KPMG's network of data & analytics professionals solve complex business challenges using analytics learn more

Trusting the Machine

How artificial intelligence can optimize data and improve your company

Companies are collecting and creating more data than ever before

double line rule
2,500,000,000,000,000,000

There are 2.5 quintillion bytes of data created every day.

44x Data production will be 44 times greater in 2020 than it was in 2009. Data production will be 44 times greater in 2020 than it was in 2009. 2009 2020
More data has been created in the last two years than in the enire previous history of the human race. More data has been created in the last two years than in the enire previous history of the human race. 2 YEARS PREVIOUS HISTORY OF THE HUMAN RACE
25 BILLION By 2020, there will be more than 25 billion connected devices transmitting data. 25 BILLION By 2020, there will be more than 25 billion connected devices transmitting data
1,500 PIECES OF INFORMATION. Data collection on one person can include an averge of Facebook collects 500 TERABYTES of data daily.

And they are using that data to drive decisions

A recent study quantified how much companies lean on data to make decisions regarding:

double line rule
  1. How products are used
  2. When to develop new products
  3. How to understand fraud
  4. How to manage risk
  5. How to be cost effective
70 %

AI can automate those data-driven decisions, but the GIGO (Garbage In, Garbage Out) effect looms large

double line rule
216,000 The average adult can read 216,000 words per day. The average adult can read 216,000 words per day.
84% of CEOs are concerned about the quality of the data on which they're basing decisions. 84% of CEOs are concerned about the quality of the data on which they're basing decisions. 84 %
$3,000,000,000,000
Bad data costs the U.S. $3 trillion every year.
10-25% The business cost of bad data may be as high as 10-25% of an organization's revenue. The business cost of bad data may be as high as 10-25% of an organization's revenue.
Multiple state revenue departments routinely report capturing millions —sometimes even billions —of dollars in revenue by using data and analytics technologies to detect and prevent fraud. Multiple state revenue departments routinely report capturing millions— sometimes even billions—of dollars in revenue by using data and analytics technologies to detect and prevent fraud.

Putting humans in the loop ensures AI systems learn signal and not noise

double line rule

  1. Develop AI (programs or algorithms) that reads data more quickly and efficiently than humans.

    • Expand your company's spectrum of human cognition and capabilities.

  2. Allow people to assess the task work done by AI and determine areas of weakness and insufficiency as well as opportunities for improvement.

    • Dedicate resources to create a “learning loop” for humans.
    • Understand that you can't buy the answer, you must build it.

  3. People adjust programs and algorithms to allow for the most cost-effective and insightful analysis.

    • Allow machines to learn from humans.
    • Unlock the value of AI to accelerate intelligent automation and cost management, growth and customer engagement, and risk and regulatory compliance.

  4. After fine-tuning is complete, it's time to expand. People give machines larger volumes of data to work faster and more completely.

    • Pair technology with ability to develop and tweak algorithms.

  5. People monitor the output and allow AI time to learn from the volumes of data.

    • Aim to automate decisions that drive growth and profitability.
    • But always value humans as the key decision makers.

  6. AI outputs insights for clients faster, smarter and more cost effectively.

    • Improve business efficiencies and increase productivity
    • Reduce costs, increase profitability and create faster innovation cycles