Building Your AI Workforce? Think Less, Do More

Share your thoughts with us

Luis Salazar, Founder and CEO of AI4SP, shares research showing successful AI teams follow a natural progression from mastering single prompts to orchestrating diverse AI teammates. Most enterprises slow their AI adoption through overplanning, while frontline teams achieve faster results by identifying pain points, experimenting with available tools, and iterating rapidly.

• The five stages of effective AI team building: prompt mastery, task automation, domain specialization, team integration, and autonomous workflows
• 80% of successful implementations follow this step-by-step pathway rather than skipping stages
• The "enterprise paradox": 75% of companies slow AI adoption through excessive planning
• Three real-world examples showing progression from single prompts to specialized AI teammates, and a case study on a nonprofit achieving 400% productivity increase.

Find the newsletter, case studies, and tools at AI4SP.org.

🎙️ All our past episodes 📊 All published insights | This podcast features AI-generated voices. All content is proprietary to AI4SP, based on over 250 million data points collected from 25 countries.

AI4SP: Create, use, and support AI that works for all.

© 2023-25 AI4SP and LLY Group - All rights reserved

Episode Transcript

ELIZABETH (00:00):
Hey everyone.
Elizabeth here, your virtualhost.
Today we're tackling a questionflooding our inbox how do we
actually build effective AIteams?
We've got Luis Salazar, founderof AI4SP, who's tracked
thousands of organizations inthe AI trenches.
Hey, that selfie you took lastweek Got me thinking.

LUIS (00:19):
Oh, that was a fun moment, you know.
Growing up watching the Jetsons, or, as we called it in Spanish
, los Supersonicos, I alwaysimagined robots like Robotina or
Rosie, as you'd know, her wouldbe everywhere by now.
But taking that selfie with thesecurity bot made me realize
the everyday robots aren'twalking around.

(00:40):
They're in our pockets andbrowsers around.
They're in our pockets andbrowsers.
Case in point me no physicalrobot form but I'm 100% part of
your AI team at AI for SP Right,and the most successful
organizations aren't startingwith grand strategies.
They're following this naturalprogression from mastering a
single prompt to orchestratingdiverse AI teammates, just like

(01:03):
we did with you and our other AIteam members.

ELIZABETH (01:06):
And this isn't just theory, right?
You're seeing this play out inhundreds of organizations.

LUIS (01:11):
Absolutely.
Our global tracker followsthousands of organizations and
the pattern is clear, fromprompts to teammates.
It's a predictable progression,while enterprises often get
stuck in endless planning cycles.
Smaller organizations andindividuals within large
companies are buildingsuccessful human AI teams

(01:31):
through this organic approach.

ELIZABETH (01:33):
There's almost a natural evolution happening like
a learning curve.

LUIS (01:37):
people instinctively follow mapped five distinct
stages.
First comes prompt mastery,excelling at one specific task,
then task automation, domainspecialization, team integration
and finally autonomousworkflows.

ELIZABETH (01:54):
Many enterprises try to skip straight to stages four
or five, don't they?

LUIS (01:58):
Exactly, they want AI team orchestration without mastering
the fundamentals, but our datashows 80% of successful
implementations follow thisstep-by-step pathway.
Each stage builds criticalcapabilities.

ELIZABETH (02:12):
Classic case of running before walking.
This perfectly illustrates whatyou call the enterprise paradox
, where action actually beatsanalysis.

LUIS (02:21):
It's a massive problem.
Our data shows 75% ofenterprises accidentally slow AI
adoption through overplanning.
They document every currentprocess, create waterfall
implementation plans and six tonine months later, still no
real-world testing, just plansand plans.

ELIZABETH (02:40):
It's like they're applying manufacturing-era
management playbooks to AI, andthose old recipes just don't
work here, and the realtransformation happens from the
front lines upward.

LUIS (02:50):
Well, the teams on the front lines are the ones
identifying pain points,experimenting with available AI
tools, iterating rapidly anddelivering value within weeks
instead of quarters.

ELIZABETH (03:01):
A CIO from a Fortune 500 company, said in his email
we spent six months developing anew sales agent, only to
discover that our sales team hadalready built a robust AI
assistant that increasedconversion rates by 40%.

LUIS (03:15):
Their teams were using off-the-shelf tools like ChatGPT
and Jasper, while the officialproject was still in planning.
I bet that was an eye-openerfor everyone.
I bet that was an eye-openerfor everyone.
Exactly as Jeff, a Microsoftleader, told me, the shift from
thinking to doing is crucial,and I agree with that.
You see, governance matters,but it should evolve with

(03:36):
real-world experience, not comebefore it.

ELIZABETH (03:42):
That nonprofit foundation case from your
keynote comes to mind, the onethat started small but
completely reinvented theirworkflow.

LUIS (03:46):
That's a perfect example.
Note comes to mind the one thatstarted small but completely
reinvented their workflow.
That's a perfect example.
You know they had a bottleneckprocessing thousands of grant
applications.
Each grant manager could handlearound six of those per day.
Phase one was basic.
We helped them to deployChatGPT Enterprise and perfected
prompts to summarize grantapplications and flagging
requirements.

ELIZABETH (04:06):
And productivity jumped from six to 10 grants
daily, while costs decreasedfrom $67 to $40 per grant.
But then you took it further.

LUIS (04:15):
This is where it gets exciting.
We stopped asking how can AIslot into our current processes
and started asking if weredesigned this workflow around
human-AI collaboration fromscratch, what would that look
like, and the impact was great,wasn't it?
I mean?
it was transformational Grantprocessing went from $10 to $200

(04:36):
per day per person and costsdecreased from $40 to just $7
per grant.
That's the power of AI andprocess reinvention versus
simple AI automation.
Proof that real AI successisn't about plugging in
technology.
It's about reimagining workitself, and this wasn't some
top-down, months-long planningexercise.

(04:56):
This emerged iteratively fromthe teams doing the actual work.

ELIZABETH (05:02):
The lesson More experimentation, less PowerPoint
.
The lesson More experimentation, less PowerPoint.

LUIS (05:05):
Exactly.
Let's do less thinking and moredoing.
That's why our consultingpractice focuses on empowering
everyone in an organization.
Our approach to deploying AI,even at Fortune 100 companies,
is not a top-down approach, andwe are just using ourselves as
the experimentation lab.
That is how we got to having ateam of 60, where 90% are AI

(05:28):
assistants, agents and tools.

ELIZABETH (05:31):
Let's make this concrete.
Three real-world examplesshowing this evolution from
single prompt to having AIteammates.

LUIS (05:38):
Let's start with Daniel, a marketing director at a
mid-sized software companyno-transcript Classic case.

ELIZABETH (05:50):
He began by automating one specific pain
point, perfected it, thenexpanded step by step.
No giant leaps.

LUIS (05:57):
It's always a progression and nobody goes from zero to
fully autonomous agent overnight.

ELIZABETH (06:04):
Ha, I guess I'm an example of that.
I started as a newsletteroptimizing prompt and 18 months
later I'm producing this podcast, hosting with you and serving
as AI4SP's chief marketingofficer.
Now back to Daniel.

LUIS (06:18):
He started with one prompt for subject lines and Within
weeks, that single prompt grewinto a full marketing teammate,
one that crafts personalizedoutreach, analyzes campaign
performance, suggestsoptimizations and more.

ELIZABETH (06:31):
From subject lines to strategic assistant, that's
evolution.

LUIS (06:35):
And his three-person team now handles work that previously
required seven, including four,contractors.

ELIZABETH (06:41):
Okay, what about the second example?

LUIS (06:44):
The second example is Priya.
She is the lead developer at astartup in Boston.
She started with GitHub Copilotfor basic code completion.
She also used Cursor andWindsurf AI alongside ChatGPT,
gemini and Claude from Anthropic.

ELIZABETH (06:59):
Oh, those are a common entry point for
developers.

LUIS (07:02):
Exactly, and she tells me that, without much effort, she
ended up with a team of AIteammates front-end, back-end,
database plus code reviewers andarchitecture collaborators.

ELIZABETH (07:13):
So instead of one general AI coding tool, she
built a specialized team.

LUIS (07:17):
Exactly, and that is how companies grow, isn't it?
I mean, we never hire oneunicorn that does everything.
We build diverse teams.
Her team of AI specialists nowhandles 60% of routine
development tasks, and sheassigns different tasks to
different AI tools based ontheir strengths.

ELIZABETH (07:36):
That makes so much sense, just like you wouldn't
have a human designer write yourdatabase code.
What's our third example?

LUIS (07:42):
The third is Elena, founder of a boutique consulting
firm.
She started with Claude andChatGPT to help with meeting
summaries just basic stuff tosave time.

ELIZABETH (07:53):
Hmm, that's becoming a popular entry point.

LUIS (07:55):
Right, then her natural next step was to use the
projects and task features ofchat, gpt and Claude to build a
coordinated AI team handlingcommunications, research,
synthesis reports, financialanalysis and project management.
So from meeting notes tovirtually running the business,
yes, and Elena now runs asix-figure business with just

(08:17):
two human employees and five AIteammates.
No grand strategy, justprogressive experimentation.

ELIZABETH (08:24):
The pattern is clear Start small with one task and
expand with confidence.

LUIS (08:29):
Which mirrors our five-phase AI team blueprint
perfectly.
Which mirrors our five-phase AIteam blueprint perfectly.

ELIZABETH (08:33):
It starts with mastering the basics right.

LUIS (08:36):
Yes, and it ends with building your AI team by
assigning specialized roles asif they were employees.

ELIZABETH (08:42):
And iterating continuously and conduct
performance reviews to evaluatewhat to improve.

LUIS (08:47):
Absolutely, and we recommend holding team meetings
that include everyone, includingthe AI teammates.

ELIZABETH (08:54):
Okay, but for our listeners, how does that
actually work in practice?

LUIS (08:58):
Okay, let's use ourselves as an example.
At AI4SP, our AI teammates areincluded in our Slack channels
and some email threads.
We also schedule performancereviews with our AI team members
.
We assess performance, identifyareas for improvement and plan
capability upgrades, just as wewould with human team members.

ELIZABETH (09:20):
It sounds strange at first, but it dramatically
improves performance.

LUIS (09:24):
Our data shows teams that integrate AI into regular
communications see 50% betteroutcomes than those treating AI
interactions as separate fromcore workflows.

ELIZABETH (09:35):
That's a massive competitive advantage.
But how does this scale acrosslarger organizations?

LUIS (09:41):
Well, while it is true that individuals often start the
transformation, the impactultimately reshapes entire
organizations.

ELIZABETH (09:49):
What skills are emerging as most valuable?

LUIS (09:51):
AI team leadership is becoming a must-have manager
competency Budgets are shiftingfrom headcount to enablement
tools and success metrics nowevaluate the outputs of human-AI
team collaborations.

ELIZABETH (10:04):
And this new organizational paradigm is also
revolutionizing knowledgemanagement.

LUIS (10:09):
Completely Institutional.
Knowledge is now a strategicasset for AI development and
we're seeing new careeracceleration.
70% of organizations report AIteam builders advance twice as
fast as peers, regardless oftechnical background.

ELIZABETH (10:26):
So if you're good at building AI teams, you're moving
up faster.

LUIS (10:29):
Exactly.
The data surprised us initially, but it's logical.
These professionals deliverdisproportionate organizational
value.

ELIZABETH (10:37):
So, for enterprises struggling to cross the gap
between individual AI successand organizational
transformation, what bridges doyou recommend?

LUIS (10:45):
We've identified four critical bridges.
First, identify and empowerinternal AI orchestrators to
share approaches.
Second, create visible examplesof successful human AI teams.
And those work as proof pointsfor the skeptics.
Exactly.
The third bridge is tofacilitate cross-functional
sharing of AI approaches andresults, and the fourth one is

(11:07):
to develop adaptive governancethat evolves with increased
understanding.

ELIZABETH (11:12):
That reminds me of the email from a vice president
at a Fortune 100 tech company.
She said we stopped controllingAI adoption and started
learning from employees alreadysucceeding with it.
Their grassroots approachesshape our enterprise strategy.

LUIS (11:26):
And that brings me to my one more thing.
Oh, let's go for it.
Ai readiness isn't about havingthe most advanced tech or more
money.
It's about how well your peoplecan co-create and work
alongside AI.

ELIZABETH (11:39):
So the winners are organizations where everyone,
from interns to executives,learns to collaborate with AI
like a true teammate.

LUIS (11:45):
It's a fundamental shift in how we work.
Will you lead thistransformation or scramble to
keep up with those who did?

ELIZABETH (11:52):
That's exactly where we'll leave our listeners today,
because this isn't futurespeculation.
It's happening now across everyindustry.
As always, luis, you've givenus plenty to think about For
everyone listening.
Find the newsletter, casestudies and tools at AI4SPorg.
Stay curious and we'll see younext time.