Q&A
The Case for Starting Small with AI
Josh Rozman explains why successful AI adoption begins with fixing one frustrating workflow, not buying another platform.
By this point, the scales have fallen from most organizations' eyes; they know that AI won't transform a business simply because it's installed. The point is to apply AI to real-world problems that consume time and limit growth.
That's the message Josh Rozman will bring to the in-person AI Pivot event this fall, taking place in Anaheim, Calif. Rather than encouraging organizations to chase the latest AI tools, the founder of AI Implementer argues that businesses should focus on improving one repetitive workflow at a time, building repeatable systems from those early wins.
In this Q&A, Rozman discusses how AI can help small teams create more personalized marketing, uncover valuable customer signals already hiding in their CRM data, and prioritize technology investments that generate actual ROI. He also shares why organizations that achieve lasting success with AI focus on building deep operational foundations instead of pursuing quick wins.
Pure AI: Many small businesses feel overwhelmed by the number of AI tools available today. What is the
first AI-driven workflow you recommend implementing to help them get the ball rolling?
Rozman: The key is to look beyond the tools for a moment. The number of tools out there only adds
to the noise, and chasing after all of them is why most people feel stuck before they even
choose to start.
Start by picking one task you already dread. I'm talking about that one thing you do over
and over every week that eats away at your time and never feels worth it. It could be
something like writing the same email repeatedly. Maybe it's turning meeting notes that are
all over the place into something useful. You want to pick that one thing, open up your AI of
choice, and talk to it like you would a new intern using plain words. Don't try to get fancy
here.
The first answer AI spits out will get you about 70% of what you wanted. So then tell it what
to fix and it gets to 90%. Now pay attention to this moment, because right here, in just a
few minutes you just did the thing that used to cost you half an hour.
That is the whole starting point. You don't need a course. You don't need an online
certificate. You just need one real task that actually matters to you and the will to go back
and forth a few times.
What happens next is the part nobody tells you. Once you feel that first win, you start
spotting these tasks everywhere. It snowballs on its own. The goal on day one is not to
transform your business. The goal is to win one small race so you can see how the engine
runs.
In your AI for Marketing Content, Campaigns, and Customer Personalization session, you demo
turning a single idea into a personalized campaign. What are the biggest mistakes companies make
when trying to do content creation with AI while maintaining an authentic brand voice?
The biggest one is that most people try to win the race on lap one. When was the last time
anyone saw that happen in real life?
Let's use a long car race for example. Nobody wins it on the first lap. You have hundreds of
laps to go, and the car that wins is the one built right, one part at a time, every piece in sync
with the next. A lot of companies skip all of that. They want AI to write the whole campaign,
sound perfect, and win on the first draft. So, they start from a place of failure without even
knowing it.
The second mistake hides inside the first. They chase. Every week there is a new hot online
tip, some new trick everyone is suddenly doing, and they go running after it.
The problem
with chasing is simple. The only way to chase is to follow. The only way to follow is to stay
behind. You cannot lead from back there.
The companies that keep their real voice are the quiet ones building their own engine in the
garage while everyone else is asleep. They are not copying the trick that worked for
someone else, because that trick was built for a different business with different
customers. Your true authentic voice comes from understanding your own business
deeply enough to build something custom to it. That work is slow and unglamorous, which
is exactly why most people skip it. Double down on this.
What techniques have you found most effective for making AI-assisted communications genuinely
relevant to customers and prospects?
It starts with one question most people skip. Who am I actually talking to?
So many of us want to jump straight into the message. We tell AI to write something
persuasive before we have any idea who is on the other end. That is backwards. If you do
not know who you are talking to, you cannot understand their problem. Miss the problem,
and every word you write is a guess.
So, I recommend using AI as a thought partner. First, I'd have it help me break my audience
into real segments. Then I drill into each segment to understand what that specific group is
actually dealing with. Only then do I let it help me shape a message in their language, the
way they would say it themselves. When that email lands in their inbox, it feels like it was
written for them, because it was.
The technique most people miss comes after all of that. This is my favorite part. The part
that actually makes someone reply, or call, or buy, comes down to the details. The human
details. The emotion sitting underneath the words. AI gets you 80 percent of the way there
fast. The last 20 percent, the part that makes someone feel something, that is all you.
People notice it. They can't always name it, but they feel it. The magic lives in the details,
and that is wide open ground right now, especially now with AI, because almost nobody
slows down enough to get them right.
In From CRM Chaos to Customer Insight: Practical AI for Small Teams, you describe uncovering
insights from data that companies already have. What are some examples have you seen in the real
world of valuable customer signals that businesses often overlook?
The one I come back to most is the lead that has gone quiet. The cold leads sitting in the
CRM that nobody is touching anymore.
I want you to think about a lead like a cup of coffee. When it first comes in, it is hot. Fresh
off the pour. The salesperson grabs the cup, adds a little cream and sugar, sweetens the
deal, and just like that everyone is paying attention. That part is easy. The trouble is what
happens next. Does coffee stay hot on its own? No of course not. Left alone on the desk, it
goes cold. A cold cup of stale coffee is something nobody wants to pick back up.
That is the signal hiding in plain sight inside almost every CRM. The activity timeline. The
last-touched date. Most businesses look right past it. They assume that once a lead is in
the system, it stays warm. It does not. To keep that coffee hot, you have to put energy back
in, the way a microwave gets the molecules moving again. For a lead, that energy is touch
points. Not one. Many. An email, a call, a text, spread out over time.
Where it goes wrong is the reheating. Most owners try to do all of it by hand, and the buck
stops with them. So, weeks slip. Sometimes months. The customer slowly forgets you
exist. The fix is not more hustle. It is a system that watches the timeline and tells your
people exactly when a cup is going cold, so the energy goes back in before the lead is gone.
The data to do this is already sitting in the CRM. Almost nobody uses it.
For organizations with limited budgets, how do you prioritize investments across marketing
automation, sales automation and intelligent CRM?
For me the order is really clear. Marketing automation is always first, a smarter CRM
second, and sales automation third.
Start with marketing, because nothing else matters if you cannot get attention. You could
have the best closers on the planet sitting by the phone. However, if you can't cut through
the noise and get someone to stop what they are doing and listen, you lose before you
begin. Marketing automation, done right, means you stop being the bottleneck. You are no
longer the one person who needs to remember to send the marketing emails every week.
Systems handle that, so it happens thousands of times without depending on you to never
drop the ball.
A smarter CRM comes second, because data that just sits there will spoil if it's not
maintained.
Your customer data works the same way. If it does not flow and stays clean
from the very beginning, it rots, and you end up making decisions off garbage.
Sales automation comes third, and only third. Once you have the attention coming in and
clean signals to work with, then you get to build the system that helps you close. Try to flip
that order, and you are pouring money into closing deals you never had the attention to
create, off data you cannot trust. That is the trap. Most people fund the end of the race
before they have built the start of it.
AI tools are evolving rapidly. Which capabilities do you think will become standard for marketing
and sales teams over the next two to three years?
In the next two to three years, AI stops being a thing you go open and starts being
something that's baked into everything you already use. Like electricity. You stop noticing it
is there.
For marketing and sales teams, here are a few things that you should look for in the next
few years. Your CRM stops being a filing cabinet you feed by hand and starts surfacing the
right lead at the right moment on its own, telling your team who is getting cold and who is
ready for a conversation.
Next, expect your content tools to draft the first version of
everything, so nobody stares at a blank page anymore. Personalization at scale will stop
being a luxury for big companies with big teams. It will no longer take a week of planning
and work to make things like this happen.
The bigger shift, the one I would watch closely, is the move from AI that answers to AI that
acts. Right now, most people use AI to get a faster answer. Soon you will hand it a full job,
and it will go and do it. It will follow up with the lead. It will book the sales call. It will
manage the routine back and forth from start to finish, and will only pull in a human when
they need to step in.
The teams that start building these habits now will look up in two years already fluent,
while the ones who waited are still reading the manual. Don't be that person.
From your experience, what separates organizations that successfully adopt AI from those that
experiment with it but don't get lasting results?
Depth. The winners are already going deep. The experimenters stay shallow and call it a
strategy.
Just take bamboo for example. A farmer plants it, and for a long stretch it looks like nothing
is happening above the ground. No height. No show. Underneath, though, it is laying down
a root system nobody can see. When those roots are finally ready, that bamboo sends up a
single shoot that climbs close to three feet in a day and passes sixty feet in a matter of
weeks. People love to say it sat there doing nothing for years, but it was never doing
nothing. The surge everyone claps for was paid for by the roots it built while no one was
watching.
That is the whole difference.
In my opinion, companies that will get lasting results from AI
over the next decade are already building deep roots. They are taking the time to
understand how AI wires into their business, where it connects, where it breaks. The ones
who experiment and fizzle out went straight for height. They bolted a shiny tool onto a
broken process, grabbed a quick win they could post about, and stopped there.
When AI first showed up, I watched people sprint out the gate showing off products they
had built in a weekend. Most days it felt like I was the one falling behind. I was measuring
the wrong thing. I was staring at everyone else's height while I was busy building depth.
We are now entering into the season where that gets exposed. The company that went all
in and the company that went for show are about to stand next to each other, and you will
be able to feel the difference the second you use what they built. The part that should light
a fire under anyone still on the fence is this. Learning compounds. The company that truly
started a year ago is not a year ahead of you. It is closer to three years ahead, because the
learning stacked on itself the whole time.
If someone leaves your sessions and has only one afternoon to improve a business process with AI,
what project should they tackle first?
Take the one process that drains the most time for the least reward and rebuild just that
one in a single sitting.
Do not pick something big or strategic. Don't overthink it when you re first starting out. Pick
the boring, repetitive thing that quietly eats your week. The same report you rewrite. The
follow-up emails you keep meaning to send and never do. It's not the most glamorous but
remember who it's for. It's for you. It doesn't matter what it looks like, it only matters if it
works for you. One process is the key, start to finish.
Spend the afternoon doing three things. First, write down every step of how that process
works today, exactly the way you do it now. That part alone is worth the time, because
most people have never actually seen their own process laid out in front of them word for
word. Second, walk AI through each step in plain English, and have it help you do each
piece faster. Third, the step people skip: save what you built. Write down how you got there
so you can scale it and run it again next week without starting over.
That last step is the whole game. A win you cannot repeat is just a good afternoon. A win
you wrote down becomes a system. Do this with one process this week, another next
week, and by the end of the year you are not starting at level ten anymore. You are starting
at level 100, because every piece you built is still working for you. You only had to build it
once.