Å·±¦ÓéÀÖ

Jump to ratings and reviews
Rate this book

Doing AI: A Business-Centric Examination of AI Culture, Goals, and Values

Rate this book
Artificial intelligence (AI) has captured our imaginations—and become a distraction. Too many leaders embrace the oversized narratives of artificial minds outpacing human intelligence and lose sight of the original problems they were meant to solve.

When businesses try to “do AI,� they place an abstract solution before problems and customers without fully considering whether it is wise, whether the hype is true, or how AI will impact their organization in the long term. Often absent is sound reasoning for why they should go down this path in the first place.

Doing AI explores AI for what it actually is—and what it is not� and the problems it can truly solve. In these pages, author Richard Heimann unravels the tricky relationship between problems and high-tech solutions, exploring the pitfalls in solution-centric thinking and explaining how businesses should rethink AI in a way that aligns with their cultures, goals, and values.

As the Chief AI Officer at Cybraics Inc., Richard Heimann knows from experience that AI-specific strategies are often bad for business. Doing AI is his comprehensive guide that will help readers understand AI, avoid common pitfalls, and identify beneficial applications for their companies.

This book is a must-read for anyone looking for clarity and practical guidance for identifying problems and effectively solving them, rather than getting sidetracked by a shiny new “solution� that doesn’t solve anything.

272 pages, Hardcover

Published December 14, 2021

1 person is currently reading
39 people want to read

About the author

Richard Heimann

10Ìýbooks1Ìýfollower

Ratings & Reviews

What do you think?
Rate this book

Friends & Following

Create a free account to discover what your friends think of this book!

Community Reviews

5 stars
0 (0%)
4 stars
7 (35%)
3 stars
10 (50%)
2 stars
3 (15%)
1 star
0 (0%)
Displaying 1 - 5 of 5 reviews
Profile Image for Giulio Ciacchini.
351 reviews10 followers
March 2, 2024
More than a book on Artificial Intelligence, this is an examination on how to formulate problems and search for solutions.
I strongly agree with the main thesis, that is first define the problem and then think about (potentially AI driven?) solutions.
His message is perfectly summarised in these few lines
Being Al-first means doing Al last. Doing Al means doing it last or not doing it at all. The reason is rather simple: Solution-focused strategies are more complex than problem-focused strategies; and solution-focused thinking ignores the most important part of business, which is the problems they solve and the customers they create.
Keep in mind that solution-centric thinking results from the following:
Focusing on what our solutions ought to be rather than what they are.
Focusing on the impact of future solutions rather than the future impact of today's solutions.
Conflating our goals with the goals of others.
Focusing too much on abstract problems with some arbitrary solution or focusing too much on someone else's problem and ignore your problems. The former is solution solving and the latter often means we are working problem solving backward and finding problems to solve in the context of someone else's solution.
Do not define your solution. The search for analytical exactitude in verbal definition will not lead to economic progress. Ignore recycling glib, textbook definitions of artificial intelligence mainly because consumers don't care about textbook definitions. Customers care about themselves. If you want to make your life better, make their lives better. Help them accomplish their goals in a better, faster, safer, or cheaper way. They're generally interested in a value proposition that contains problem-specific information, not in a definition of intelligence. Your journey starts with more comprehension of problems, not the names or definitions of solutions. Besides, creating definitions for our solutions means we are creating external goals for them, which is nonsense.


He then highlights the differences between doing research in the AI field and work with AI: in the latter case we must be pragmatic and problem oriented.
Working in a company means that your manager does not care if the solution in called AI or BI, but whether you solve the problem in the first place.
Remember that insiders seek epistemological discoveries, not economic ones. The more epistemological a pursuit is, the less likely it is to become something that could be turned into a business. Entrepreneur, venture capitalist, and author Paul Graham discusses the value of problems at length and explains that good business ideas are unlikely to come from scholars who write and defend dissertations.
The reason is that the subset of ideas that count as "research" is so narrow that it's unlikely to satisfy academic constraints and also satisfy the orthogonal constraints of business. The incentives for success in the academic world are not consistent with what it takes to start and grow a business. Ultimately, business pursuits are much more complicated than academic ones. Managers ought to acknowledge that solving intelligence is not likely your goal, and in many ways it's oppressive to problem solving. AGI may be possible, but it is not desirable as a business goal.


However, he soon starts to repeat those concepts over and over again, making most of the text redundant.
He then takes a (boring) philosophical tangent on what's a problem and its different types, which weights down the narrative flow.
Just to give you an intuition here is an extract on this topic
There is something magical about writing down a problem. It's almost as though by writing about what is wrong, we start to discover new ways of making it right. Writing things down will also remind oneself and our teams of the problem and the goal. Once a problem is written down, don't forget to come back to the problem statement. It is a guide. Problem solving often starts with great intentions and alignment, but when it counts most-when the work is actually being done-we often don't hold on to the problem we set out to solve, and that's the most important part of problem solving: what the problem is and why we are solving it to begin with.
Furthermore, do not needlessly seek out complexity by making larger solutions to solve needlessly bigger problems. Complexity bias is the logical fallacy where we find it easier to seek out complex solutions rather than a simple one. Without a problem statement, solutions tend to become more complex and expand to fill in the available time we've allocated for problem solving. Parkinson's law, named after Cyril Northcote Parkinson, states that "work expands so as to fill the time available for its completion." This is a sort of solution sprawl, similar to the urban sprawl that expands to fill in geographic spaces immaterial to how well the urban landscape serves it citizenry.


At last he gives good advice on general problem solving, particularly regarding Divide and Impera, which comes from the fact that small problems are often simpler problems
Always start small and take small steps to ensure that performance is what you want. Don't try to boil the ocean with the whole of a problem. With smaller steps almost everything can be reduced to something more manageable. Working in smaller sizes and smaller steps goes for your team as well. Rather than having your whole team work on something for six months, think about what one person can do in six weeks. The Basecamp team uses six weeks, which I think is a good size. If you are an Agile team, you may have batches of two weeks.75 That is fine, too. The point is that constraining batch size will force everyone to find the best bad solution, rather than working into the abyss of perfection.
Of course, simple problems are different. Simple problems can often be solved by applying a single solution to the whole of the problem. In practice you may not know the best solution a priori. One strategy to find the best solution for a simple problem may be to simply guess. Guessing, however, will have a high error rate in the face of increasing complexity.
Profile Image for Said.
188 reviews1 follower
March 31, 2022
A bit repetitive but raises fair concerns about the current approach to AI

First of all, while I do think it's worth a quick read, this book could easily be an article. That said, I think the perspective in connection with focusing on problems rather than solutions and not being obsessed with using AI for the sake of using AI (which seems to be a huge hype today). It was also not filled with technical language that many other AI books struggle with. I think the book would be most useful not to a random person who has a general interest in AI but more to business leaders who are considering either introducing AI to their companies or who are already working with AI as part of their job/business.
Profile Image for Noelle N..
86 reviews2 followers
June 16, 2024
Important book for those getting swept away in nebulous asks such as, “how do we incorporate AI?�

I liked the clear construct the author uses by labeling AI Insiders (those working to build AI) versus outsiders (businesses with real problems to solve).

Definitely worth reading.
Profile Image for Debashish Mishra.
44 reviews1 follower
July 31, 2022
A business centric introduction to AI. Going through the book won't disappoint you.
40 reviews
May 27, 2024
Not sure who the audience is supposed to be. Not very well organized. Constantly repeating the notion that you should start with the problem and not the solution. Some good ideas and useful citations and quotes.
Displaying 1 - 5 of 5 reviews

Can't find what you're looking for?

Get help and learn more about the design.