Building bespoke AI products to win in the age of software

7 minute read

The age of mass manufacturing forged a popular managerial mindset anchored on a belief that there exists an optimal flow for any business process. For instance, an experienced welder should be able to weld eighty shielded-metal-arc-carbon-steel-joints in an hour, with less than one percent defects, and it takes roughly eight weeks and three hundred and twenty hours of training to become proficient. That mindset culminated in Taylorism (i.e., Scientific management) and Fordism at the beginning of the 20th century.

1. Winners in the age of services learned to apply Taylorism at scale

After the age of manufacturing came the age of services but many elements of Taylorism persisted because the most successful businesses during the transition were the ones who had applied those principles best. The winners applied the principles globally via branding and acquisitions, and achieved mass economies of scale and scope.

For example, Disney trained legions of employees on how to provide a distinctive customer service using employee evaluation benchmarks and training handbooks that reverse engineer our limbic system to create instant joy and happiness. Similarly, another incremental improvement on Taylorism is the famous Toyota Production System, which focuses on waste reduction and minimization of process inconsistency and overburden.

2. Most leaders today were forged in the age of services

The majority of today’s C-suites, boards, and government leaders have been academically trained and have had their formative work experiences in the age of services. It could therefore be counter-intuitive for them to understand how the future is going to be dominated by organizations that today put AI product development (and software development) as the most important investment they can make.

Furthermore, short term bottom line concerns are exacerbating the problem: the very nature of making a budget and forecasts where the return on investment is a percentage improvement over prior years fails to capture the dramatic impact that a successful AI product will have. Even planning a fiscal year using the popular waterfall model, while comforting, is counter-productive to the discovery of innovative AI products with great user experiences. In fact, if CEOs were agents in a reinforcement learning environment, we could say that many fail to do proper “exploration” and will inevitably experience a high level of “regret” down the line because they are not considering the long term consequences of their actions.

3. Leaders can learn to win in the age of software

I could argue that corporate leaders would greatly benefit from learning from the bottom: how to design (e.g., learn Figma and experience design sprints), how to code (e.g., learn Python), and how to solve problems with AI (e.g., learn PyTorch).

But, there are things all leaders can do at the top to influence their organization.

  1. First, they can adopt a venture capital mindset when investing in product development (see my post on how to weave your digital transformations out of silk to learn more on that topic).
  2. Second, they can learn from startups, especially how they fail (see Paul Graham’s essay on the hardest lessons for startups to learn, and an audio version of it).
  3. Third, they can make room on their board and leadership teams for people who are proficient with some of the tools mentioned above and who continue to be hands on with building AI products.

One major benefit of having decision makers at the top who are building AI products is an increased willingness to build rather than buy an AI product. These decisions are absolutely critical because whenever an organization decides to buy from a vendor, they are essentially conceding a significant portion of their identity, control and ability to grow as an organization. When they build bespoke AI products for which they own all of the IP, they are creating a differentiated customer experience and shareholder value for decades to come.

4. Winners in the age of software will build proprietary AI products

In the age of software, innovative organizations invest significantly to build their own proprietary digital products. These products win because their user experience leverages AI including both the more recent advances in machine learning (e.g., predictions, recommendations, and explorations) but also more foundational techniques such as operations research and optimization which have been around for a good 60+ years but are still widely underutilized in 2020 - which perhaps speaks to the precarious situation of many companies since those are easily explainable.

This omnipresent AI software layer opens a new dimension where products and services merge into one user experience. Whether the customer/user is a consumer, an employee, or another business, the cost, speed, and ease of use of the AI layer becomes one of the most important differentiators between winners and losers in globally competitive marketplaces. Internally, this AI layer provides innovative organizations with greater nimbleness, efficiency, and speed.

Two trailblazers who are examples of proprietary digital products that have achieved global scale over the past decades are Google and Amazon. At one end of the spectrum is Google (founded in 1998), an organization that showed the world how a pure product company could leverage AI (i.e., pagerank algorithm) to create a product (search) that provides unrivaled quality (relevant content), speed (fraction of a second), and cost (free). At the other end is Amazon (founded in 1994), an organization that applied this digital AI layer internally, to optimize the speed and efficiency of its supply chain from the distribution center to your door.

If you are looking for inspiration on what kinds of AI products you could build for your organization, I wrote a post on how to build a digital twin in 6 months for 1M USD.

5. There is room for improvement at both end of the spectrum

For every company that has entered the age of software (and started winning), there are hundreds that are still firmly in the age of services: banks, insurers, oil and energy producers, construction companies, transportation, hospitals, governments, and the list goes on.

Some companies seem reluctant to be first movers. For instance, American Airlines latest 10-K filing does not even mention “user experience” or “machine learning” once (or similar words). “Do you want ice in your water?” - that is a question I’ve been asked more times that I can remember and yet my answer is always the same. You can’t expect humans to know passenger preferences, but today’s technology can. Imagine adding a recommendation to the current app flight attendants use that shows if the passenger has status on the frequent flyer program and also what the customer’s preferences are in terms of food and beverage. It would improve the overall user experience, save a lot of time, and allow airlines to plan better the food/beverages on board.

Some early movers try to go fast but make critical mistakes or omissions such as in the domain of cybersecurity. For instance Capital One sees itself as a technology company doing banking and its CEO, Chairman and Founder seems to have read Gene Kim’s DevOps Handbook. And yet, Capital One was hacked in 2019, exposing records of over 100M people. An unfortunate event that could have been prevented. They appear to have since made very concrete improvements (notable hires) to address cybersecurity better. If you are interested on the topic of improving cybersecurity, I wrote a post on how ACID can protect against computer hacking.

Some darling innovators are pushing the limits, but they too have a lot of room for improvement. When the Model S first came out, some investors mocked it as an iPad with a big battery on wheels. I think a more suitable description of Tesla cars is that they are the first mass market consumer AI powered robot, and soon could make their way into our homes. But even Tesla has major product deficiencies. While the cars may be great, the experience of buying the car (which still relies on many industry/incumbent standards and processes) is only marginally better than other car manufacturers, and definitely worse than buying on Amazon or ordering a pizza from Domino’s.

6. How prepared is your organization today to win in the age of software?

Ask yourself one simple question: how many board members at your company have written code this week? This year? Since 2000? How many of them have gone through the PyTorch or Tensorflow tutorials to understand the capabilities and limitations of today’s AI resources? If more than half of them have and you are a publicly traded company, I would like to invest in you. If not, I recommend that you upgrade your board and your leadership team so that more of the critical decisions of your organization can be aligned with what lies ahead and not what has been. In plain words: If you don’t want to become a relic of the service economy, you better ingest new blood into your leadership team so that they can best incorporate bespoke AI solutions into your products, services, and internal processes.

Mik Kersten, author of Project to Product: How to Survive and Thrive in the Age of Digital Disruption with the Flow Framework, explains how a large number of companies went out of business because they were not able to adapt to the industrial revolution. He argues eloquently in his book that we are now in the midst of another one of these cycles that will witness many more great institutions disappearing.

In my next post, Finding your Asana to win in the age of software, I discuss how crucial your talent strategy is, and the necessity of looking outside your organization for help.