What is the first step you take to set up a successful AI initiative in a new organization or domain, and why?
Most answers emphasize defining a clear problem or objective as the first step, while others focus on leadership buy-in, process readiness, data quality, or building foundational AI understanding.
12 answers
Answers

Michael Bruun
Partner, Global co-Head Private Equity, Goldman Sachs Asset Management · Goldman Sachs
Get CEO buy-in
Michael Bruun says it needs to start with CEO buy-in, because it has to come from the top: the CEO must make clear to all employees that AI is an absolute must.
The answer starts at 32:30.
How I Invest with David Weisburd · Listen on Apple Podcasts

Amit Mathradas
Chief Executive Officer and Board Member · Five9
Identify the problem
Amit Mathradas says the first step is to figure out what problem you are trying to solve, since many people rush to throw AI at a problem before considering whether AI actually solves it.
SiliconANGLE theCUBE · Watch on YouTube

Vinod Muthukrishnan
VP / GM Webex Customer Experience · Cisco
AI readiness assessment
Vinod Muthukrishnan says the first job is to help people put points on the board, pointing to a document on AI readiness assessment, so teams can learn empirically at a small scale before scaling up.
SiliconANGLE theCUBE · Watch on YouTube

Autumn D. Krauss, Ph.D.
Chief Scientist, Future of Work Research Lab · SAP SuccessFactors
Build AI literacy
Autumn D. Krauss, Ph.D. says the first step is building AI literacy, which she describes as a foundational understanding of how AI works, including its ethics and fit-for-purpose use cases, rather than training on specific tools.
NZ Tech Podcast · Watch on YouTube
Conny Ploth
AI Transformation Vice President, Global Data & AI · Santander
Fix the process
Conny Ploth says the first step is to fix the process, because putting an AI layer on a messy process only speeds up inefficiencies instead of resolving their root cause.
The answer starts at 4:45.
Marketer of the Month · Listen on Apple Podcasts

Paula Goldman
Chief Ethical and Humane Use Officer & EVP, Product · Salesforce
Pick a few big bets
Paula Goldman says the first step is to pick a few big bets and go all in on them, because most companies already have enough data for that.
The answer starts at 27:48.
Thinkers & Ideas · Listen on Apple Podcasts

Sergio Torres
Senior Vice President Product Management · Descartes Systems Group
Define the first decision
Sergio Torres says to start by defining the first decision the agent should make, because that decision dictates what data is non-negotiable and other things are distractions until that decision works.
SupplyChainBrain · Watch on YouTube

Neil Morelli, PhD
Senior Director, Human-AI Collaboration and Workforce Transformation · Salesforce
Start small with a business case
Neil Morelli advises starting small but with a higher-level business case that keeps the big picture and a strong why in mind, because that builds more momentum.
The Science of Excellence · The Science of Excellence · Listen at 28:59

R "Ray" Wang
Principal Analyst, Founder & Chairman · Constellation Research, Inc.
Define the problem
R "Ray" Wang says to define the problem and understand its stakeholders, so the cool solution does not take over.
DisrupTV · DisrupTV · Listen at 14:21

Andrew Embry
Sr. Director Insights Innovation Capabilities- Global Core Market Research · Eli Lilly And Company
Identify a real problem
Andrew Embry says the first part of AI adoption that sticks is to identify a real, tangible problem for AI to solve, since claiming AI can do everything means it can’t really do anything.
The Curiosity Current: A Market Research Podcast · The Curiosity Current: A Market Research Podcast · Listen at 10:18

Valentina Jordan
CEO & Co-Founder · Nauta
Define a clear objective
Valentina Jordan says the first step is to have a very clear objective, because without a clear expectation of the work it will never give ROI.
TalkingLogistics · Watch on YouTube
JOSE MARTINEZ
Head of Costing Engineering and Program Control · Nissan Do Brasil
Ensure clean, curated data
Jose Martinez says you need clean, curated data that works and has the right value to run statistical tests, machine learning or LLM models; with unnormalized or duplicated data, you won't get the result you're looking for.
The Agile Brand with Greg Kihlström®: Expert Mode Marketing Technology, AI, & CX · The Agile Brand with Greg Kihlström®: Expert Mode Marketing Technology, AI, & CX · Listen at 16:21