Please welcome to the stage,
Madhav Thattai, Senior Vice President and
Chief Operating Officer of Agentforce Product at Salesforce.
Chief Technology Officer at OpenTable.
Well, it’s great to be here with all of you.
Sagar, thank you for being a great customer at Salesforce.
You know, a lot of customers, a lot of companies
where do they get started on their agentic journey.
What’s the thing that I should go after?
You guys have done a lot across
your customer experiences, your employees.
So how did you get started?
How did you think about where you should focus?
Yeah, so when we started, really, we were looking for efficiencies
for our three audiences: restaurants, diners,
and our internal employees.
conversations in a much more fluid and authentic way.
So we started by saying, where do we have
those conversations across those three audiences?
Customer service was a natural first step.
Thank you, Salesforce, we partnered with you
to launch our agents and our chat experiences.
For us, a key metric was how many questions
are we resolving without escalating to a human agent?
And we were able to improve that pretty dramatically.
But the potential is just so much higher as well, right?
how does the agent interact with our own internal tools and APIs?
We know that, you know, over time we can just get better and better.
For our restaurants, actually, a common thing
that they had to deal with was people calling the restaurant
And it’s funny because OpenTable, when it was founded 26 years ago,
was founded with the simple premise of make a reservation online
You don’t have to talk to anyone, right?
But despite that, people still pick up the phone, call the restaurant.
And so what we said is, okay, how can we
partner with great voice AI companies to bring a solution to market?
Allows the restaurant to answer the phone automatically.
Make a reservation, answer questions.
It’s seen great success so far.
Millions of calls answered,
you know, hundreds of thousands of reservations.
the person calling doesn’t even realize they’re talking to an AI.
What I love about that is, I think picking an outcome,
This is what I want to orient my experience towards, I think is a much
healthier approach to this versus
just kind of trying to boil the ocean with a bunch of experiences.
Now, as you’ve been on this journey,
would love to hear about what are some of the challenges
What are the things that you felt like
you had to put more energy around?
How did you get ready for it?
love to hear what that implementation journey has looked like for you.
in addition to figuring out what’s your metric,
I think understanding the data and,
is your data in a state where it can be consumed by this
AI is a critical concern.
So, for internal efficiencies, we use a tool called Glean.
It’s, you know, enterprise search, knowledge management.
And what it’s based on is it says,
hey, I’m going to take all this internal data and provide you
an answer in a more intuitive way
without having to read through all the documents, right?
So previously, I might have just shared the links.
Now it’s going to give you an answer.
Obviously, like anything, it’s only as good as the inputs, right?
And so ultimately you need to really think about, okay,
if I’m creating a customer-facing experience, especially,
what is the data going into it
and how do I evaluate the answers coming out?
And so that can be testing, right?
I have a set of questions that I know
the answer to and I’m going to compare and do some evaluations.
It can also be things like monitoring for sentiment.
So is the person that is interacting with this AI frustrated?
Can you detect that from the conversation trace or the voice?
then incorporate that into your learnings and your feedback loop?
Because if you’re not doing that,
have this thing out there and you just don’t know how to evaluate.
Is it doing the right thing or not?
every step of that journey from let’s pick the outcome.
Let’s think about the configuration.
Let’s think about what the data sources
are all the way through deployment,
implementation, monitoring, testing.
It’s like we’re in this world
now where we’re kind of creating an entirely new development
cycle for these agents as they’re starting to experiment.
I’d love to hear how are you managing across your teams,
across these experiences, particularly on are we getting the outcomes
How do we know how to improve things?
How have you thought about that at OpenTable?
I mean, when we started, it was definitely more top-down
of, hey, here are the initiatives we want to go after.
At this point, we’ve tried
to democratize it to some extent where we said, look,
as an executive team or as a leadership team,
figure out all the potential use cases.
You as employees know what you’re spending your time on.
And we want you to set goals
for your own teams and organizations on how you’re going to use
AI to make your processes, your workflows more efficient.
And that’s actually been, you know, pretty interesting in that
you can potentially take a lot of wrong turns.
But in this world, experimentation is key.
it helps us bring the employees along for the journey, right?
Like, okay, how are you going to use AI to make yourself more efficient
is different than us saying, here’s a new tool.
been pretty empowering for the organization to hear that message of,
you know, actually part of your job as a leader or as a person on
a team is to actually get better at AI and use AI
You know, if you’re a recruiter and you’re
thinking about how do I more effectively source candidates or,
you know, filter through applications, if you’re a marketer saying,
the right influencer on Instagram or TikTok so I can run a campaign?
actually, we start seeing bits and pieces bottom-up being built out.
Like, hey, I built a small agent workflow using the tools you gave us.
Or, you know, the development budget you gave us.
And it’s been actually pretty phenomenal.
And with the right controls and governance mechanisms, etc.
and governance mechanisms, but like different problems need different
governance and controls, right?
So like, what does OpenTable do?
We match diners to restaurants, right?
I’m not building a self-driving car.
I’m not making medical recommendations.
I think we can take a little bit more risk as an organization.
And that’s this thing I try to tell my teams is like,
but we can get feedback and learn
and improve it, it’s better than the alternative.
And so you just have to put that into context of like,
what is the actual risk in what we’re doing and in our products.
And, you know, can we take more risk?
And I think you’ve kind of led to my last question here, which is
when you think about the future of OpenTable as an organization,
you know, from Salesforce’s vantage point, we believe
customers are going to have lots of different kinds of agents: agents
facing their customers, agents facing their employees,
and then a whole class of agents
that are just doing kind of work in the background,
you know, kind of at scale work.
How do you think about your agentic workforce?
How are you all planning for that?
What does that future look like for you?
Yeah, and it’s interesting because we’re a hospitality company.
So we bring people together, right?
And that human touch is really important.
That said, agents are here and they’re here to stay.
One thing I think about is,
how do I take my human workforce and help it be
the people who are training and building out
the digital workforce, and really involving them in the journey.
I have like seven or eight different coding copilots in trial
And, you know, for some people like me, well, that’s a lot.
Do you need all of those?
It’s like, look, I want to engage with my engineers.
I want them to try and test different things and find out what works,
Because ultimately, you know, your most valuable employees
are going to be the ones who learn how to use AI very effectively.
And, you know, I want to empower them to do so.
Well, Sagar, thank you so much for spending some time with us.
All right. All right. Good to see you. Thank you all. Appreciate it.
words in the world that we know are “What if?”
And today thousands of organizations are saying
“What if” with a power to act on those words
What if you could empower
every employee with a limitless, digital workforce?
One that listens, learns, and acts all day, every day?
Well, this isn’t some far-fetched science fiction fantasy anymore – no.
It’s happening right now – with AI from Salesforce.
Agentforce turns your vision into reality.
Think about it – a future where every student, client, citizen,
and customer experiences the best service possible.
Well, now they're free to focus on the work that matters most.
All you have to do is bring your dreams.
