Take Control of Your Company's Data-- Or AI Will.
The vast amounts of data that your company is now producing is the raw material of AI. You need to learn how to harness it to create better experiences for your customers. Focus on these four key areas.
EXPERT OPINION BY HOWARD
TULLMAN, GENERAL MANAGING PARTNER, G2T3V AND CHICAGO HIGH TECH
INVESTORS @HOWARDTULLMAN1
JUL 2, 2024
It seems that 200% of
the conversations that I'm having with the teams running our portfolio
companies, as well as those with other entrepreneurs across a dozen different
industries, are about chatbots, the dominance of the primary large language
models (LLMs,) and the unlikely prospects of the other emerging AI tools and
models. Very few of these discussions are focused on the much more pressing and
immediate concern: who's creating, who's aggregating, who's vetting, and who
owns the masses of underlying data that every business is now generating daily?
Further, how is all of this material going to be used to move those
businesses forward, and by whom?
The excessive focus on
the AI "engines" seems somewhat misplaced because it's the data
that's the fuel-- the source material, the content if you will-- which
ultimately drives these AI machines and determines the accuracy, quality and
value of their outputs. Every business is going to need to develop its own data
guidelines and monetization strategies, if any, and they'll need to decide how
they will use their accumulated data to improve and extend their internal
operations. Growing their customer attachment, loyalty and engagement and
improving the overall experiences of their users and clients will be one of the
most important data applications for the next several years. And because few of
these new firms have the requisite AI talent and skill sets already onboard,
figuring out how to employ the wealth of new digital information now at their
fingertips isn't going to be easy.
We've already learned
that these kinds of questions and concerns aren't as simple a matter as
"garbage in, garbage out" because the machines themselves have
already learned to optimize their responses for believability over accuracy.
That's because humans, being the lazy and Pollyannish folks that we
are, clearly prefer some answer-- in fact, any answer-- rather than a
system that responds with a goose egg or "I don't know." The tech
bros immediately started calling the LLMs programmatic tendencies to output
lies, fake statistics, and made-up fabrications by the anodyne term
"hallucinations," highlighting the fact that we're already losing
control of these systems and the ability to determine what is real and true in
many cases. Sadly, the developers have done a great job of simplifying the
output and a terrible job of providing audit trails, references, or other
checks and balances describing how the output and conclusions were reached and
what specific information they were based upon.
An even bigger question
will be whether and how these data-rich firms will share their information and
results with third parties while protecting the privacy and confidentiality of
their own users, customers, clients and partners. The exponential and voracious
demand of the LLMs for more and more content means that almost every company
will be approached and solicited to sell its data to users, brokers and other
aggregators. These financially appealing pitches will be especially attractive
to startups, because the vast majority of new businesses have their hands full
running their own shops, can always use additional cash for little effort, and
simply don't have the resources, people or time to develop and deliver
solutions that can economically market their own data to others.
But selling off their
data is likely to be a serious mistake in the long run. Even if they can't do
it all or even do it very well at the moment, successful businesses will need
to commit to figuring out how to use their own data to drive and enhance their
operations. Because the big winners of tomorrow will be those data-driven
businesses that are focused on two key objectives: (a) learning as much as
possible about each and every customer and prospect and (b) using those
findings to accurately predict and ultimately favorably alter the behavior of
these individuals to drive improved financial performance.
Every company can start
down this road to greater enlightenment by understanding that-- specifically in
the eyes of the consumer-- there are pretty clear lines between proper and
improper uses of customer data, even apart from any legal requirements. Most of
us long ago made a series of decisions and trades where we largely gave up on
transactional (as opposed to personal) privacy in the interests of ease of
access, convenience, and saving time. As a practical matter, this means that,
so long as businesses stay within four basic guardrails, we're more than
willing to have data about our behavior and activities in their possession used
to make our daily digital lives and our dealings with them easier and simpler.
The four key areas of
permissible data use are also the key elements for any company's comprehensive
data management plan. And while they may ultimately be accelerated, streamlined
and strengthened by the processing, analytical and predicative capabilities of
AI, the fact is that the foundations for each component already exist in many
firms and can be developed and utilized in a less technical and complicated
fashion right now. The main requirement is to get focused and get started. The
four areas are:
(1)
Speed and Ease of Access and Use
This is probably the
easiest and most obvious area where customer information management systems can
be used to speed the selection and shopping process, eliminate redundant
actions, supply prior information and apply preferences, accelerate payments,
and offer alternative delivery options. No one is unhappy about having their
time saved and their patience salved and only prudes care about their
"privacy" when they're just trying to get something done and done
right as quickly as possible.
(2) Service and
Support for Customers, Products and Services
As hard as it is to
sometimes believe, millions of businesses treat their returning clients and
customers as strangers, newbies or random visitors rather than immediately
acknowledging their prior connections, service history, and product or service
needs. On the other hand, we're also seeing more and more instances of
proactive service alerts and interventions (smart washing machines come to
mind) so that problems, interruptions in use and service, and other issues can
be resolved before they arise-- and often even behind the scenes without any
involvement from the user. Here again, no news is the best news if it means I
don't have to worry about something abruptly breaking down in my home or
office.
(3) Suggestions,
Bundles, Subscriptions and Sales
Anyone who has made
multiple trips to the store or a website because you forgot certain simple but
critical items quickly comes to appreciate smart systems, which do some modest
suggestive selling, remind you unobtrusively about packages, bundles, recurring
purchase discounts, and related purchases; or the cables and other pieces
necessary to operate and enjoy your new gizmos. Some firms, like Amazon, do
this far better than others in large part because Amazon knows what you have
recently purchased and doesn't waste your time or its opportunities offering
you a chance over and over again to buy the same item you selected three days
ago and won't need to restock for another six months.
(4) Surveys and
Satisfaction Measurement and Management
Some things never
change. Happy customers may tell a friend or two about a successful purchase or
service experience; unhappy customers badmouth your business all over town.
This is why customer service and satisfaction management programs are
absolutely essential but also must be executed with care and in a manner that
isn't unduly intrusive. This is just as difficult as it sounds, but the best
organizations make it their business to actively and promptly ask their
customers for feedback and to act on the information they receive. Getting a
survey in the mail 90 days after you visit some car dealer or have your vehicle
serviced is a waste of everyone's time. Real time inquiries and responses make
an important impression on customers and clients, demonstrate a sincere
interest in them individually (assuming it's not some pro forma phone
call from a call center in a distant place) and absolutely lead to additional
customer engagement and revenue.
Bottom line: you don't
need a magical machine or some new AI genius to start using your wealth of data
to do a better and more comprehensive job of serving and supporting your
customers. Rather than being offended by the fact that you know something relevant
about them and their connection with your firm, they'll be the first to say
that they're grateful that you aren't wasting their time.
