Professional Documents
Culture Documents
Exec Guide Gen Ai
Exec Guide Gen Ai
executive’s
guide to
generative AI.
Table of Introduction
About this guide.......................................................................................................................................................04
contents Chapter 1
A quick primer on generative AI......................................................................................................................... 05
Core capabilities and applications........................................................................................................................ 09
Chapter 2
The step-by-step guide to getting started with generative AI................................................................ 10
How to launch your first use case in 30 days...................................................................................................... 12
KPIs to measure the impact of generative AI......................................................................................................26
Chapter 3
The value of generative AI in every industry.................................................................................................. 28
See what your peers are doing............................................................................................................................... 31
Retail and CPG............................................................................................................................................................32
Financial services.......................................................................................................................................................35
Healthcare and life sciences....................................................................................................................................37
Media and entertainment.........................................................................................................................................39
Manufacturing............................................................................................................................................................ 42
Communications service providers...................................................................................................................... 44
Conclusion
Innovate faster with generative AI for business........................................................................................... 45
Introduction
Generative AI marks one of the most significant technological shifts in history.
Its impact on individual and business productivity can Even better, a single generative AI platform can deliver What can change, though, is how you use this
be significant, with the potential to rival the advent solutions for multiple use cases, creating a network technology to enable your teams to improve core
of the internet or the mobile device. Indeed, among effect. As the number of users and applications offerings, and how you solve fundamental problems
organizations considering or using AI, 82% believe it will increases, the model is exposed to more data and that get in the way of delivering them. Indeed, with
either significantly change or transform their industry.1 becomes increasingly accurate and useful — which in the right tools, you could even identify and deliver new
turn encourages more users. points of difference.
What makes generative AI different from other forms
of AI that have come before is its ease of use in Organizations that use generative AI to speed up,
helping solve everyday problems in people’s personal automate, scale, and improve business processes stand
or professional lives. Anyone who knows how to to reap big benefits. According to McKinsey & Company,
ask a question of a search engine can use everyday generative AI’s impact on productivity could add
language to interact with a generative AI chatbot or between $2.6 trillion and $4.4 trillion USD annually to the
virtual agent — getting it to answer questions, create global economy.2
content, produce images, summarize documents, and
No technology ever takes away the fundamentals of
much more.
your value proposition and the core value chain in
your industry. At the end of the day in healthcare, for
example, you’re still trying to improve patient care.
4
Chapter 01
A quick
primer on
generative AI
5
Chapter 01 | A quick primer on generative AI
Philip Moyer
Global VP, AI & Business Solutions, Google Cloud
6
Chapter 01 | A quick primer on generative AI
Every day, people in your From executives digging into strategic trends to
salespeople creating product demos or new employees
business spend time and with benefits questions — everyone in your organization
energy digging for information can relate to this frustration. But that’s changing.
to make decisions, serve Imagine giving each person in your company not only
a personal assistant, but an expert in every piece of
customers, and move your data relevant to their job and, indeed, potentially every
business forward. Informed piece of data across your whole organization. With an
assistant like that, impatient moments of indecision
decisions require information, would dwindle. Everyone would be empowered to
and collecting the right inputs spend less time waiting and more time doing.
can take time. With generative AI, this is achievable. And it’s just
one example of the many disruptions this technology
Suppose you need to know how ad spend impacts
has unlocked. Always-on coding collaborators.
customer perceptions. Or you want to see the trends
Brainstorming assistants to draft and iterate content.
in a competitor’s patent filings, R&D investments,
Personalized self-education on any subject. Human-
and technology acquisitions. Information like this
like interactions with customers wherever they need
exists somewhere in your organization, often across
you, for whatever reason. Generative AI is all these
a bunch of places. You need it to decide the next step
things and more. In time, it will affect almost every
— and, to obtain it, experts need to be convened,
aspect of every business.
research completed, and the information compiled
and synthesized. If you have a follow-up question, the
whole process might have to start again.
7
Chapter 01 | A quick primer on generative AI
Foundation models are the content generation — with little or no additional data or
tuning. Yet they can also be expensive to train and run,
engine powering generative AI. prone to inaccurate outputs, and difficult to work with.
Generative AI applications are powered by foundation For these reasons, generative applications aren’t
models, which are trained on vast amounts of content. reducible to generative models. Your intelligent
For example, Large Language Models (LLMs) are applications will need to mix probabilistic
one type of foundation model, trained on text or foundation models with traditional, deterministic
language. Other multi-modal types can be trained on (in other words, constrained) programming.
images/photos, video, music, software code, medical Deterministic models are constrained to what
information, or cyber security data. But access to can flow out of them, and are limited by endless
models alone won’t position your business for success. options that have to be pre-established.
Foundation models are best thought of as probability Traditional AI is purpose-built for the task at hand. It is
engines that can be nudged and shaped by human all about optimizing and tweaking existing processes
input. Because they are probabilistic, they are such as forecasting specific patterns that have been
fundamentally different from traditional software pre-determined by humans. It’s why traditional AI can be
paradigms. When today’s apps need to look up product used to automate discrete, standardized processes in
prices or validate customer information, they use specific areas of the business, such as customer service.
deterministic functions that call a database. In contrast, In comparison, generative AI models have emergent
foundation models use the patterns they learn during capabilities where they can perform multiple tasks,
training and tuning to calculate the most probable even if they were not explicitly trained for them during
output, such as the most likely answer to a question or the instruction tuning phase. It is this ability to multitask,
an accurate caption for an image. plus the freedom afforded by the prompting interface,
Since they’re not confined by rows and columns in a that enables these models to perform in a wide variety
database, foundation models are extremely powerful. of use cases.
They’re often capable of performing many downstream
tasks — such as Q&A, summarization, or open-ended
8
Chapter 01 | A quick primer on generative AI
Creation
rapid popularity and adoption through simple chat speech, and code has enormous potential.
interfaces. Chat is a natural way to interact with Whether it’s speeding up processes or helping
powerful generative AI models. You can use it to employees turn ideas into output faster,
Summarization improve customer interactions, enhance product
capabilities, train employees, and more.
generative capabilities can be deployed into
products, tools, and workflows.
The step-by-step
guide to getting
started
10
Chapter 02 | The step-by-step guide to getting started with generative AI
11
Chapter 02 | The step-by-step guide to getting started with generative AI
12
Chapter 02 | The step-by-step guide to getting started with generative AI
Step 1 of 10
persona Such roles are often repetitive and offer little career
advancement. Automating these tasks can free up
employees to focus on more strategic work.
Many industries must meet strict requirements to be
compliant. In life sciences, for example, every claim
about the efficacy of a medicine must be reviewed by
an attorney to ensure the language is compliant. An
Determine which job Find opportunities to attorney must then review the claim to ensure certain
clauses are included in every instance of the fine print.
category or function automate repetitive, tedious This can be time consuming and repetitive, but is an
within the chosen domain tasks that are necessary to essential part of the process for making medicine
you want to make more generate revenue. publicly available. Generative AI can help automate
critical tasks that help organizations improve accuracy
productive. For example, the multi-trillion dollar pre-authorization and reduce risk.
industry in healthcare is incredibly frustrating for
patients. It can take hours or even days to get
authorization for a simple MRI or specialist visit,
and the process is often paper-based. Investment
memorandums are another example in which the
same information must be gathered over and over
again. Generative AI can help automate these tasks
so employees can focus on more strategic work.
14
Chapter 02 | The step-by-step guide to getting started with generative AI
Step 3 of 10 Your generative AI model will be trained on the data you gather.
This data should be honed for the specific business or domain-
Determine level problem it is trying to solve, and accessible through
sources the
For example, if you select a marketing manager as This individual job role requires several data
your first persona, it’s important to understand the sources, including:
persona
specifics of their job role. Say they are responsible • A word processing tool like Google Docs to write,
for creating digital campaign content like ebooks edit, and collaboratively review the copy
and reports. These assets are promoted and used to
needs to be
• A design tool to lay out and format the final copy
capture leads via online forms. If the individual fills
• A web platform to publish the content online
out the form and opts in to receive outreach about
15
Chapter 02 | The step-by-step guide to getting started with generative AI
Mitigate hallucinations
AI models are trained to give users what they want, which means they
occasionally spit out made-up answers that sound convincing and can be hard
to spot. To avoid this problem, you can ground responses in specific data rather
than relying on the LLM alone.
16
Chapter 02 | The step-by-step guide to getting started with generative AI
Step 4 of 10
person
workflows, challenges, and needs of the day-to-day
the business tasks executed by the chosen persona.
17
Chapter 02 | The step-by-step guide to getting started with generative AI
Step 5 of 10
objectives,
efficiency
cases and provide
oversight. 57% Improved customer
and the
experience
49%
output you
Keep in mind, the value of a Accelerated innovation
generative AI project can come
from a number of sources. There’s
48% Improved employee
are trying to
productivity
direct business value, incremental
value of generative AI over legacy
39% Developed new
Step 6 of 10
tiger team
prompt samples to move quickly in this step.
19
Chapter 02 | The step-by-step guide to getting started with generative AI
(UX) and user Begin with a selection screen which allows users to choose the personality
to apply to the generated text, images, or output. For example, options
Create a logical and intuitive user flow that guides users through the AI
Create a user-friendly experience
model’s functionality. Ensure that the interface design aligns with the
and interface that will run the
expected user journey.
generative AI model in production
for the chosen persona’s use case. Consider how the new interface fits within the larger ecosystem of
existing apps, like email and chat, that might have their own built-in
generative AI capabilities.
Ensure the UI/UX is responsive and accessible across different devices and
screen sizes, including mobile phones, tablets, and desktops.
20
Chapter 02 | The step-by-step guide to getting started with generative AI
Step 8 of 10
Step 9 of 10
plan
Do we have cost controls during evaluation and experimentation?
22
Chapter 02 | The step-by-step guide to getting started with generative AI
23
Chapter 02 | The step-by-step guide to getting started with generative AI
Step 10 of 10
Expand
usage to In the beginning of this chapter, we explained how
an organization could start with one generative AI
use case and naturally expand into three use cases
use cases
question; second, it summarized the frequently
asked questions agents received over the phone;
and third, it produced answers that could be posted
same domain
itself became more accurate in the domain.
24
Chapter 02 | The step-by-step guide to getting started with generative AI
Looking ahead
at days 60-90
Once you are ready
to extend your use
case to external
01
Host a hackathon
02
Bring in partners
03
Create a center of excellence
users and/or third-
Harness enthusiasm across your In addition to sharing their expertise, Excitement about new technologies
party data, use these teams by hosting a hackathon, which partners can consult on business value can often lead to widespread use. A
methods and tactics encourages employees to brainstorm and technical implementations, provide center of excellence in models, tuning,
ideas and get hands-on with AI — all training, and even work side-by-side with and application integration can help to
to scale quickly and within a matter of days. your teams to transfer knowledge as standardize processes, share knowledge,
safely: they build out your implementation. and ultimately drive innovation.
25
Chapter 02 | The step-by-step guide to getting started with generative AI
generative AI
anticipated business value, and ultimate
return on investment of each generative
AI project.
Like any technology investment, you need to prove its worth.
Embed ROI measures into every use case and project, and establish
KPIs to keep a pulse on progress along the way.
26
Chapter 02 | The step-by-step guide to getting started with generative AI
Accuracy Productivity Customer satisfaction Cost savings Turnaround time Quality of output
Measure the accuracy of Assess the impact of If the generative AI use case Measure the cost savings Evaluate the time taken for Assess the quality of the
the generative AI model generative AI on the involves customer-facing achieved through the use the generative AI model generative AI outputs against
in producing relevant and productivity of the target applications, use customer of generative AI. This may to generate responses predefined criteria. This can
correct outputs. This can persona or department. This satisfaction surveys or involve comparing the costs or outputs compared to be done through manual
be quantified using metrics could include metrics like the feedback to gauge how of employing the AI system traditional methods. Faster review or automated quality
such as precision, recall, F1 number of tasks completed well the AI system meets to the expenses associated turnaround times can lead checks, depending on the
score, or mean squared error, per unit of time, response customer needs and with traditional manual to increased efficiency use case.
depending on the nature of time, or reduction in manual expectations. processes or outsourcing. and improved customer
the use case. effort required. experience.
Error rate Business impact Training time and Human-in-the-loop Scalability Regulatory compliance
Quantify the rate at which Identify specific business
cost metrics Assess how well the For sensitive domains like
the generative AI model metrics that are directly Measure the time and If human intervention is generative AI model scales healthcare or finance,
produces incorrect or impacted by the generative resources required to train involved in the generative to accommodate increased monitor how well the
undesirable outputs. AI use case, such as and fine-tune the generative AI process, track metrics usage or higher demands. generative AI system adheres
Minimizing error rates is increased sales, reduced AI model. Efficient training related to the efficiency Scalability is essential for to relevant regulatory
crucial for maintaining customer complaints, processes can lead to faster and effectiveness of human long-term success. requirements and data
accuracy and reliability. or improved employee implementation and quicker oversight. privacy standards.
retention. time-to-value.
27
Chapter 03
The value of
generative AI in
every industry
28
Chapter 03 | The value of generative AI in every industry
29
Chapter 03 | The value of generative AI in every industry
30
Chapter 03 | The value of generative AI in every industry
31
Chapter 03 | The value of generative AI in every industry
Retail
and CPG
82% Priority use cases
Insight
Todd Penegor
President and CEO, The Wendy’s Company
34
Chapter 03 | The value of generative AI in every industry
Financial
services
79% Priority use cases
Insight
35
Real-world example
Deutsche Bank accelerates
financial decision-making
Deutsche Bank is testing Google’s generative AI and LLMs at scale to provide
new insights to financial analysts, driving operational efficiencies and execution
velocity. There is an opportunity to significantly reduce the time it takes
to perform banking operations and financial analysts’ tasks, empowering
employees by increasing their productivity.
36
Chapter 03 | The value of generative AI in every industry
Healthcare and
life sciences
75% Priority use cases
Insight
37
Chapter 03 | The value of generative AI in every industry
Real-world example
US national medical center
transforms healthcare with
generative AI
One of the leading national medical However, this data is scattered across
centers and hospitals is transforming multiple formats and locations, making
healthcare with generative AI. The it challenging for clinicians to find the
organization started by using Vertex AI necessary information promptly.
Search and Conversation to improve the
Vertex AI Search and Conversation
efficiency of clinical workflows, helping
unifies data across dispersed
clinicians and researchers to find the
documents, databases, and
information they need, and ultimately
intranets, making it easier to search,
improve patient outcomes.
analyze, and identify the most
Healthcare professionals often rely on relevant results.
information from various sources such
as medical records, research papers,
and clinical guidelines to aid them Learn more about how
in diagnosing and treating patients. generative AI can improve
healthcare patient service.
38
Chapter 03 | The value of generative AI in every industry
Media and
entertainment
87% Priority use cases
Insight
Real-world example
TIME wants to build
community, not just
creativity, with LLMs
With trusted sources and chat “As publishers, what we’ve done for
resources, TIME wants to do more a hundred years, it’s been a one-
than deliver headline news — playing a way street: we put out content for
bigger role as a beacon for accuracy. the consumer and they consume
As media companies explore the it. With generative AI prompts
possibilities of generative AI, the and chat, we actually become
publisher sees an opportunity to able to understand, and have an
strengthen its role as a trusted source interaction with, the consumer
and community builder. — creating experiences that are
two-way in many ways. That’s
For years, it’s been using AI-powered why I actually see generative AI
recommendations to build affinity as a powerful tool for building
and loyalty with readers. Now, with community.”
generative AI, TIME hopes to turn a
one-way conversation into a dialogue. Burhan Hamid
Senior VP Data, Product,
Read the full story and Engineering, TIME
40
Chapter 03 | The value of generative AI in every
Introduction
industry
Real-world example “There is a quote that I love from the science fiction
writer Arthur C. Clarke, ‘Any sufficiently advanced
Canva is solving for technology is indistinguishable from magic.’ Canva
AI-powered design, has always been about removing as much friction
from the design process as possible and AI technology
for everyone allows us to make the design process even easier with
even less friction.
Canva is using the latest AI technology to empower their
customers and make the design process as frictionless as
We are incredibly excited to be working with Google
possible. From enabling users to translate their designs into Cloud as we test and explore ways to bring even more
over 100 languages with just a few clicks, to turning short magic to our community. Earlier this year we launched
videos into longer and more compelling clips with Google Magic Translate and what that enables you to do with
PaLM technology, they are unlocking the magic of AI with just a couple of clicks turn any of your designs into
Google Cloud.
over a hundred languages. Magic video is a time-saver
for marketers and teachers who want to be able to
Watch the full video cater to a diverse audience.”
Melanie Perkins
Co-founder & CEO, Canva
41
Chapter 03 | The value of generative AI in every industry
Manufacturing
monitoring to be Provide an easy, informative, and value-added Efficiently match requirements to the specifications
customer service experience that automates of products purchased.
valuable9 and accelerates time-to-resolution for common
Supply chain advisor
interactions.
Optimize fulfillment by recommending best-suited
suppliers based on relevant criteria.
42
Chapter 03 | The value of generative AI in every industry
Real-world example
Leading global airline supplier, GA Telesis,
integrates generative AI technology
As a major supplier of essential equipment in the airline to help businesses tune and deploy machine learning
industry, where long-term relationships and trust are the models, to help it quickly build innovative AI applications.
bedrock of many business transactions, GA Telesis’ sales Leveraging a new data extraction solution the GA Telesis
staff receive inquiries from global customers requesting technology team built internally, GA Telesis will be able
quotes for all sorts of commercial aircraft and jet engine to automatically synthesize purchase orders and quickly
replacement parts. provide customers a quote, eliminating the need for
sales teams to manually cross-reference emails with
The typical inquiry is not standardized, requiring
their inventory availability.
sales representatives to quickly decipher the
relevant aircraft or jet engine model, applicable
Read the full story
codes, quantity required, preferred condition and
provenance, and often most importantly, where
the part is needed and when. Additionally, in order
for airlines to meet their on-time performance
metrics, inquiries are often urgent and logistics have “In aerospace, GA Telesis will deploy Google Cloud’s generative AI
to be factored into the equation. GA Telesis’ team technology to revolutionize the sales and service processes for the parts
is expected to accomplish what can resemble an
the company supplies to major global passenger and cargo carriers.”
impossible feat in minutes, not hours.
43
Chapter 03 | The value of generative AI in every industry
Communications
service providers
Priority use cases
Customer or employee service automation Employee knowledge search
Make online customer service more conversational with Make it easier and more effective for staff to
human-like support and search. get things done, with human-like bots providing
IT support, self-service, and T2/T3 guidance to
Network planning and operations
field techs.
Easily access and understand complex data on network
Test or code script generation
performance, faults, inventory, infrastructure, and
anomaly detection. Generate and test experiments using real work
experiences.
Advertising and creative content assistance
Contract analysis and negotiation
Generate interactive and relevant content with highly
personalized messaging. Automate contract negotiations with suppliers by
analyzing bills, trends, and other supply data.
44
Conclusion
Innovate faster
with generative
AI for business
45
Conclusion | Innovate faster with generative AI for business
46
Conclusion | Innovate faster with generative AI for business
47
Conclusion | Innovate faster with generative AI for business
Contact us to
get started today.
cloud.google.com/ai/generative-ai