Wednesday, August 23, 2023

Estimating AI Direct Contribution to Revenue Will be Difficult

One recurrent problem with revenue forecasts for all new technologies expected to be widely embedded in most existing products is how to isolate the specific contributions made by the one new technology, compared to all the other processes, features, packaging, distribution and manufacturing changes that might also have occurred simultaneously. 


Consider some studies of revenue upside from artificial intelligence for a variety of use cases and products, including virtual assistants, self-driving vehicles, smart factories, health care, fraud prevention, marketing, education or content creation. The numbers are big.


Product

Estimated revenue upside

Study

Key assumptions

Virtual assistants

$30 billion

IDC (2022)

AI-powered virtual assistants are expected to be used in a wide range of applications, including customer service, healthcare, and education. The study assumes that AI virtual assistants will be adopted by 50% of households by 2025.

Self-driving cars

$8 trillion

McKinsey (2018)

Self-driving cars are expected to revolutionize transportation, leading to significant cost savings and productivity gains. The study assumes that self-driving cars will account for 10% of new car sales by 2030.

Smart factories

$1.5 trillion

PwC (2018)

AI-powered smart factories are expected to improve efficiency, productivity, and quality. The study assumes that AI will be adopted by 75% of global manufacturing companies by 2030.

Personalized medicine

$100 billion

Grand View Research (2022)

AI-powered personalized medicine is expected to lead to more effective and targeted treatments. The study assumes that AI will be used to develop new drugs and treatments for a variety of diseases.

Fraud detection

$300 billion

Juniper Research (2022)

AI-powered fraud detection systems are expected to help businesses reduce losses from fraud. The study assumes that AI will be used to detect fraud in a variety of industries, including banking, insurance, and retail.

Marketing automation

$15 billion

Gartner (2022)

AI-powered marketing automation systems are expected to help businesses improve their marketing campaigns. The study assumes that AI will be used to automate tasks such as lead generation and customer segmentation.

Customer service

$10 billion

IDC (2022)

AI-powered customer service chatbots are expected to reduce costs and improve customer satisfaction. The study assumes that AI will be used to answer customer questions and resolve issues more quickly and efficiently.

Education

$25 billion

Research and Markets (2022)

AI-powered educational tools are expected to personalize learning and improve student outcomes. The study assumes that AI will be used to develop new learning materials and assessments that are tailored to individual students.

Content creation

$10 billion

MarketsandMarkets (2022)

AI-powered content creation tools are expected to automate tasks and improve the quality of content. The study assumes that AI will be used to create content such as videos, images, and articles that are more engaging and informative.


Some of that revenue will be earned by suppliers of computing infrastructure necessary to support widespread AI operations; operators of cloud computing services or suppliers of software, training and advice to implement AI is specific business functions. That might be the most-clear and most-direct AI contributions to firm and industry revenue.


That will be easier to measure than revenue upside for all other industries that might use AI in some way. 


Obviously many of those products have existing customers, markets, business models and revenue. AI is expected to enable lower costs, add features to existing products and might also allow creation of some new products. 


In many cases--probably most--AI impact might be indirect. Often AI features will be embedded in products with a revenue impact that is measured by new account additions, product model or version upgrades, lower churn or some other indirect outcome. 


No comments:

Have LLMs Hit an Improvement Wall, or Not?

Some might argue it is way too early to worry about a slowdown in large language model performance improvement rates . But some already voic...