At Dreamforce, Paul Tatum, executive vice president of global public sector solutions at Salesforce, joins theCUBE’s Gemma Allen and George Gilbert to unpack how AI and agents are moving from experiments to mission impact in government. Tatum explains why agencies are leaning in responsibly – spurred by recent U.S. AI directives and RFPs/RFIs that now mandate AI capabilities – while still insisting on standards, compliance and security. He shares real-world citizen-service pain points (think CMS.gov/IRS.gov “help” pages and 4,000-page Medicare/Medicaid policy PDFs) and how Agentforce plus large language models can read policies, ground responses and augment overworked civil servants for faster, more consistent adjudication, always with a human in the loop.
The discussion explores Salesforce’s approach to scaling the “agentic enterprise” in the public sector: start from existing systems of record on Salesforce and surround them with agents, use Data Cloud’s “zero-copy” data fabric to harmonize siloed information without multi-year migrations, and prioritize trust and governance with FedRAMP accreditation covering the full Agentforce stack, Data Cloud and the trust layer, including SLAs for vulnerability remediation. Tatum predicts the government will become the largest user of agentic technology because public policies are published and machine-readable, making this a compelling arena to measure what’s production-ready, how adoption happens internally vs. externally, where approvals remain essential and where early automation can relieve backlogs before tackling edge cases.
Forgot Password
Almost there!
We just sent you a verification email. Please verify your account to gain access to
Dreamforce 2025. If you don’t think you received an email check your
spam folder.
In order to sign in, enter the email address you used to registered for the event. Once completed, you will receive an email with a verification link. Open the link to automatically sign into the site.
Register for Dreamforce 2025
Please fill out the information below. You will receive an email with a verification link confirming your registration. Click the link to automatically sign into the site.
You’re almost there!
We just sent you a verification email. Please click the verification button in the email. Once your email address is verified, you will have full access to all event content for Dreamforce 2025.
I want my badge and interests to be visible to all attendees.
Checking this box will display your presense on the attendees list, view your profile and allow other attendees to contact you via 1-1 chat. Read the Privacy Policy. At any time, you can choose to disable this preference.
Select your Interests!
add
Upload your photo
Uploading..
OR
Connect via Twitter
Connect via Linkedin
EDIT PASSWORD
Share
Forgot Password
Almost there!
We just sent you a verification email. Please verify your account to gain access to
Dreamforce 2025. If you don’t think you received an email check your
spam folder.
In order to sign in, enter the email address you used to registered for the event. Once completed, you will receive an email with a verification link. Open the link to automatically sign into the site.
Sign in to gain access to Dreamforce 2025
Please sign in with LinkedIn to continue to Dreamforce 2025. Signing in with LinkedIn ensures a professional environment.
Are you sure you want to remove access rights for this user?
Details
Manage Access
email address
Community Invitation
Paul Tatum, Salesforce
At Dreamforce, Paul Tatum, executive vice president of global public sector solutions at Salesforce, joins theCUBE’s Gemma Allen and George Gilbert to unpack how AI and agents are moving from experiments to mission impact in government. Tatum explains why agencies are leaning in responsibly – spurred by recent U.S. AI directives and RFPs/RFIs that now mandate AI capabilities – while still insisting on standards, compliance and security. He shares real-world citizen-service pain points (think CMS.gov/IRS.gov “help” pages and 4,000-page Medicare/Medicaid policy PDFs) and how Agentforce plus large language models can read policies, ground responses and augment overworked civil servants for faster, more consistent adjudication, always with a human in the loop.
The discussion explores Salesforce’s approach to scaling the “agentic enterprise” in the public sector: start from existing systems of record on Salesforce and surround them with agents, use Data Cloud’s “zero-copy” data fabric to harmonize siloed information without multi-year migrations, and prioritize trust and governance with FedRAMP accreditation covering the full Agentforce stack, Data Cloud and the trust layer, including SLAs for vulnerability remediation. Tatum predicts the government will become the largest user of agentic technology because public policies are published and machine-readable, making this a compelling arena to measure what’s production-ready, how adoption happens internally vs. externally, where approvals remain essential and where early automation can relieve backlogs before tackling edge cases.
At Dreamforce, Paul Tatum, executive vice president of global public sector solutions at Salesforce, joins theCUBE’s Gemma Allen and George Gilbert to unpack how AI and agents are moving from experiments to mission impact in government. Tatum explains why agencies are leaning in responsibly – spurred by recent U.S. AI directives and RFPs/RFIs that now mandate AI capabilities – while still insisting on standards, compliance and security. He shares real-world citizen-service pain points (think CMS.gov/IRS.gov “help” pages and 4,000-page Medicare/Medicaid policy...Read more
exploreKeep Exploring
What is the current approach of government regarding the use of AI in citizen services?add
What is the current stance of governments regarding the adoption and integration of AI technologies?add
What is the current requirement for government RFPs and RFIs regarding AI capabilities?add
What challenges does the government face regarding data when adopting new technology, and how does Salesforce address these challenges?add
>> Hi, and welcome back to Dreamforce 2025 Live here from the Moscone Center in San Francisco where the energy is absolutely palpable. I'm Gemma Allen, your host here in theCUBE, joined by my co-host and one of our wonderful analysts, George Gilbert. And joining us today we have Paul Tatum, who is EVP of Solution Engineering for Global Public Sector at Salesforce. Welcome, Paul.
Paul Tatum
>> Yeah. Thank you for having me. It's great to be here. We're so excited about Dreamforce. Tons of energy, tons of customers, so thanks for having me today.
Gemma Allen
>> Well, we're so happy to have you. So we had a good chat there off camera about public sector, but when we think AI and government, should that make us nervous, or excited? Tell us what that means in real life.
Paul Tatum
>> It makes me excited. I think it should make us all excited. We all have been working with government organizations as citizens for our entire lives, whether it be getting your birth certificate, your driver's license, your passport, et cetera. And typically, it hasn't been a great experience. I think we all have our stories of the DMVs and et cetera. And I think if you look at where AI is going, where Salesforce and Agentforce is going, like the government could use some help in this space of citizen services, engaging responsiveness. The nice thing about government is they are very responsible with our data, with the approach to technology. Some might call it slow, methodical, but they are careful and they do have standards and compliance and security requirements that we have to abide by at Salesforce as a company, the industry does. So I'm really confident that government is taking it seriously. They're leaning into it, but they're doing it in a responsible way.
Gemma Allen
>> Tell us a little bit about the day-to-day, the cultural transition. We all know AI is happening and what feels like the speed of sound. And government hasn't necessarily been famous or known for its agility in tech adoption. If we think about the cloud era, first AI era, it's a fundamental shift change in speed. And I think really also in terms of that feeling of existential crisis, right?
Paul Tatum
>> Yeah.
Gemma Allen
>> So tell us about what sort of conversations you're having with your customers.
Paul Tatum
>> It always starts with AI now. So it's interesting. I'm based in Washington, D.C. I obviously cover our federal customers, but also state and local and global government. And what you're seeing is governments ultimately are pretty risk averse. We feel that in the industry. What we saw here in the United States with this administration is the AI action plan that came out a few months ago, which was very, very kind of a bold statement for the United States as a country to say, "Listen, we need to really, really double down on AI at the infrastructure level, at the chip level, at the enterprise software level. And oh, by the way, as government organizations, we need to do more with AI."
I think in the past it had been a little bit of an experiment. "Hey, let's go try it in the corner." Today, every RFP, every RFI that comes out of government has a AI component. You must deliver this particular capability, and there has to be AI capacity and capability within it. So government has done a lot of really good things over the years, all the way from introducing the internet to us that we all enjoy today, so there is innovation that's happening, and I think they see what's happening in the private sector. We need to deliver the same level of service using AI that the commercial and private industry is going to do, so they're definitely leaning in.
Gemma Allen
>> Absolutely.
George Gilbert
>> So maybe, Paul, help us unpack when these RFIs come out and they say, "We want AI in it." Well, I've talked to some of your colleagues like Madhav, who's the COO of Agentforce product, and he says when he hears from a CEO or a CIO, we got to do something with agents or we got to put agents in the experience. It's a red flag. That a better signal of likely success is when the request comes down, I have a process that I want to re-engineer and an outcome that I want to change, and even like a magnitude by which I want to change it. I want to onboard a customer in 50% of the time, that sort of thing. Is government still in the, "I need to experiment with AI," or are there some departments, functions, whatever, where they are thinking, "Okay, I know how to prioritize a process. I want to re-engineer first?"
Paul Tatum
>> So I think they're on the journey to the second. It's kind of funny. I joke with people and my customers, if you type agentic into your mobile phone, your iPhone, it shows up as being misspelled. That's how new this whole agentic era is. A year ago, the spell checkers still haven't caught up with it. And so government is in the process of going from, "Let's play with it, let's kind of see what it is." To today, basically 18 months later saying, "This is for real. This is here to stay, and it's transformational."
So what that is showing up in these RFIs, RFPs is let's go build the new systems. It is taking what is frankly mountains of a workload, overworked great civil servants that are serving us as the public and as citizens, and trying to augment their work and help them. And so I'll give you a perfect example. If you were to go to the CMS.gov website or IRS.gov and hit the help button up in the corner of their website, which I did this morning, there are over 200 URLs, little blue URLs to go out to various articles around the tax code. Not super helpful. You're trying to figure out where to go, what to do. The day will come where there will be an agentic helper on that webpage, right? But that takes time, right? You got to test it, you got to secure it. If you're trying to help your mother, grandmother with Medicaid, Medicare, there are 4,000 pages of PDF that are the Medicaid/Medicare policies. It's like, "Mom, should we get reimbursed for this?" How are you going to know that? So what's really interesting with this technology, Agentforce or LLMs, they're really good at reading these PDF documents, bringing them into the RAG models, producing them out into Agentforce to have a conversation either with the CMS employee, the Medicaid/Medicare person that's saying, "Should we reimburse this hospital? Should we reimburse?"
All of a sudden, this Agentforce agent is now the specialist on those policies. And I, as the employee, don't go, "Oh shoot, I have to memorize 4,000 pages of tax code or Medicaid and Medicare code." The agent can help advise me on that and really help me adjudicate the answer.
George Gilbert
>> And is Salesforce's advantage in that organization is already in Salesforce and the structured data and increasingly the unstructured data is in data cloud, so it's all ready to be served up naturally by the agents?
Paul Tatum
>> Exactly right. So we have thousands of government customers around the world and in Washington D.C. They started with us to do all kinds of programs, whether it be grants or call centers, contact centers, you name it, they have done it with Salesforce. Now, we have an opportunity to go back and say, "You have a system of record using Salesforce, now let's surround it with agents. Let's put the agentic enterprise in place for your organization."
And we've been a big believer throughout our entire existence over 26 years of making the complexity of technology simple for our users, out of the box, simplicity, low code, build an agent in a matter of minutes. You can go here at Dreamforce and go build one, and then really start to transform government. We want a more responsive government. We want to be part of that at Salesforce, and we believe we can do that with Agentforce.
Gemma Allen
>> Well, you've certainly given two examples that speak to me, I have to say, because I've spent more hours than I want to will ever get back at the DMV in New York. And as a European as well living in the US, navigating health insurance is another minefield, so I can totally see the value. In terms of the bottleneck, so you mentioned data, and we know that there is a lot of challenge around legacy systems and around that last mile of excellence that can sometimes be where I guess benefits are won and lost for industries as well that maybe don't have the same capability or the same even spend as some of the larger tech players, et cetera. So from your perspective, how do you think about that? How do you think about that real last mile of adoption? Where does Salesforce play in that space?
Paul Tatum
>> So you kind of touched on two topics. You touched on data and adoption. So the data piece of it, up until very recently, the data piece, especially for government, is a major stumbling block. So you figure they save everything. They have basements, they have mountains full of data, paper, electronic. And whenever you would come up with a new idea, new technology idea, they would go, "That sounds great, Paul, but what about all of that data? Do we have to move it? Do we have to massage it?" And historically, the answer was kind of yes, which was kind of a stumbling block. So several years ago, we embarked on trying to deliver a data fabric. We call it data cloud. You guys are familiar with that if you've been at Dreamforce here today, and what that has really done is provided a data fabric for our customers to keep the data in place, do a zero copy to allow those agents at agentic layer to really have the data they need to fuel them, because the answers to conversations are only as good as the data they have access to, and government is not going to bring all those silos of data together. That would be a multi-decade project. The days of like MDM, like trying to get five organizations to agree on the data format, bring it all together so everybody can share it and use it are gone. That will never happen. So we're very bullish about data cloud and its ability to harmonize and aggregate the data in a zero copy fashion to feed the agentic player. So that's the data part. Now, the adoption part I think is a two-fold. There's internal adoption for the government employee going, "Hey, what's this new buddy I have? What's this new agent that I'm working alongside to help me understand the tax code?" I don't think that will be very hard to adapt to because, basically, here's what's going to happen in my belief, you submit for a tax return, tax refund. In fact, this happened to me. I'll tell you a little story. So I had a little mix-up on my 2023 taxes, got sent to the international department for some reason. I called, "Hey, I'm supposed to get my tax return." Very nice IRS employee. "Oh, Paul, yes, I see it here. I don't know why it ended up in this department." I'm like, "Can you send it back to the normal department?" "Oh, no, I can't do that." I'm like, "Well, when can I see my tax refund?" "Oh, I'm sorry, we're working on 2022s. You'll see yours in about 18 months." And there's nothing I could do about it. Now, you're wondering, did I ever get it? I just got it two years later. So that IRS employee is overworked, right? There's 300 million Americans in the country, taxpayers, et cetera. And so when the technology shows up for agentic assistant that is adjudicating tax returns and saying, "Yes, no, maybe. Here's my recommendation. I've read the tax code, I've looked at the return. Here's what I think should happen." I can come in the morning. That thing can run through a thousand of them overnight. I can come in and start to kind of go, "Yeah, yes, no, maybe, I agree. Paul needs more information, et cetera." So I think the internal will be great. The external is where you and I interface with government. That website I just described that has 200 URLs of data. I want to have a conversation with the data. I want to have a conversation about the policy. I don't want to search through dozens of webpages to get the answer. So it will take a little time, but it is inevitable. It's too alluring. It is too impactful to ignore for government.
George Gilbert
>> What is the role then? How does the IRS, let's say, tax refund process work in that world? Will the agents learn progressively more of the edge cases, so they go from handling 10% to handling 80%? And then the IRS examiners, there can be more focused on the really weird edge cases, or they can go into doing more audits, spot audits?
Gemma Allen
>> I hope not.
Paul Tatum
>> I'm going into the audit side of it, but I will say, I think the technology in general is designed right now and its capabilities to do the simple, easy stuff and allow the human employee, whether it be government or private industry to pursue the more complex things, right? The edge cases, maybe there's not a policy about this. Maybe I need to kind of use multiple data sources, get more information, et cetera. So I think that what we'll see is the early and simple ... I call it adjudication. My other prediction is government will be the largest user of agentic technology of any industry by far. And you may like, "Well, Paul, why would that be?" If you think it, everything the government has ever done or needs to do has been published. Private industry, they hide behind their firewalls on their policies. But the policy for taxes, the policy for health insurance is you can all find it, and the agents are extremely good at reading it, understanding it, and then making or helping augment the human around decisions for it. So why wouldn't this industry embrace it?
Gemma Allen
>> I guess the one thing too about government is there's a lot of governance. One of the reasons why it's probably such a great battleground for AI is it's so procedural. But with that too comes an expectation around governance and security. How are those conversations playing out? Especially when you think about a world where you go from work automation to AI orchestration for government employees? That must incite a level of risk or at least a sense of risk.
Paul Tatum
>> It's definitely a public-private partnership between the government industry to deliver trustworthy, accurate AI. The first thing the government did, which has been fantastic years ago, like here in the US, they've established the FedRAMP program, which is the security aspects of cloud computing for government. Salesforce made a big investment to bring the entire Agentforce stack through that accreditation process. We were awarded that in July of this year. Took a little while, because we took the whole stack through there. We took the data cloud through there. We took Agentforce through there. We took the trust layer through there. None of our competition, no one else in the industry has taken the entire platform, the Agentic platform through that very stringent government requirement, and we had to maintain that. There are SLAs for remediating, vulnerabilities, et cetera, that we are required to do by government standards. So the government has set up a nice standard. It's very clear for the security compliance of the stack. And then there is what is appropriate use. Where should we be using it? And government is starting to develop those policies, it's moving very quickly about what's appropriate. You always want a human in the middle of it to say, "Should we say yes or no? Here's my recommendation." But ultimately, I think there needs to be a government employee involved in the final decision.
Gemma Allen
>> Great. Well, Paul, it's been wonderful to have you on. Thanks so much for chatting with us today. I think we've all learned a lot about Salesforce's mission-oriented value towards government. And I think as a taxpayer, it's certainly something that I welcome, so thanks so much.
Paul Tatum
>> Well, thank you so much. I really appreciate the questions, the conversation. Thanks for what you all do to get the word out on behalf of Salesforce and Agentforce in government, appreciate it.
Gemma Allen
>> Great. I'm Gemma Allen here live in San Francisco at Dreamforce 2025. We'll be right back with our next guest.
Gemma Allen
George Gilbert