
Artificial Intelligence (AI) is reinventing industry after industry. In China, AI is tutoring children in more than 1700 schools across 200 cities. In Australia, an AI created a flu vaccine that far outperformed all other existing flu vaccines. In the US, a machine-learning robot is autonomously cooking burgers. There are many incredible real world-implementations of AI. And now, with OpenAI’s recent ground-breaking language-generating algorithm called GPT-3, remarkably powerful AI solutions are pouring down like rain. Crafty developers have already deployed GPT-3 to autonomously write viral blog posts, generate web designs, and create role-playing adventures.
Machine learning technologies are more accessible than ever, but finding the business case for AI isn’t always straightforward. In this piece of writing, I would like to make AI business strategies more concrete by walking you through four AI strategies that you can use to improve any activity you could imagine. After walking through the four strategies, I’ll help you figure out which strategy to use for any given activity.
The Premise
But first, there is one central point to artificial intelligence that it is important that you understand: AI can empower any activity through either automation or augmentation.
- Automation is the removal of humans from an activity.
- Augmentation is the empowering of humans in an activity.
Automation and augmentation are opposite extremes, and few AI solutions are fully automated or fully augmented. Automation and augmentation is a scale that encompasses our four strategies.
- The efficiency strategy, in which activities are optimized through automation.
- The effectiveness strategy, in which activities are made seamless, enabling easier communication.
- The expert strategy, in which AI empowers decision-making.
- The innovation strategy, in which AI enables creativity.

The Efficiency Strategy
For many people, the first thing that comes to mind when they think about AI is automation. While there is a lot more to AI than just automation, as we will soon discuss, this particular strategy is indeed all about just that.
The efficiency strategy can be used for activities with very clearly defined rules and routines. Companies adopt this strategy to optimize their processes, generally with the intent of reducing costs.
For instance, the taxi industry is at risk of becoming fully automated in the not-so-distant future, as self-driving cars are outperforming humans. There are very clearly defined rules for how to drive a vehicle; hence driving can be optimized. Fraud detection is also generally automated these days, as it turned out that detecting fraud is quite straightforward. Warehouses are being increasingly populated with autonomous robots as well, and we are also starting to see staffless stores. Note that staffless stores are not without their issues, however. Some activities, such as maintaining fresh groceries, are challenging to automate.
As more routines are being discovered for complex activities and as machine learning algorithms are becoming ever more competent at comprehending data that was previously considered to be complex, more and more activities can be automated. However, even if an activity can very clearly be automated, it may be preferable not to do so for a number of reasons:
- In some businesses, customers may require a high level of service, beyond what machine learning AI is capable of providing today.
- Many medical, military, and financial processes could be automated already with the algorithms we possess in 2021. However, the decisions made in these industries often have severe consequences. For that reason, decision-makers in these trades may be hesitant to automate certain activities. It is sometimes preferable to develop augmentative AI solutions that empower human decision-making rather than replacing human decision-making altogether.
- The ethical challenges of automation are often far greater than one might imagine. Even Amazon, well-known for its competence in Machine Learning, developed a resume-scanning AI that became sexist. As companies rush to automate their business processes, risks are high that ethics will take a low priority. It’s easy to create a discriminatory algorithm unintentionally, which is why one should be careful with automating certain activities.
The Effectiveness Strategy
This second strategy revolves around using AI to empower the communication and coordination of workers. Here, AI takes the role of an assistant. Typically, the goal of the effectiveness strategy is to make humans more effective by either eliminating or simplifying the act of scheduling, communicating, or monitoring.
It has become trendy for companies to implement basic AI solutions with this strategy in recent years, thanks to the rise of chatbots. Artificial intelligence is often used to partially automate customer support errands. If a support errand proves too difficult for the AI to solve, a human agent can join the errand. Partial automation is common for activities built for this AI strategy.
The effectiveness strategy can be deployed for more complex activities than just customer support, however. AI can be used to autonomously schedule meetings between coworkers, for instance. In large consultancy firms, AI can be used to suggest which consultant(s) to assign to a project, based on skills, experience, and employee desires.
Applications built for services like Google Assistant, Siri, and Alexa tends to fall within this strategy. As over one-third of American adults own a smart speaker for their home, products built with the effectiveness strategy in mind have become crucial for consumer-facing enterprises.

The Expert Strategy
Unlike the two AI strategies mentioned above, the expert strategy is closely associated with augmentation. For activities that contain complicated work processes, typically with ad hoc tasks, the expert strategy can be used to elevate human decision-making. In this strategy, humans always have the final say. Expert systems deal with anything from large amounts of money to human lives, meaning humans must always be responsible for the consequences of the decisions made.
Expert AI systems can be used to empower professionals in all manner of industries. Doctors, lawyers, judges, politicians, military staff, financial advisors, and teachers are just a few examples of professions that AI can advise. AI solutions can help teachers create tests, evaluate students, determine the best method for helping individual students, suggest curriculum’s, and much more. Financial advisors can receive powerful insights to make monetary decisions. Doctors can receive assistance with diagnosing patients. The list goes on, but the common pattern is that humans always make the final decision.
Product development is an area that entices me a lot personally. Through human-AI-collaboration, companies and scientists have invented new varieties of whiskey, vaccines, perfumes, and spices. We are likely to see a lot more AI-invented products in the future.
The above examples are all expert professions, but note that this strategy’s name can be slightly misleading. In this context, the term expert doesn’t necessarily mean an expert in the traditional sense. For example, Hopper has made an AI-powered app based on the expert strategy that predicts the best time to book flight tickets. The app empowers consumers to buy a ticket when the AI predicts it to be at its lowest price. In this scenario, the consumer is the expert.
The expert strategy can make complex jobs more manageable and help transition workers from roles that have been automated. Some jobs that currently require three to five years of University studies may not require such long educations in the future, as augmentative AI makes complicated jobs more manageable.
The Innovation Strategy
Finally, the innovation strategy is the most advanced of all the AI strategies. The polar opposite of the efficiency strategy, this strategy revolves around augmenting humans to enable creativity.
Imagine a music composer using some software to create a new song. The composer adds a piano, a guitar, and some drums. Suddenly, the machine learning algorithm, which has learned the composer’s music style from observing them, suggests adding a certain bass. The composer listens to the bass presented by the AI and decides for themselves whether or not to include it in the song. The AI will then learn from the composer’s decision. Much like in the expert strategy, the innovation strategy gives humans complete freedom in decision-making.
Recruiters conducting interviews with job aspirants can be augmented by an AI that proposes follow-up questions in real-time. Meanwhile, writers could be augmented by an AI that not only advises them on their vocabulary and grammar but even notifies them when some topic needs to be explained more clearly.
When you’re replying to an e-mail with Gmail, the software might suggest a quick reply for you based on the e-mail and your style of writing. This is a basic example of the innovation strategy in action.
Which strategy should you adopt?
To understand what strategy to adopt for any given activity, we need to look at two variables: data complexity and work complexity. Based on the complexity contained in an activity, we can find out what AI strategy to use.

Activities with low data complexity are typically structured and simple: these activities usually consist of simple pieces of text or numbers. This data is easy to interpret for a computer. High data complexity, on the other hand, is often unstructured and up for interpretation. Images, videos, music, and voices are examples of complex data. While a machine learning algorithm may be able to tell whether the subject of the photo is a cat or a human, it’s up to subjective interpretation to determine whether the human looks tired, annoyed, upset, or are just resting their face.
Work complexity is all about determining whether or not an activity has clearly defined rules and routines. If an activity has rules and routines, it becomes predictable and has a low work complexity. However, if the activity is generally unpredictable and ad hoc, it requires decision-making skills, which would result in high work complexity. Note that the time it takes to complete an activity is, in this context, irrelevant to its complexity.
By analyzing the complexity of an activity, we can roughly determine the appropriate AI strategy:
- Low data and work complexity: The efficiency strategy.
- High data but low work complexity: The effectiveness strategy.
- Low data but high work complexity: The expert strategy.
- High data and work complexity: The innovation strategy.
Note that it’s not always this cut and dry. As mentioned previously, other factors can come into play when determining whether to automate an activity. Furthermore, what is considered to be complex is ever-changing. Data that is considered complicated today might be considered to be simple in the future. Furthermore, if an activity lacks clear routines and rules, maybe it’s possible to create them? Activities that we consider to be unpredictable today might be activities that we can streamline in the future.
In summary
AI can be used to automate and augment any activity.
- Automation is the removal of humans from an activity.
- Augmentation is the empowering of humans in an activity.
Every AI solution in the world can be placed within one or more of these four AI strategies:
- The efficiency strategy, in which activities are optimized through automation.
- The effectiveness strategy, in which activities are made seamless, enabling easier communication.
- The expert strategy, in which AI empowers decision-making.
- The innovation strategy, in which AI enables creativity.
In fully automated solutions, the AI makes decisions, but humans must be responsible for the AI’s choices. In augmented solutions, the AI does not make decisions autonomously. To find the optimal AI strategy for your activity, start by looking at the activity’s data and work complexity.
Implementing the right strategy for the right activity is crucial to succeeding in one’s AI journey.
Thanks for reading! If you’d like to go a little deeper, check out a case study I did on Domino’s Pizza and three of their AI strategies:
Also, you can find a more detailed presentation of the AI strategies, along with 100 real-world implementations of artificial intelligence, in my non-technical book on AI:
This Is Real AI: 100 Real-World Implementations of Artificial Intelligence