I recently finished reading the 2023 book “HBR Guide to AI Basics for Managers” by Harvard Business Review, and this post summarizes some of the most important ideas therein.
All in all, I found the book a bit fluffy, but there are some interesting ideas nevertheless.
Introduction
The book starts with some interesting platitudes observations:
78% of the surveyed managers believe that they will trust the advice of intelligent systems in making business decisions in the future.
Here’s the gist:
AI will ultimately prove to be cheaper, more efficient, and potentially more impartial in its actions than human beings.
Now we’re talking. Well, hallucinations and bias notwithstanding, of course.
Section One: AI Fundamentals
Chapter 1: Three Questions about AI that Every Employee Should be Able to Answer
What will humans do in a world with AI?
Problems that are novel, or which lack meaningful data to explain them, remain squarely in the realm of human specialities.
Phew.
A machine cannot understand the biases that data reveals, for example, nor the consequences of the advice it gives.
Tru dat.
Chapter 2: What Every Manager Should Know about Machine Learning
This chapter provides some simple explanations for technical terms. For example:
Regularization is a way to split the difference between a flexible model and a conservative model, and this is usually calculated by adding a penalty for complexity, which forces the model to stay simple.
Fair enough. What else?
The more stable the environment, the more useful a prediction model will be.
Yes, that’s right. And?
There is no guarantee the model will be useful if we move to a new environment.
Might seem obvious, but it’s not.
Chapter 3: The Three Types of AI
This chapter provides a taxonomy as simple as it is useful.
Robotic process automation (RPA) technologies automate digital and physical tasks–typically back-office administrative and financial activities.
(…)
RPA is the least expensive and easiest to implement of the cognitive technologies we’ll discuss here, and typically brings a quick and high return on investment.
Now to the gist of the matter:
If you can outsource a task, you can probably automate it.
Ouch. What another type of AI do we have?
Cognitive insight applications are typically used to improve performance on jobs only machines can do (…) so they’re not generally a threat to human jobs.
And the third type?
Projects that engage employees and customers using natural language processing chatbots, intelligent agents, and machine learning were the least common type in our study.
Where should businesses start then?
Before embarking on an AI initiative, companies must understand which technologies perform what types of tasks, and the strengths and limitations of each.
(…)
A main success factor is your people’s willingness to learn. Some will leap at the opportunity, while others will want to stick with tools they’re familiar with. Strive to have a high percentage of the former.
Chapter 4: AI Doesn’t Have to Be Too Complicated or Expensive for your Business
Who can use AI? Everyone.
But industries such as manufacturing, agriculture, and health care still need to find ways to make this technology work for them.
(…)
… executives in all industries should adopt a new, data-centric approach to building AI.
(…)
We’ve heard about the benefits of big data, but we now know that for many applications, it is more fruitful to focus on making sure we have good data–data that clearly illustrates the concepts we need the AI to learn.
But how far are they?
Accenture estimates that 80% to 85% of companies’ AI projects are in the proof-of-concept stage.
Section Two: Building your AI Team
Chapter 5: How AI Fits into your Data Science Team
Let’s assume you have invested in Data, what now?
Now, today, we talk about AI. The term itself (…) it’s essentially about using machine learning, and specifically deep learning.
There are some capabilities that are tantamount to success:
The first capability is understanding the business.
Duh.
The second capability is product data science.
And?
The last data capability is one that tends to get neglected or lumped in with product data science. It’s an R&D capability–using data to open up new product, new business, and new revenue opportunities.
Et voilà. But remember this:
Data science requires having that cultural space to experiment and work on things that might fail.
This is where Switzerland shines. I’m kidding.
Chapter 6: Ramp Up your Team’s Predictive Analytics Skills
Your AI projects might need help at the beginning, and that’s OK:
At least for your first pilot projects, you’ll need to bring in an external machine learning consultant for key parts of the process.
Chapter 7: Assembling your AI Operations Team
What happens once your AI solution is ready? Deployment. But beware!
Well, it’s hard–in fact, very hard–to integrate AI models into a company’s overall technology architecture.
(…)
Top-notch AI won’t do you any good if you can’t connect it to your existing systems.
You’ve been warned.
Section Three: Picking the Right Projects
Chapter 9: A Simple Tool to Start Making Decisions with the Help of AI
Recent developments in AI are about lowering the cost of prediction. AI makes prediction better, faster, and cheaper.
The book shows an “AI canvas” (figure 9-1) that I won’t reproduce here, but it’s gold.
Section Four: Working with AI
Chapter 11: Collaborative Intelligence: Humans and AI are Joining Forces
The goal is the augmented corporation, where AI increases the productivity and potential of everyone at every step:
Companies must understand how humans can most effectively augment machines, how machines can enhance what humans do best, and how to redesign business processes to support the partnership.
Chapter 12: How to Get Employees to Embrace AI
This is the $10.000.000 question.
If you give control over AI experiments to employees to keep them involved, and to allow them to see what the AI does well, you can leverage the best of both humans and machines.
No shit Sherlock.
Chapter 13: A Better Way to Onboard AI
We continue the touchy subject started in the previous chapter:
We advise managers and systems designers to involve employees in design: Engage them as experts to define the data that will be used and to determine ground truth; familiarize them with models during development; and provide training and interaction as those models are deployed.
Chapter 14: Managing AI Decision-Making Tools
Once the AI is running, you need to manage it (and the surrounding humans), and for that we have frameworks, each with its mandatory acronym:
The four main management models we developed vary based on the level and nature of the human intervention: We call them “human in the loop” (HITL), “human in the look for exceptions” (HITFLE), “human on the loop” (HOTL), and “human out of the loop” (HOOTL).
Chapter 15: Your Company’s Algorithms Will Go Wrong. Have a Plan In Place.
Murphy’s Law strikes back:
Analyzing the list of AI failures above, we can arrive at a simple generalization: An AI designed to do X will eventually fail to do X.
🎉
Section Five: Managing Ethics and Bias
Chapter 16: A Practical Guide to Ethical AI
Handling hallucination and biases is hard work:
Today the biggest tech companies in the world are putting together fast-growing teams to tackle the ethical problems that arise from the widespread collection, analysis, and use of massive troves of data, particularly when that data is used to train machine learning models, aka AI.
Your small business should be no exception.
Companies need a plan for mitigating risk–how to use data and develop AI products without falling into ethical pitfalls along the way.
How can you do this?
Leaders should take inspiration from health care, an industry that has been systematically focused on ethical risk mitigation since at least the 1970s.
Can you name a risk you have to manage?
Take, for example, the oft-lauded value of explainability in AI, a highly valued feature of ML models that will likely be part of your framework.
Or rather, its lack thereof.
Chapter 18: Take Action to Mitigate Ethical Risks
Let’s be real:
Having productive conversations about what AI ethical risk-management goals are achievable requires keeping an eye on what is technologically feasible for your organization.
Section Six: Taking the next Steps with AI and Machine Learning
Chapter 19: How No-Code Platforms Can Bring AI to Small and Midsize Businesses
You can start small with no-code platforms, and achieve interesting returns on investment:
Work with the data you already have. There is often more value to be captured there than you may initially think.
(…)
Get quick wins in common areas, such as sales funnel optimization or churn reduction, so your team can learn how AI applies to a wide range of use cases.
Chapter 20: The Power of Natural Language Processing
What about the GPTs, LLMs, and GenAIs of the world?
Language-based AI won’t replace jobs, but it will automate many tasks, even for decision-makers.
Disruption is coming, also for managers:
Remember that while current AI might not be poised to replace managers, managers who understand AI are poised to replace managers who don’t.
Chapter 21: Reinforcement Learning is Ready for Business
What is reinforcement learning?
Game-playing systems like AlphaGo use reinforcement learning–a mature machine learning technology that’s good at optimizing tasks.
What is it good for?
In fact, any time you have to make decisions in sequence–what AI practitioners call sequential decision tasks–there is a chance to deploy reinforcement learning.
How does it work?
Reinforcement learning systems produce actions, not predictions–they’ll suggest the action most likely to maximize (or minimize) a metric.
When should you use it?
Reinforcement learning is helpful when you lack sufficient historical data to train an algorithm.
(…)
The technology shines when used to automate or optimize business processes that generate dense data.
Epilogue: Scaling AI
Chapter 22: How to Scale AI in your Organization
The time of MLOps is here.
MLOps seeks to establish best practices and tools to facilitate rapid, safe, and efficient development and operationalization of AI.
(…)
Implementing MLOps requires investing time and resources in three key areas: processes, people, and tools.
The holy trinity of management.
And please, don’t reinvent the wheel.
On the production side, model services must work with DevOps tools already approved by IT.
(…)
Of these, data science and IT tend to have opposing needs.
And thus tension arises.