Introduction: Become An AI Company in 90 Days

Even if there were a trustworthy way to send money over the Internet– which there isn’t–the network is missing a most essential ingredient of capitalism: salespeople.

Clifford Stoll
“The Internet? Bah!” Newsweek 1995


Congratulations. You’re living through a once-in-a-generation technology shift—the era of artificial intelligence, or AI. Like previous fundamental shifts such as electricity, the computer, or the Internet, AI will change everything.

Skeptical about such big claims? I don’t blame you. Over the past few decades we’ve been bombarded by an endless parade of new technologies promising to “disrupt” everything and solve all our problems.

Most of these technologies never yield more than incremental benefits:

  • Object-oriented programming made software only slightly more reusable.
  • Social media moved us closer to only a fraction of our customers.
  • NoSQL databases solved some scalability problems and created many others.

Big data, fuzzy logic, mobile, ontologies, . . . year after year the buzzwords keep coming. And most of them merely come and go.

But how about electricity? The microprocessor? Personal computers? The Internet? Enthusiasts welcomed these technologies with lots of hype, and everyone underestimated their long-term impact. Today no large-scale business can survive without all of them. These fundamental technology shifts created value by:

  • Replacing or augmenting labor (industrial automation, appliances, robotics)
  • Helping us work more efficiently (spreadsheets, databases, enterprise software)
  • Helping us communicate more efficiently (phones, email, web sites)

AI, the current technology shift, creates value in an additional way:

by replacing or augmenting human thinking.

No doubt you’re already seeing examples of AI in consumer products. Facial recognition works almost flawlessly in the iPhone X. Siri and Alexa can interpret and answer simple questions, and voice-to-text on our smartphones improves daily.

But these narrow AI solutions are only the beginning. Soon business processes that involve thousands of interactions between people, and software will be replaced with AI models.


Let’s consider how AI might be applied to something that’s relatively easy to accomplish, like vacation planning.

Twenty years ago vacation planning was straightforward. A travel agent would ask a few questions and then present us with a handful of options. “Would you like to visit Paris, Rome, or Athens?” This simplicity came at a cost—it was relatively expensive and generic.

Today we have unlimited travel choices for nearly every budget.Unfortunately the volume of data and options creates a hassle. We have to make dozens of decisions:

  • When to go
  • Where to go
  • Where to stay
  • How much to spend
  • How to get there
  • What to do

Each decision impacts every other. We have nearly infinite good and bad options, and all of them can change by the minute as prices, weather, events, and schedules fluctuate. So we spend hours Googling, reading reviews, and looking at prices and weather patterns. We’ve exchanged simplicity for choice.

But an AI-powered vacation planner could give us both simplicity and choice. Such a system could ingest thousands of data sources:

  • Your social media pictures and activity
  • Your friends’ social media activity
  • Your email
  • Your calendar
  • The real-time pricing and schedule of every transportation resource
  • The availability and cost of every lodging option
  • Every local event
  • The weather forecast
AI models could process these disparate data sources and suggest vacation options. We can imagine a conversational interface built into Alexa:

Me: Alexa, can you suggest some vacation options in August?

Alexa: Sure, Kevin. I’m getting some options from the Intelligent Vacation Planner. What do you think of golfing in Costa Rica, wine tasting in southern France, or relaxing in Chengdu?

Me: China is too far. What are the best times for Costa Rica and France?

Alexa: Anytime in August is good for Costa Rica. France is too expensive until the 25th.

Me: What else can I do in Costa Rica?

Alexa: Costa Rica is known for sauvignon blanc, pinot noir, and Syrah wines. You could visit some wineries.

Me: Awesome. Send me three itineraries for a five-day vacation in Costa Rica. I’d like to play at least three rounds of golf and want to stay in the same hotel I had last time.

I’m thrilled to let the Intelligent Vacation Planner augment and replace my own thinking. Why should I have to comb through TripAdvisor reviews to pick a good golf course? Or use Kayak to find good flights? Airbnb to find appropriate lodging? Since there is no “best” vacation, I’d rather offload this intellectual hassle to AI so I can focus on doing client work—or playing golf.

Building the Intelligent Vacation Planner doesn’t require a scientific breakthrough or the visionary leadership of Elon Musk. The fundamental AI building blocks already exist. It’s only a matter of time before a company creates it. And by the end of this book you’ll know how to do it.


To understand how you might use AI in your own business contexts, let’s break down the business workflow of the Intelligent Vacation Planner:

  1. Ingests data from many different data sources.
  2. Analyzes complex data such as images and documents.
  3. Generates the best solutions from sometimes infinite choices.
  4. Provides the user a means to iterate options and make a final decision.

Many businesses have entire departments dedicated to performing similar workflow functions. Their processes can be exceedingly complex and time consuming. Consider the processes involved in some of our major industries:

  • How do financial services generate investment recommendations for their clients?
  • How do accountants process invoices?
  • How do hospitals diagnose problems?
  • How do banks review correspondence for compliance?
  • How do insurance companies process claims?
  • How do construction companies develop cost and schedule estimates?
  • How do governments process tax returns?

Here’s the answer: inefficiently.

I know because I’ve talked to Fortune 500 executives about applying AI to every one of these business processes. I have yet to encounter a single large company whose business won’t be dramatically impacted by AI. Your company is no different.


I try to help my clients understand AI by drawing analogies with the Internet, the most recent fundamental technology shift. I agree with Chris Benson—AI in 2018 is like the Internet in 1996. Let’s take a trip down memory lane and talk about the Internet’s beginnings.

The current state of enterprise AI is like the Internet in 1996.

— Chris Benson,
Chief Scientist of AI at Honeywell

I was a graduate student at Stanford in the mid-1990s, the dawn of the Internet. I had been using early applications like email, FTP, Listserv, AOL, and Gopher, but none of these were indispensible yet. Then in October 1994 a classmate showed me Netscape 0.9, the first widely adopted web browser.

Like thousands of other early Internet enthusiasts, I instantly realized the world was about to change: everyone would connect through this simple computer interface. Anyone would be able to instantly find answers to questions. Commerce would move to online catalogs. Business processes between companies would become transparent and automated. Computer games would be global. Best of all, utopian democracies would emerge when everyone had access to the same information on the Internet. (Yeah, got that one wrong, didn’t I? Ah, youth).

I knew a revolution was coming and I wanted to be part of it. I started telling everyone about the Internet—and quickly realized most people had no idea what I was talking about.

Most people asked me questions like:

  • So . . . what is the Internet?
  • Where is it?
  • CompuServe already does all this for $10/hour—who do you pay for the Internet?
  • What is the difference between the World Wide Web and the Internet?
  • What is the market for it? (First meeting with a Silicon Valley investor.)

Most early Internet enthusiasts had the same experience. A Wired reporter registered and couldn’t get McDonald’s corporation to take it for free.1 Of course everything changed on August 9, 1995, when investors valued Netscape at a $3B on its IPO.

Fundamental technology shifts

Why did it take several years for most people to recognize the Internet’s potential? Because fundamental technology shifts are initially abstract. It is hard to explain “everything will change” to someone who isn’t trying to understand what is happening.

Most new technology isn’t so abstract, and it meets known needs. NoSQL, for example, is a relatively new technology that promises to overcome known problems with relational databases—that’s why it isn’t a fundamental shift. NoSQL also has a natural home in your infrastructure or product team.

Fundamental shifts don’t necessarily solve known problems. They don’t fit neatly into existing departments. And they don’t replace technologies so much as they propel additional innovations. Fundamental shifts include electricity, computers, the Internet, and now AI.

That’s why you’re having a hard time identifying specific AI use cases for your company. The biggest opportunities for AI are not obvious. Most people give up when they don’t see easy answers to hard questions. “I just don’t see the value,” they say.

But not you. You’re reading this book because you’re willing to invest the time and energy to lead your company into the AI future. Ironically, planning for a big, broad impact like AI is actually easier than identifying specific use cases because you can start with a blank slate.

AI’s predictable path forward

Fundamental technology shifts don’t happen instantly. Researchers worked on electricity, heavier-than-air flight, the microprocessor, and the Internet for decades before the technologies became practical.

AI is no different. Neural networks—the algorithms which have enabled the recent AI breakthroughs—have been in the research-and-development shop for 70 years! They only recently became useful because the right enabling technology and market forces converged.

In the following table, consider the parallels between the adoption curves of the Internet and AI.

Of course trying to predict which specific technologies or companies will emerge as winners in this technology shift is impossible. Google wasn’t predictable, but the migration of advertising dollars to the web was clear from the outset. Any company with significant revenue from advertising could have begun this transition in 1996, even though the ultimate impact and timing of online advertising was unknown.

Today we’re in a similar situation with AI. Venture capitalists are pouring billions into AI initiatives. Most of these will fail.3 Most initiatives by your competitors will fail, and many of your own initial assumptions about AI will be completely wrong.

Yet there is little doubt that AI will radically change your business. You have entire departments which will transform to a workflow like that of the Intelligent Vacation Planner. You just don’t know when or how.

Regardless, you can begin making small bets and taking incremental steps to prepare for AI. You’re about to learn about those bets and steps.


So here we are. You’re trying to decide whether this book is worth your time. You’re skeptical because, well, most of what you read about AI is useless. Vendors are pitching you magical solutions you don’t understand. The blogs, podcasts, online courses, and books are too technical. The last AI conference you attended presented nothing but fluffy concepts which didn’t get you closer to your only real question: What can we actually do with AI?

Let’s decide whether this book is worth investing your most valuable asset: your time.

Why should you listen to me?

I caught the AI bug as a graduate student at Stanford in the mid-1990s doing neural network research under Dr. Bernard Widrow, a legend in computer science. I was fascinated by “computers that think” even though the technology at that time fell short of even modest expectations. I decided to pursue a career in industry and had the fortune to help drive innovation in many related projects on the road to AI. I’ve run massive data processing applications for anomaly detection (FINRA), helped launch the first version of Palantir into the US intelligence community (In-Q-Tel), and served as CMO for a venture-backed data science startup (MadKudu).

I’ve also got the battle scars of many failed startups, failed investments, and failed enterprise software initiatives—experiences which taught me far more than my successes.

So what, right? My career looks like that of everyone else who’s promising to help you achieve great things with AI. Smart guy, good schools, great
experience . . . blah, blah, blah.

Why should you listen to me over anyone else? Because I’m investing all of my energy into bringing AI to the enterprise—that is, I focus on solving only your problems. I’m not concerned with the latest AI fads, research papers, or hot new startups. I’m not trying to impress you with how much I know by using needlessly complex jargon. I care about helping you transform your enterprise with AI.

A few years ago I realized that AI had finally matured to the point where I could start using it to solve real problems. With high hopes and big dreams, I launched an AI startup with a colleague from Apple. Here’s a summary of most customer sales conversations: “Thanks Kevin, but your product sucks. But while you’re here, can you explain AI to us?”

As I engaged with these companies, I started to realize that the most interesting AI problems were those that involved bringing the technology to large enterprises which had the data and business-process complexity to take advantage of AI.

Coincidentally, my friend Russ Rands was running a data science startup and had come to a similar conclusion. Russ and I decided to join forces and founded Prolego (which means predict in Greek) with the mission of bringing AI to Fortune 500 enterprises. We’ve spent the last few years helping dozens of executives in industries like banking, insurance, defense, financial services, healthcare, transportation, hospitality, automotive, and real estate.

We started recognizing similar problems across these industries:

  • Confusion about basic concepts like machine learning.
  • Unfamiliarity with how to value data assets and estimate data gaps.
  • False assumptions about AI talent shortage and necessary skills.
  • No frameworks for generalizing fundamental technologies (e.g., recurrent neural networks) into solutions.
  • No frameworks for exposing the major risks and opportunities with AI projects.
  • No heuristics for deciding what to buy or build.
  • No means of choosing among many potential AI opportunities.
  • Uncertainty about the first steps.

Let’s talk about you

You’re not Google: you don’t have $1B to buy AI startups and invest in
everything that sparks your interest. You’re not a startup: you can’t stand up a
server tomorrow and start building whatever you want. You’re not a research
organization: you don’t have the resources to push the state of the art in AI.

You work in a big, complex organization that has evolved over decades. Your
technology team is still trying to finish the CRM and ERP initiatives started
five years ago. Your data? Nobody can tell you where it all resides, what is in
it, and who owns it. Getting anything done at your company is harder than
outsiders expect. And. It. Takes. Forever.

Sound familiar? Relax. I promise you’re not alone. Your competitors are in the
same position, their fancy demos and press releases notwithstanding. The
CEO of that hot AI startup you met is terrified you will realize he doesn’t have
anything magical you can’t build yourself.

AI is still very new, and you have time to develop your strategy and
systematically test ideas. You don’t have to recruit PhDs from Stanford and MIT,
pay outrageous prices to acquire AI startups, waste hours listening to vendors
pitch you magical AI solutions, or invest three months learning Python so you
can take Andrew Ng’s new online course. You don’t even need to hire me to
build your AI strategy, because I’m about to show you how to do it yourself.

This book shows you how to build an AI strategy and launch your first
product. You’ll find out how to generate AI ideas and systematically vet
them for business opportunities. You’ll learn how to get your organization
comfortable with experimentation and to balance competing near-term and
long-term opportunities. You’ll see how to set up your AI infrastructure, build
a team, and organize your data. You’ll know when and how to deploy your AI
solution and how to avoid common mistakes.

AI is once-in-generation opportunity—one which will create fortunes and
make careers. Why can’t it be yours?

Let’s Future Proof Your Business.