Almost every founder I talk to is asking

the same question. Am I falling behind

on AI? They're worried that they aren't

using AI enough. And if they don't keep

up, their business could become

irrelevant. Here's the thing. They're

not wrong to worry, but they're asking

the wrong question. Instead of am I

using enough AI, they should be asking,

am I using AI the right way? I've been

building out a framework around how

successful startups are using AI. And so

far, I've found that there are only five

distinct categories, and each one has

completely different risks and rewards.

Today, I'll walk you through all five.

And by the end of this video, you'll

know exactly which ones make sense for

your startup and which ones to avoid. As

you watch this video, I want you to

think about all the ways you're using AI

and think about which category each of

those fall into. And if you think of an

additional category that isn't covered

in this video, please drop a note in the

comments. Before we dive into the five

categories, let me tell you about

today's sponsor, G2I. Because

implementing almost all of these AI

strategies means finding qualified

engineers to build them. G2I doesn't

just give you access to over 8,000

prevetted engineers. They clear away the

chaos and clutter when it comes to

hiring. There's no AI generated resumes

and no time wasters, just solid

candidates with at least 5 years of

proven experience. G2I does the vetting

for you with customized live technical

interviews so you can actually see how a

candidate might work with your team.

Companies like Meta, Microsoft, and

ShopMonkey trust G2I, and so do

firsttime founders who just need to get

this hire right the first time. If you

need a prevetted engineer to join your

team quickly, head to g2i.co/microcom

to start your 7-day free trial. And if

you mention microcom, you'll get $1,500

off your first invoice. That's g2i.co.

co/microconf.

All right. So, category number one is

where AI is your core business. This is

where your product is AI. Meaning, if

you were to remove the AI, there's no

business left. This includes both

foundational models like Gemini 2.5,

Claude Opus, GPT5, but it also includes

companies like fiscal.ai, Jasper,

MidJourney. These companies aren't just

enhancing their product. The AI is

effectively the product. Some additional

examples are Rosie, founded by Jordan

Gaul. It's an AI answering service. One

Accord, it's a tiny seed company. They

offer live translation for churches. And

AI is a huge piece of this. And Pod

Squeeze. This is a bootstrapped startup

that helps you create and prepare

summaries and show notes of podcast

episodes. Does actually a lot more than

that. And we've used it for a few years

on startups with the rest of us and it

saves us a tremendous amount of time. So

with each of these AI categories, I'm

going to talk about upside potential and

primary risks. And so if AI is your core

business, the upside potential is that

you can potentially become the default

solution in your category. And there are

massive market opportunities in emerging

spaces right now. The primary risks are

commoditization, meaning today's moat

can be tomorrow's API call. Market

education burden. Customers don't know

what's possible. So oftent times with

stuff that's this new, you have to

educate the market. platform dependency.

Think about OpenAI and Anthropic and how

they could make a single change that

ends you. So if you have a PDF converter

and OpenAI just bakes that into the next

model release, your business will

probably go to zero. And the last one

only really applies if you're building a

foundational model, but it's capital

intensity. You need a huge amount of

funding for that. Category number two is

AI as a feature. So AI is not your core

product here. The key difference here is

that AI enhances your core product, but

it isn't the product itself. So, for

example, if Notion removed their AI

writing assistant, they'd still be

Notion. If Zoom killed their AI

summaries, meetings would still work.

And if Loom removed their AI titles and

chapters, which are very helpful, they'd

still be Loom. AI makes these products

better, makes them faster, easier, but

their core product value proposition

still survives without it. Let's look at

the upside potential for having AI as a

feature. One is you might be able to

charge more. So in conversations with

chat GPT, it suggests that some

companies are able to increase prices

between 20 and 50% for premium AI

features. Another potential upside is

adding AI now can offer differentiation

in competitive markets and it can

improve retention through habit forming

workflows. Finally, AI as a feature can

be a natural upsell path for existing

customers. Let's talk about the risks.

The first is that this isn't really a

competitive advantage because

competitors can copy AI features

honestly usually within a few weeks or a

few months. So, it is not a longlasting

advantage. Second is quality

expectations. Bad AI is often worse than

no AI and so you really have to vet your

results and what you're delivering for

your customers and that it is actually

providing benefits and it's not just a

checklist item that you build. The third

is cost management. AI API costs can

destroy your unit economics if your

customers use them often. And the fourth

one is overpromising, which really ties

back into quality expectations. But the

idea is in your marketing or your sales,

are you writing checks that the product

can't cash? The third is AI for building

your product. So this is for the devs in

the audience. This is where you're not

building an AI product. You're using AI

to build faster. GitHub Copilot, Cursor,

Claude Code. These tools mean an

engineer can now build more faster. The

AI never touches your customers, but

it's transforming how quickly you can

ship. Wind surf, lovable, so many

examples in this space. The upside

potential here is faster velocity. I've

heard someone quote, you can get a 3x or

10x development velocity improvement.

So, anytime you can make developers more

efficient and ship code faster,

obviously that's a good thing. This

allows small teams to potentially

compete with larger competitors. Your

iteration cycles can be cut down

especially in the early days from months

to weeks or weeks to days. In theory,

you can have a dramatic reduction in

technical debt and an expansion of test

coverage cuz like what developer really

loves writing tests and frankly AI is

actually quite good at that. So now

let's look at the primary risks. The

first is code quality. you can introduce

subtle bugs that are hard to spot. And

if you're not a senior or a mid-level

dev really combing through the code and

really looking at the test, it's easy to

build up technical debt. That's number

two. If you move fast without

understanding your code, you can

introduce technical debt and it's easy

to get kind of lazy or kind of sloppy

and start building technical debt over

time. The third one is over reliance.

Your team can lose the ability to code

without AI. Now, you're not going to

lose all of it, but you're going to lose

certain skills over time. And the fourth

one is security vulnerabilities. Is AI

introducing exploitable patterns? Today

it is. There are many examples of

specific tools that have introduced

security vulnerabilities. And you know

another one is that the development

velocity improvements are up for debate.

There was a recent study where folks

were saying that developers are not

actually faster with AI coding

assistance. Now I think the best

developers are actually better with it,

but I think that will shake out over

time. Before I get into category 4, I

wanted to let you know that tickets for

Microconf US 2026 are on sale. The event

is April 12th through the 14th in

Portland, Oregon. Microcom is where all

of this started. We've been running this

event for 15 years. It's a gathering of

275 to 300 of your favorite bootstrapped

SAS founder friends. It's an incredible

event. You're going to make connections

that will last you years. You're going

to build your network. You're going to

see incredible talks and be in the best

hallway track in the world for Bootstrap

SAS. To get all the details, head to

microcom.com/

us. And for subscribers of this channel,

you can use promo code rob50 to get $50

off your ticket. So, category 4 is AI

for growing your business. This is AI as

your growth engine. It's not in your

product, but it's in how you acquire

customers. This is AI powered cold

outreach that personalizes at scale.

It's content generation that helps you

dominate SEO, ad copy that tests

hundreds of variations automatically.

This is where you're using AI to grow

faster and cheaper than your

competitors. So, some examples of this

are Clay for outbound sales, Jasper.ai,

and Copy.AI for content, AI email

personalization tools, AI powered ad

optimization. And the upsides, I feel

like, are obvious. 10 to 100x increase

in outreach capacity, dramatically lower

cost to acquire a customer through

better targeting, content production at

crazy scale, and true personalization

that can convert better. But there are

some risks, and you might not have

thought of these. The first is

authenticity. Are customers going to

detect or reject AI outreach? Are you

going to cause brand damage when AI goes

off-brand in public? Are there

compliance issues? AI can violate

regulations if you're not paying

attention. And finally, channel

saturation. Everyone's using the same

tactics. Are they going to work now and

in the future? All right, my fifth and

final category is AI for operating your

business. This is where you use AI

internally. It's the stuff that helps

you run your company more efficiently.

Examples include AI handling tier one

support tickets, AI screening resumes,

analyzing customer feedback for

patterns. The key here is that your team

is the user, not your customers. It's

about operational leverage, not product

differentiation. Tools you might use

here include chatgpt claude or whatever

chat tool you prefer, intercom's fin for

customer support, any AI hiring or

screening tool, automated data entry,

processing, categorization, internal

knowledge bases, and AI powered

analytics. The upside potential is

significant. You can see 30, 50, 80%

operational cost reductions in certain

departments. You can scale support

without linear headcount. It offers 24

by7 availability. And of course, it can

achieve consistency that humans can't

match. But the risks are significant as

well. The customer experience can

suffer. What if AI mishandles sensitive

issues or frustrates your customers?

Employee morale might be an issue. What

if your team fears being replaced? And

finally, compliance and legal. We all

know that without human oversight, AI

can easily violate laws or internal

company policies. So again, those five

categories are AI as your core business,

AI as a feature, AI for building your

product, AI for growing your product,

and AI for operating your business. As

you look at your list of how you use AI,

if everything generally falls into these

categories, can you hit like on this

video? And if you have a use case that

doesn't fit into these categories or an

edge case that I should be considering,

I'd love it if you drop it in the

comments below. If you're considering AI

as your core business or as a feature,

you really need to think through

platform risk. In this next video, I go

indepth on the risks associated with

relying too heavily on an outside

partner and what you can do to mitigate

that risk. Thanks for watching. I'll see

you next time.