The
1 Thing AI Still Can’t Do Better Than a Human (and How Startups Can Use It to
Build a Moat)
As AI kills the traditional software
moat, this strategy can help startup founders survive.
EXPERT OPINION BY HOWARD TULLMAN, GENERAL MANAGING PARTNER, G2T3V
AND CHICAGO HIGH TECH INVESTORS @TULLMAN
May 12, 2026
For quite a long time,
the most common assertion regarding the threats presented by AI has been that even the most advanced systems
will never reach the point where they will replace humans who are creative and
innovative and who are designing new ideas, products, and solutions every day.
At the same time, these are the very folks looking most anxiously over their
shoulders at the oncoming onslaught.
To the extent that
millions of these new ideas are never commercialized or even implemented—and
worse yet, that other millions of these concepts turn out to be nothing more
than incremental changes, enhancements or extensions of existing products and
services which essentially add nothing to the aggregate base of human
knowledge—it’s no great loss to humanity that these kinds of menial materials
will soon be left to the tender mercies of AI-driven applications and programs.
Ideally, getting rid of the scut work will free up folks to do more challenging
and valuable tasks. Hopefully sooner rather than later.
The absolute flood of AI
slop that’s already overwhelming every social media channel and other
information delivery system serves as the first of many proof points in this
regard. If humans who previously toiled in the creation of massive mounds of
this kind of promotion, media, marketing and
advertising crap are soon replaced by machines doing the same work more
efficiently and economically, there’s an argument to be made that we’ll be
doing those poor impoverished souls a favor to put them out of their misery so
they have some prospect of finding meaningful work. And it’s not like the
ultimate AI-driven output is likely to be materially better or worse. The truth
is that you can’t polish a turd no matter how hard you try.
More recently, there has
been a despondent group of technologists who dejectedly argue—especially
because AI has proven so overwhelmingly adept at coding—that there are simply
no longer any barriers to the advance across the board of these technologies. They
believe that even building the best and most novel software offers an
entrepreneur, a new business builder or even a senior and highly talented
developer no sustainable moat or substantial protection from readily
available AI tools simply
copying, rewriting, reverse engineering or otherwise duplicating any of their
new offerings and solutions in a matter of hours or days. How accurate this
threat turns out to actually be is an open question, but it doesn’t take much
talk like this to scare away early-stage investors and prospective employees.
Years ago, the fear was that Microsoft (and then Facebook) would either buy you
or roll right over and crush you. Today, AI is the new boogeyman.: 102006)
So, the real strategy
for software startups looking to survive beyond only a
momentary flash in the pan seems to be a two-fold approach. First, take what
you need in the way of funding but stay lean and don’t be a pig in terms of
raising capital because that only makes it harder for you to pull off the
easiest of the most likely positive outcomes. In these crazy times, if you hit
on a compelling idea and can build an early viable offering, you want to always
have one eye on a quick exit.
I call this plan: Build to be Bought. You want to make sure that, when an
eager buyer shows up, you haven’t created too many financial or other
impediments to an attractive sale which can get in the way of giving your
investors and your team a great return and give you all the ability to happily
move on to the next challenge.
The second plan, if
you’re planning to stay in the race for the long run, is to keep moving forward
and head to where the machines can’t follow. Your most effective moat is that
you’re constantly in motion and that you’re always at the tip of the spear, which
is essentially and inevitably the point of human contact and interaction. This
is hard, but not as difficult as you might imagine because what it translates
into is always being focused on and building to the front end—upgrading,
simplifying, and extending the points at which the end users access and
interact with your product or service. That relationship which is so central to
every part of our lives will never be fully appreciated and mastered by the
machines because they don’t appreciate that we’re never going to be willing to
fully abdicate our actions to any of our devices or machines.
The machines keep
getting swifter and smarter, but they will never bridge the final space which
will always be defined by EQ emotional considerations (the user interfaces)
rather than IQ technical attributes (the operating core). We see this dilemma
every day when prospects are presented with and swayed by powerful utility
claims and new levels of agency and then are quickly and totally turned off by
the substantial technical implementation and onboarding challenges which they
never signed up for. New users and even early adopters don’t want to build
these things; they just want them to work. They don’t want to read a manual or
learn a new trade or even invest a reasonable amount of time learning the
basics.
This is where AI falls
off the cliff because the machines’ tendencies are always to growing
complexity, increasing bells and whistles, and expanding functionality while
the target users want simplicity, rapid access, obvious controls and inputs and
useable results and outputs. This is the old curse of engineers who are
building to impress their peers and not to satisfy the real needs of their
customers. It’s the reason that less than 5 percent of the tens of millions of
users of all of the major Microsoft products never use or even discover 95
percent of the bloated and buried functions and features of the software.
The moat that still
makes the difference and the key to sustainable success is to be constantly
focusing on the quality of the end users’ experience, building customer
confidence and continuity, and managing and meeting or exceeding the buyers’
expectations. The goal is simple: you want the competition to find your warm
campfires and by then, you’ll be over the next hill.
