For buyers of AI technology, the two largest gaps in expectations vs. reality are the current capability of A.I. and what implementation of these tools actually looks like.

Artificial intelligence is facing a PR problem today. Ironically, many of the people building AI
software can take a bit of blame for it. As a business building a new category of software, it’s
difficult to strike the balance between exciting people about your solution and over-hyping the
possibilities, and there are still a lot of misconceptions about both the technology and the human
impact of deploying A.I. within companies.

Our company, Talla, has spoken with thousands of executives in the last couple years and now
I want to set the record straight regarding the state of the ecosystem and what adopting A.I.
means for your organization. The two largest mismatches in expectation vs. reality for buyers
are:

1) The state of A.I. technology today and
2) What implementing intelligent assistants or other A.I. software at your company really looks
like.

How much can A.I. agents really help you and your people?

First, it’s worth being clear about what A.I. is and is not. If you’ve spent any amount of time on
the internet in the last few years, you have run into chatbots on websites (or elsewhere) that
have no intelligence behind them. They may be a simple selection menu or information
submission, just through a conversational interface. Those systems typically have limited or no
flexibility in responses and run like a decision tree.

The true mark of machine learning is that they continue to make better decisions based on
additional data points or input. They may not necessarily learn instantaneously with a single
new data point or have flexibility in all areas.

Similarly, like a human, an A.I. assistant with no exposure to your company-specific data isn’t
very helpful. It would be like asking someone what the best sales tactics are, without giving
them any insight on the product being sold. The answer is, of course, dependent on the market,
the buyers, the product itself and pricing tiers, among other distinct variables. Now, imagine a
sales expert shows up with a vast, intimate knowledge of your product, company sales history
and the nuances of your market and customer. That’s the power A.I. software will have down
the road, when provided with access to training data that’s specific to your organization.

There are various ways to get ahold of that data and “feed” the system, but it’s typically more
difficult than people suspect. It’s not to say it’s not worthwhile—in fact, the longer you wait to
deploy, the more company information you have that’s not “ready.” A.I. isn’t something you can
just sprinkle on any old data—it needs to be properly formatted, esentially A.I. ready. Much of
what lives within businesses wasn’t optimized for that, as it’s been created over years or even
decades.

Typically, there are two paths for implementation:

1. Opening up a large, multi-participant deployment project where information about your
company is formatted in a way that machines can understand, so they can make
increasingly interesting conclusions. That involves multiple people within your
organization, and the right external business partner to get you set up. There’s a strong
case for this in companies who are willing and ready to spend the time.

2. To change your workflows slightly today with software that makes it easy to annotate
data so that it is machine readable, so that, over time, that data is increasingly useful
(with a longer time-to-value than option 1).

There’s very clear ROI for systems that can powerfully assist your employees down the line. For
instance, a study published by McKinsey showed that employees spend 19% of their time just
searching for and gathering information.

The point we try to drive home is the value in considering this now. Since A.I. software learns
and grows over time, like a human employee, that means the earlier you deploy it, the sooner it
will be smarter. And when your competitors deploy the same software 2 years from now, you
will have a headstart on learning— it will be years before they catch up.

What does the human-side of rolling out A.I. agents at companies look like?

Hopefully at this point it’s clear that rolling out an A.I. system often takes more work than simple
tools, but the long term benefits continue to grow. Besides employee time and resources, there
are other human factors to consider when you introduce digital workers to your organization.
You, as the leader, have to manage expectations for the project. Magic won’t happen on day
one. An HR executive or other senior level driver of adopting assistants needs to be explicit with
their teams that over time, when fed more data, the system becomes increasingly useful. Much
like with training a new employee, they can’t be put in a box and expected to be insightful. Some
amount of patience must be expected.

The good news is that we’ve seen employees that are receptive of and excited about engaging
with digital agents as a category. An office within the U.S. government’s General Services
Administration created their own employee onboarding bot, who they called Dolores
Landingham, in homage to the secretary from The West Wing. It was popular enough that it
wasn’t just new hires who wanted to connect with and receive information from Dolores. Existing
employees requested access and for additional content to be added so that they could receive
more information via the bot. While this is a simple example of information successfully
delivered via bot-interface, we’ve seen the same in deployments of Talla and other products
when they are providing a useful service to humans.

Finally, an important factor to consider when evaluating A.I. products for HR (or any division, for
that matter) is to select a partner who knows how to work with you and your people to make it a
long-term successful project. Beyond just the right technology, working to understand the
human side of bringing A.I. to your business is a success can’t be missed.