so it’s important to constantly take action items from what I’ve learned and apply them to real life.
Focus. I use a term called TTV (Time to Vision). Focus inherently increases TTV; you spend a lot of time in a very narrow area (your core) while vision is the ability to dream big. Where do you put your focus and still make progress on your vision?
Dataloop is facing a huge opportunity involving building a company with thousands of employees working to build AI cloud services. We had to identify our business core and focus on it, believing that if that core is strong and meaningful, it will carry the company and product to its goal.
Paradoxically, we started an MLOps platform with virtually zero attention to AI models and instead focused on a data-centric platform of management, labeling, and pipelines. While all around us the focus was on model-centric AI, we focused on building a data-centric company that supports AI. With the core assumption that there are no barriers to AI model management, experiment monitoring a data-centric core, we were able to support capabilities that will bring us into the future, either organically or by acquisitions, enabling us to be more competitive and maintain the gap, compared to MLOps platforms that were developed on a model-centric core.
It’s worth noting that even our data focus was narrow; for 3 years we focused on visual data only (no text, no audio, no signals), supported by management and labeling (no pipelines). Choosing what not to do is critical. Moving forward I expect the challenge will continue, where we carefully define our core and focus our capabilities as we outsource everything else: ExO – the exceptional organization!