Like my great mentors told me, I advocate first time entrepreneurs focus on simplicity when building their first products. I tell them to focus on a single product with a concise vision, that the odds of succeeding are already stacked against us so there is no need to further complicate the process.

With one highly focused vision, I achieved lifestyle startup success at my first company, Planus, using this exact line of thinking. But I learned first hand at my second company, Granite Networks, the detriment of launching with a multi-product company.

At Granite, we attempted to launch with 3 products simultaneously. We thought our users would need a range of functionality — a full suite rather than a single module– in order to benefit. As it turned out, we were wrong.

Our first product, which required significant one time upfront costs generated over 90% of our revenue, and we broke even near our eighteenth month in business. However, alongside this success, we hemorrhaged $60K/month on the remaining products, which only generated approximately 10% of our revenue. I swore never to repeat the mistake of launching with multiple products again following this rookie move and subsequent costly disaster.

Learning from my mistakes, I religiously adopted a lean startup mentality at my third startup, ROOT Data Center. ROOT became cash flow positive within 24 months and, in less than 5 years, quickly grew to become the third largest Data Center provider in Canada– about 10 times the size of my initial forecast. This solidified for me the importance of doing one thing, and only one thing, really really well.

Until now.

Contrary to the single-product launch rule of thumb espoused by so many wise industry mentors and the mountain of supporting evidence, I posit that in the AI space this line of thinking could not be more wrong. Does that not defy the logic laid out in my preceding paragraphs, you might be asking?

Put simply: no. I explain:

When it comes to data, more is better.

Unlike with traditional software, developing a significant amount of data is required in order to train a Machine Learning/Artificial Intelligence pipeline. Some types of generic data are publically available through large open source networks, like ConceptNet. Other types of data can be acquired through systems, like with Mechanical Turk. However, more sensitive proprietary data is only available when permission is directly granted by users. It is access to this proprietary data that is critical for training AI models that address specific pain points and why developing multiple products in the AI space is really a means of collecting more of this data than it is about the product itself. Multiple products serve as direct pipelines to a range of proprietary data which in turn compresses the process of data collection and accelerates the rate at which you can deliver increased value to customers. With multiple data acquisition pipelines, the advantage is exponential, just like with accruing interest in a bank account.

There is no “I” in R&D

It’s no secret that researchers work better as a team. Achieving a minimum scale in a research team allows them to explore ideas in parallel and helps improve the probability that a radically innovative outlier idea will surface. This is a complexity that does not typically need to be addressed as part of a typical software development process, where one idea often guides researchers.

The ‘C’ in C-Suite stands for challenge

If there is one thing that really helps when it comes to attracting and attaining the best minds to the top of your organization, it’s making sure that they have complex and challenging issues to work on day in and day out from the very beginning. But in order to justify the overhead of an executive team whose titles are appropriately paired with their skill set, your organization needs to achieve a minimum scale, which multiple data pipelines will assist.

Time isn’t all we have

At Soul Labs we believe we have hit a local maximum in terms of remaining market opportunity and development affordability in the AI space. The talent pool supply shortage of qualified engineers, and researchers will be unblocked in the coming years which will accelerate commercialization efforts. Striking the appropriate balance of risk and velocity are a critical element of any business plan. Scaling too aggressively such that there is multi-layered removal of the founding culture from new employees before that culture can be properly ingrained is a mistake that would likely tip the risk/reward scale of a business model in an unfavourable direction.

There are of course risks involved with these points above that need to be managed very carefully. These include:

  • Resource conflict:
    • Top talent within an organization also has a tendency to want to be involved in the most interesting projects. To help counter this, it is important to ensure it is made crystal clear which team members are assigned to which projects. Robust, scalable roadmap products will be a critical part of keeping team members from peaking over the fence at whatever other projects are underway and becoming distracted. The conversation should likely be one of: “Do we kill the project or move forward?” not “What do we give Johnny to do during this product development lull?”
  • Vision management:
    •  It is easy for a vision to become obscure over time. It is important to do checkpoints at appropriate intervals to make sure everyone and every project are still aligned. Boards serve as great forcing functions for this. Strategy weeks and weekends serve as another critical element to ensure rogue agents and management fractions do not occur.

 

That said, no matter what advice is followed or strategy is implemented, it’s impossible to definitively predict which approach will guarantee success. It’s possible a single product hits a home run and consumes your team’s attention. It’s also possible a technological development occurs that completely negates the value of data as part of the development pipeline, forcing a core business hypothesis to undergo review.

The only thing any entrepreneur in the AI space or otherwise can be sure of is the need for an agile team that is focused on the present and the future of your organization, one that is open to whatever may be thrown in their path along the way.

At Soul Labs, we are in the middle of following, and thus testing, each of these points. We aim to maximize our potential for success by remaining laser-focused on our vision, and ensuring we are prepared to pivot when there is a well-founded reason to do so.

if you’ve undergone a multi-product launch or have comments on my line of thinking, please reach out and tell us about your experience and how you and your business were affected.

 

– Jason
Soul Labs Co-Founder

 

About The Author

Jason van Gaal, CEO of Soul Labs, is a four-time serial entrepreneur with an insatiable desire to tackle complex problems, and an objective of building a business of adequate scale to create a meaningful and positive impact in Canada’s technology landscape.

Share This