1
00:00:00,000 --> 00:00:03,360
What is coming next that state leaders should be watching for?
2
00:00:03,360 --> 00:00:07,040
I don't think there's much doubt that what's coming next is a genetic AI.
3
00:00:07,040 --> 00:00:13,120
This is Real State, a podcast from CSG South.
4
00:00:13,120 --> 00:00:19,680
Join us as we talk with policymakers and public leaders shaping state-level decisions across the
5
00:00:19,680 --> 00:00:25,040
South and explore how ideas move from policy to real impact.
6
00:00:25,040 --> 00:00:30,320
[MUSIC]
7
00:00:30,320 --> 00:00:35,280
Hello and welcome to Real State. I'm your host, Lindsey Gray. Back in July,
8
00:00:35,280 --> 00:00:40,160
we recorded some real-state episodes while at the Southern Legislative Conference,
9
00:00:40,160 --> 00:00:45,600
where state leaders from across our 15 state regions and beyond are discussing the issues
10
00:00:45,600 --> 00:00:51,680
that will shape the future of state government. And few issues are evolving faster than artificial
11
00:00:51,680 --> 00:00:58,800
intelligence. For states, the conversation around AI has quickly moved from beyond the basics.
12
00:00:58,800 --> 00:01:06,240
Leaders are now asking what tools are most effective? What problems can AI help us solve?
13
00:01:06,240 --> 00:01:13,280
How do we manage the risk and how do we prepare for what's coming next? AI, including
14
00:01:13,280 --> 00:01:21,840
generative AI, a genetic AI, and machine learning is now ranking as one of the top priorities for state CIOs.
15
00:01:21,840 --> 00:01:28,320
But successful adoption isn't just about deploying new technology. It requires thoughtful decisions
16
00:01:28,320 --> 00:01:34,320
around governance, security, data, workforce readiness, and public trust.
17
00:01:34,320 --> 00:01:40,080
Joining me today is Doug Robinson, Executive Director of NASCIO,
18
00:01:40,080 --> 00:01:46,560
the National Association of State Chief Information Officers. Doug works with CIOs
19
00:01:46,560 --> 00:01:53,680
across the country and brings a unique perspective on where states are today and where this technology
20
00:01:53,680 --> 00:01:59,280
is heading. He was also gracious enough to lead an awesome session at the Southern Legislative
21
00:01:59,280 --> 00:02:04,720
Conference this year. Doug, welcome to Real State. Pleasure to be here, Lindsey. Thank you for the
22
00:02:04,720 --> 00:02:11,280
invitation. Doug, you have a unique vantage point where you're talking with technology leaders
23
00:02:11,280 --> 00:02:17,680
across the country. When you look at where states are today with AI adoption, how would you describe
24
00:02:17,680 --> 00:02:25,360
this moment? Are we still in the experimentation phase or are states beginning to think about AI as
25
00:02:25,360 --> 00:02:33,600
a long-term transformation strategy? Certainly beyond experimentation. I've characterized
26
00:02:33,600 --> 00:02:41,840
2026 as the year of adoption. So we've moved for the most part, and again, I'm generalizing
27
00:02:41,840 --> 00:02:47,040
from the states, our motto at NASCIO, if you've seen one state, you've seen one state. So I have to
28
00:02:47,040 --> 00:02:54,000
look at kind of unevenness, but for the most part, about 80% of the CIOs report that they already have
29
00:02:54,640 --> 00:03:01,920
AI solutions, generative AI solutions in production. So I would say beyond experimentation and cautiously
30
00:03:01,920 --> 00:03:09,360
moving. So it's kind of a combination of adoption and kind of risk management. What are you hearing
31
00:03:09,360 --> 00:03:17,840
from state CIOs? And are they excited, cautious, or a bit of both? A little bit of both. I think
32
00:03:17,840 --> 00:03:25,600
there's certainly a lot of excitement about the possibilities. We have not seen a large percentage of
33
00:03:25,600 --> 00:03:34,000
these pilot projects move to scale and by that, I mean beyond an agency or beyond a certain line of
34
00:03:34,000 --> 00:03:39,680
business or a project. So when we look at our data, they're fairly specific in what I'll call kind
35
00:03:39,680 --> 00:03:46,080
of the functional domains of government, health and human services, transportation, tax and revenue,
36
00:03:46,080 --> 00:03:52,160
those are the large ones. But in terms of scaling across the entire executive branch, those have
37
00:03:52,160 --> 00:03:58,480
been limited to a lot of the productivity tools, but not something that we're saying. So I think that's
38
00:03:58,480 --> 00:04:05,760
kind of the cautious nature of what their role is, part of their role is risk management, regardless. So
39
00:04:05,760 --> 00:04:13,360
I think they're looking at the exciting opportunities, transforming how states serve their citizens,
40
00:04:13,360 --> 00:04:19,840
but also doing that cautiously because you just can't let the beast loose without those guard rails.
41
00:04:19,840 --> 00:04:26,560
A lot of attention goes to the future potential of AI. But I want to talk about what is available
42
00:04:26,560 --> 00:04:33,680
today? What AI tools or capabilities are already mature enough that states can realistically
43
00:04:33,680 --> 00:04:40,640
begin leveraging them? Yeah, so that's a good question because it's again mixed, but I'll generalize for
44
00:04:40,640 --> 00:04:47,120
the most part what's widely available are the productivity tools that are already built into
45
00:04:47,120 --> 00:04:54,480
collaboration platforms. So things like Microsoft, co-pilot and Google Gemini, those are widespread. Other
46
00:04:54,480 --> 00:05:00,960
states have direct through CIO directives are an executive order by the governor. They've deployed
47
00:05:00,960 --> 00:05:08,000
other tools like chat DPT and Clawed and others throughout the executive branch for employees to
48
00:05:08,000 --> 00:05:14,080
utilize. So those are widespread. Again, what is more limited are the more specific
49
00:05:14,080 --> 00:05:19,920
applications within a line of business. So you have to recognize that the states aren't, they're not
50
00:05:19,920 --> 00:05:26,480
building AI tools. These are commercial products. That's one of the challenges in terms of acquisition
51
00:05:26,480 --> 00:05:34,560
and procurement is all of these products today are you can say putting quotes powered by AI. So every
52
00:05:34,560 --> 00:05:40,320
solution they bring in most likely already has AI embedded in it and that's kind of one of the
53
00:05:40,320 --> 00:05:45,600
challenges is making sure that you have an inventory of what's available when you look at the states that
54
00:05:45,600 --> 00:05:51,760
do have inventories. You'll see it's predominantly commercial products that are focused on document
55
00:05:51,760 --> 00:05:57,840
preparation, generation, chat bots. Those are the kind of leading use cases across all the states.
56
00:05:57,840 --> 00:06:03,040
Where do you see the biggest opportunities for AI to improve how government operates?
57
00:06:03,040 --> 00:06:09,520
Whether that's internal efficiency, citizen services, cybersecurity or decision making.
58
00:06:09,520 --> 00:06:14,720
My speculation and certainly based on the evidence out there is the greatest opportunities are to
59
00:06:14,720 --> 00:06:21,200
enhance citizen services. But that comes with the challenge of trust. And so I think that's the one
60
00:06:21,200 --> 00:06:29,680
thing that CIO's and agency leaders are looking at is they see opportunities to apply beyond things like
61
00:06:30,800 --> 00:06:36,640
responding to questions out of chat bot or doing immediate translation services much more
62
00:06:36,640 --> 00:06:44,640
transformation from end to end cybersecurity. Almost 90% of the states today are using AI in
63
00:06:44,640 --> 00:06:52,160
cybersecurity for threat detection to examine anomalies to scrub logs. Things that take a lot of
64
00:06:52,160 --> 00:07:00,000
human capacity can be done with AI. So we see that as another from cybersecurity ops. Recognizing
65
00:07:00,000 --> 00:07:08,160
that the blade swings both ways. So the bad actors are also using AI extensively to craft
66
00:07:08,160 --> 00:07:14,400
very convincing, fishing emails and to do things in social engineering that weren't available for.
67
00:07:14,400 --> 00:07:21,120
So that's certainly part of it. I think the other area that I'm really looking at, because I think
68
00:07:21,120 --> 00:07:26,880
there's tremendous opportunity there is fraud detection, fraud prevention. So in a states of
69
00:07:26,880 --> 00:07:32,960
classically done kind of what's called the pay and chase model. You've seen a lot of headline news
70
00:07:32,960 --> 00:07:38,000
over the last several years certainly since COVID about the level of fraud and public benefit programs
71
00:07:38,000 --> 00:07:44,960
and just public distribution programs. It is is significant and in the major challenge for that is
72
00:07:44,960 --> 00:07:51,280
identity management and identity verification. So I think the states are testing AI several
73
00:07:51,280 --> 00:07:59,520
state revenue and tax departments are using AI already to look at how to advance that through AI
74
00:07:59,520 --> 00:08:05,280
as opposed to some of the traditional fraud tools. So that's another I think really strong use cases
75
00:08:05,280 --> 00:08:11,120
fraud detection. Are there use cases you were seeing emerge across the states that you think could
76
00:08:11,120 --> 00:08:17,040
become standard practices over the next several years for our states? We're capturing those with our
77
00:08:17,040 --> 00:08:23,600
regular surveys and we're just now analyzing our 2026 data to look across the word kind of to see
78
00:08:23,600 --> 00:08:30,960
where are those were the patterns of of use cases. Again, top of the list is in things like case
79
00:08:30,960 --> 00:08:38,400
management and benefits management with with cautions that ultimately its humans making a decision.
80
00:08:38,400 --> 00:08:45,120
We have a number of state legislative bodies that have already enacted legislation and the past bills
81
00:08:45,120 --> 00:08:51,600
that prohibit AI from making final things like Medicaid eligibility termination or insurance.
82
00:08:51,600 --> 00:08:58,560
However, those tools can be used to obviously streamline the massive amount of rules and
83
00:08:58,560 --> 00:09:04,720
regulations that case workers have to absorb. So we've seen good examples of that good examples
84
00:09:04,720 --> 00:09:12,240
and helping citizens navigate complex bureaucratic structures because they don't understand the rules
85
00:09:12,240 --> 00:09:19,120
of the road and so a lot of those business rules can be embedded in in AI some very fascinating things
86
00:09:19,120 --> 00:09:26,800
and natural resources some very discrete kind of use cases of things like predicting the
87
00:09:26,800 --> 00:09:33,360
ovulation cycles of stripe bass. We got that one. I thought that was very interesting. So a lot of
88
00:09:33,360 --> 00:09:42,000
things California has embarked on a fairly significant effort to look at things like accidents
89
00:09:42,000 --> 00:09:47,440
and to use data historical data that again and previously this wasn't possible with the with the
90
00:09:47,440 --> 00:09:54,320
capabilities and the compute powers to look at 10 years of historical accident data and make
91
00:09:54,320 --> 00:09:59,200
a some type of revelations around the causal factors of that. So that's what they are doing. They
92
00:09:59,200 --> 00:10:05,200
have two major initiatives around vehicular and pedestrian accidents and they are even embarked
93
00:10:05,200 --> 00:10:11,040
on that where they are looking at terabytes and terabytes of data that has been going to put
94
00:10:11,040 --> 00:10:17,040
together for AI analysis. So a lot of good things in those. Which you I think if you look at
95
00:10:17,040 --> 00:10:24,480
and just think about what are the major the functions of state government and the major large
96
00:10:24,480 --> 00:10:28,240
agencies that's where you're going to find a lot of activity because they're trying to reduce
97
00:10:28,240 --> 00:10:36,000
the overall burden but I think certainly the promise in the future is enhanced citizen services
98
00:10:36,640 --> 00:10:44,240
with with AI tools. Many states have launched AI pilots or created task forces to study potential
99
00:10:44,240 --> 00:10:50,400
applications but moving from pilot project to something that truly changes how government works
100
00:10:50,400 --> 00:10:57,360
can be challenging. So what separates states that are successfully implementing AI from those that
101
00:10:57,360 --> 00:11:03,440
are still trying to determine where to start? That's the question on the table today. Certainly as we've
102
00:11:03,440 --> 00:11:11,040
seen the states that have moved that if you look at the data from a year ago 90% of our states reported
103
00:11:11,040 --> 00:11:18,880
that they had pilot projects and only about 30% said they had moved those into production. So fast
104
00:11:18,880 --> 00:11:26,960
forward to 2026 our latest data indicates about 80% of them report that they've moved these pilots
105
00:11:26,960 --> 00:11:33,200
into production. Now some of them will not succeed. They will find issues with those but for the most
106
00:11:33,200 --> 00:11:41,040
part they're reporting that I think strong governance strong oversight, strong measurable impact and so
107
00:11:41,040 --> 00:11:48,080
be basically measuring along the way to see how these things are operating because they advance and
108
00:11:48,080 --> 00:11:53,920
evolve so quickly you have new models that come out all the time and those new models obviously can
109
00:11:53,920 --> 00:12:00,160
impact the direction of the project. So they're looking to stabilize those governance is a big part
110
00:12:00,160 --> 00:12:06,320
of that. The other certainly most defining challenge of all states is data quality and data integrity.
111
00:12:06,320 --> 00:12:14,560
I always remark years ago CIO on our board said to me, "Doug, for a bad data I wouldn't have any at all."
112
00:12:14,560 --> 00:12:21,040
And so that's one of the challenges is how do you make sure that you are ingesting high quality
113
00:12:21,040 --> 00:12:28,240
reliable data into your models that you're using to make those decisions? And I think the third
114
00:12:28,240 --> 00:12:34,560
thing which is a touchstone for all of this is workforce readiness. So we're seeing a lot of states ramp
115
00:12:34,560 --> 00:12:39,920
up their capabilities and in the workforce they're combining it with their cyber security training
116
00:12:39,920 --> 00:12:45,840
perhaps. Some cases they're making it mandatory if you're using an AI tool then you're required to
117
00:12:45,840 --> 00:12:50,800
receive some type of training and that training is going to focus on responsible and ethical use.
118
00:12:50,800 --> 00:12:57,360
And so that's a that can be a major issue. There will be missteps. I am confident that something
119
00:12:57,360 --> 00:13:03,120
will happen and states will learn from that. But and right now it's they're moving ahead but they're
120
00:13:03,120 --> 00:13:09,120
being cautious. Are there foundational investments whether that's cloud modernization, workforce
121
00:13:09,120 --> 00:13:14,240
training or governance that states need before they can fully take advantage of AI?
122
00:13:14,240 --> 00:13:18,880
Yeah well I've mentioned that governance is certainly at the top of the list which is those
123
00:13:18,880 --> 00:13:26,640
is the shared kind of decision model with in my area certainly of focus is the only the executive
124
00:13:26,640 --> 00:13:34,960
branch. So you have to have stakeholder engagement. I think you you bringing in outside experts as
125
00:13:34,960 --> 00:13:38,720
you have many states that are they're not only bringing in industry but they're bringing in the
126
00:13:38,720 --> 00:13:44,480
university. They have coupled their projects with university efforts. Mississippi is a good example
127
00:13:44,480 --> 00:13:50,320
of that. They have a project called Maine which is the Mississippi AI network and it's a combination
128
00:13:50,320 --> 00:13:57,280
of state government and the private sector and industry and higher education that are that are
129
00:13:57,280 --> 00:14:02,240
involved in that. I think that's kind of a good collaborative to the matter of model to look at
130
00:14:02,240 --> 00:14:08,400
other states have done very similar. So that's that's that's the challenge is how to how to kind of
131
00:14:08,400 --> 00:14:16,080
navigate all of that. Procurement is is another piece of the acquisition cycle is another kind of
132
00:14:16,080 --> 00:14:22,000
foundational aspect you have to get right. And those that comes with some challenges because
133
00:14:22,000 --> 00:14:29,920
states today don't have that maturity in how to require things like disclosure from you know all
134
00:14:29,920 --> 00:14:35,680
their vendors during association. So they're going to have to kind of work through you know all of that
135
00:14:35,680 --> 00:14:41,280
because there's tremendous liability if they if they don't workforce readiness as I've already
136
00:14:41,280 --> 00:14:46,320
mentioned in a couple of times is another one of those foundational pieces when you look at some of
137
00:14:46,320 --> 00:14:52,880
the challenges not not is not not not I think a couple of policy areas that have not been given
138
00:14:52,880 --> 00:15:01,280
enough exposure one is things like public records laws right so how do you make sure that you're
139
00:15:01,280 --> 00:15:07,200
aligned when you're using these tools to the existing open records laws that are in the state and
140
00:15:07,200 --> 00:15:12,560
for almost every state it's very similar record as a record regardless of physical form or
141
00:15:12,560 --> 00:15:20,240
characteristics. So for the states AI results AI prompts and AI results used by state agencies are
142
00:15:20,240 --> 00:15:25,440
part of the official public record but the majority of the states today are not handling it in that
143
00:15:25,440 --> 00:15:30,080
way. So they're going to have to advance the maturity of that not surprisingly technology is
144
00:15:30,080 --> 00:15:35,760
always way out in front of the policy framework and in the policy foundations in most states. So
145
00:15:35,760 --> 00:15:41,600
they always have to speed those up and move quickly. So let's stick with the governance theme.
146
00:15:41,600 --> 00:15:48,800
Nassio has highlighted AI governance as a major priority so this technology touches every agency
147
00:15:48,800 --> 00:15:54,960
every part of government so given that from your perspective what does a strong governance model for
148
00:15:54,960 --> 00:15:59,440
state governments look like? Well when you think about you think about governance particularly in a
149
00:15:59,440 --> 00:16:06,880
technology space there are certainly lots of different opinions about architecture essentially the
150
00:16:06,880 --> 00:16:14,800
the the the the blueprint of government as we describe it and so you have to be able to have a
151
00:16:14,800 --> 00:16:23,120
venue, a policy process, a collaborative model to have those shared decision rights and if you
152
00:16:23,120 --> 00:16:30,960
look at the state today, overall majority of them there is a directive of some type of coming from
153
00:16:30,960 --> 00:16:36,800
the executive branch and from the state CIO which outlines what what is their AI framework? What
154
00:16:36,800 --> 00:16:42,160
is the responsible use? What are the ethical guidelines? What are the guardrails that you have to adhere
155
00:16:42,160 --> 00:16:49,520
to? A simple example would be employees are not permitted to use personally identifiable citizen
156
00:16:49,520 --> 00:16:56,160
information right when they're using these tools they can't they can't train the models using state
157
00:16:56,160 --> 00:17:00,720
data that's kind of a guardrail it's across all states so they're all very common they're they're
158
00:17:00,720 --> 00:17:06,400
borrowing from each other and looking out to success and so you know strong governance looks at that
159
00:17:06,400 --> 00:17:11,520
that collaborative approach and shared decision it can't be the big hammer type of authority that
160
00:17:11,520 --> 00:17:17,520
we've seen and it can't be the rigid you know you cannot use these tools it has to recognize that
161
00:17:17,520 --> 00:17:22,800
these tools are in place they're now ubiquitously available and there's got to be a flexibility
162
00:17:22,800 --> 00:17:31,520
around those around those use cases so it has to be taken seriously this is a a transformational
163
00:17:31,520 --> 00:17:38,160
technology in terms of its use but it can also inherit risks and so I think that's the role of
164
00:17:38,160 --> 00:17:45,440
of strong governance is having those conversations and adopting a rubric that works in in that in
165
00:17:45,440 --> 00:17:51,520
that state depending on their operating model so when you talk to state leaders what are the biggest
166
00:17:51,520 --> 00:17:58,560
concerns you hear around AI adoption they don't know what they don't know and I think that's the
167
00:17:58,560 --> 00:18:06,400
the major concern is you know they they they expect surprises and they're trying to mitigate those
168
00:18:06,400 --> 00:18:12,960
and so without understanding the complexities and again you even some of the data scientists and
169
00:18:12,960 --> 00:18:20,320
developers don't understand how these models are evolving and the constant flux in the marketplace
170
00:18:20,320 --> 00:18:25,600
and also the constant changing there's new new models and new capabilities that seems like coming
171
00:18:25,600 --> 00:18:32,080
out every week and so it can be very confusing because the states are trying to kind of put a
172
00:18:32,080 --> 00:18:38,800
stake in the ground at a point in time and it's constantly it's constantly moving and it creates
173
00:18:39,360 --> 00:18:44,480
you know some some friction perhaps and in whether it's because there might be some
174
00:18:44,480 --> 00:18:51,120
agency that really wants to move quickly but there's a level of uncertainty there about
175
00:18:51,120 --> 00:18:58,400
you know what the consequences might be so they've been very careful to test things go to the sandbox
176
00:18:58,400 --> 00:19:04,000
reevaluate and I think that's a big part is constant and continuous evaluation and some states
177
00:19:04,000 --> 00:19:09,440
the others report to me that they're not doing that very well yet because of the magnitude of that
178
00:19:09,440 --> 00:19:15,840
there's certainly as you know winzie a lot of hype around AI right it's at the higher that's kind of
179
00:19:15,840 --> 00:19:20,320
high hype cycle but you've got a temper that with the reality of the role of government
180
00:19:20,320 --> 00:19:28,400
and also the fact that look at the latest results from pew around trust in AI 70% of those
181
00:19:28,400 --> 00:19:35,840
respondents certainly US citizens said they don't trust AI 70% so they don't trust AI and yet we
182
00:19:35,840 --> 00:19:41,760
have state government agencies and leaderships trying to deploy AI across you know all matter
183
00:19:41,760 --> 00:19:47,920
of state services and that creates some some challenges and direct conflicts with what citizens want
184
00:19:47,920 --> 00:19:56,240
to see taking the matter of citizen data protection a step deeper states have a responsibility to
185
00:19:56,240 --> 00:20:03,440
protect citizens information and maintain trust how should leaders think about privacy transparency
186
00:20:03,440 --> 00:20:11,200
and ethical use as AI becomes more integrated in state governments again a broad constituency of
187
00:20:11,200 --> 00:20:17,360
collaborators internally so a specific example of we host the community of practice of the state chief
188
00:20:17,360 --> 00:20:21,760
privacy officers we also heist the state chief information security officers those are
189
00:20:22,640 --> 00:20:29,280
typically part of the CIO organization but not always and we are we we always recommend that the
190
00:20:29,280 --> 00:20:35,520
chief privacy officer be fully engaged in those decisions in fact in a couple of states the CPO
191
00:20:35,520 --> 00:20:43,120
is actually leading their AI effort so we now have over half the states that have a chief AI
192
00:20:43,120 --> 00:20:48,400
officer already appointed and many more that are operating in that role not without equivalent
193
00:20:48,400 --> 00:20:54,480
title but they they operate in support of that so we're seeing the growth of another chief physician
194
00:20:54,480 --> 00:20:59,920
but with tell them that they they have to have communications together and that's a big part of
195
00:20:59,920 --> 00:21:06,960
that to make sure that they are following the appropriate privacy there's no federal privacy rules
196
00:21:06,960 --> 00:21:11,680
there's no federal privacy legislation so each state have crafted their own and some of them
197
00:21:11,680 --> 00:21:17,760
you know are are very restrictive in terms of what the states can do so they have to make sure
198
00:21:17,760 --> 00:21:23,760
that they're aligned with that and that again requires that they they understand the the magnitude
199
00:21:23,760 --> 00:21:31,520
of a state data states have the most amount of personal information about all of us
200
00:21:31,520 --> 00:21:36,880
than any other entity from essentially cradle to grave right they have they issue the births
201
00:21:36,880 --> 00:21:43,520
certificate so they have all that information they host the gold standard for identity which is
202
00:21:43,520 --> 00:21:49,280
the driver's license and then of course they have a whole set of transactions which they're constantly
203
00:21:49,280 --> 00:21:54,960
collecting personally that have all information from citizens and that has to be protected and
204
00:21:54,960 --> 00:22:02,400
certainly fire walled off from a lot of these AI tools AI obviously is moving very very fast
205
00:22:02,400 --> 00:22:09,520
so many leaders are just getting their arms around generative AI now here comes the agents with
206
00:22:09,520 --> 00:22:15,600
more advanced capabilities what is coming next that state leaders should be watching for I don't
207
00:22:15,600 --> 00:22:21,600
think there's much doubt that what's coming next is a genetic AI as you mentioned that is from you
208
00:22:21,600 --> 00:22:28,640
know again our recent data about the next wave what's the the most what will be the most impactful
209
00:22:28,640 --> 00:22:34,560
emerging technology in the next three years it's two thirds of our our state CO said a
210
00:22:34,560 --> 00:22:40,800
genetic AI as they see that actually as surpassing generative AI because of its capabilities
211
00:22:40,800 --> 00:22:48,320
its capabilities to have actually navigate and and and make decisions of complex business rules
212
00:22:48,320 --> 00:22:54,400
that many citizens are not surprisingly are confounded by call it navigating the pirate map
213
00:22:54,400 --> 00:23:00,000
it's very difficult to understand that so a gentigate AI and those agents can actually give them the
214
00:23:00,000 --> 00:23:05,120
tools again comes with a lot of inherent risk a lot of security risks and privacy risks but
215
00:23:05,120 --> 00:23:11,040
they see that as the next capability to use agent technology they're using it a number of states
216
00:23:11,040 --> 00:23:18,720
already using a gentigate AI internally they're using it to make a kind of simple decisions along
217
00:23:18,720 --> 00:23:25,600
along those complex business rules internally I see them opening that up two citizens in the near
218
00:23:25,600 --> 00:23:33,040
future so it is is not widespread is I think the maturity is uneven the adoptions uneven but they
219
00:23:33,040 --> 00:23:40,080
see the promise of what of what that can do and there's certainly plenty more that generative AI
220
00:23:40,080 --> 00:23:47,840
can do when you get down into get down into the large enterprise scale by that I mean cross boundary
221
00:23:47,840 --> 00:23:53,520
interoperability across multiple agencies right now that is a fairly low level and so we're going
222
00:23:53,520 --> 00:23:59,520
to see that we see that build up but I think it's exciting to look at the I guess the innovation
223
00:23:59,520 --> 00:24:08,640
that you're seeing across the states and the fact that old old the legacy modernization topic is
224
00:24:08,640 --> 00:24:14,080
always kind of the top of the list for states see I just modernization and general every state CIO
225
00:24:14,080 --> 00:24:20,160
aspires to have a modern environment these tools can help leak accelerate that but again with
226
00:24:20,160 --> 00:24:26,480
they're doing it very cautiously understanding that things that will certainly could go wrong and
227
00:24:26,480 --> 00:24:30,960
will go wrong at some point in the future is they're trying to mitigate the risk as much as they
228
00:24:30,960 --> 00:24:36,720
can and provide guidance through their policy memorandum and their direction so that'll something
229
00:24:36,720 --> 00:24:42,960
will solidify over the next several years beyond a genetic AI what other emerging tools or
230
00:24:42,960 --> 00:24:49,520
capabilities do you see that could be especially valuable for states over the next five years I
231
00:24:49,520 --> 00:24:57,440
think we're looking at the tools that were originally deployed at a fairly low level during
232
00:24:57,440 --> 00:25:04,880
during COVID one is low code no code development and there's also vibe coding which involves AI
233
00:25:04,880 --> 00:25:11,200
but I think low code no code is really bringing the capabilities of application development out of
234
00:25:11,200 --> 00:25:17,440
the software solution shop and giving it to the lines of business meaning that that state
235
00:25:17,440 --> 00:25:23,840
employees who know the business process can now use modules and kind of plug and play
236
00:25:23,840 --> 00:25:29,760
to develop solutions we saw that happen during COVID that has become certainly more widespread
237
00:25:29,760 --> 00:25:37,360
quantum certainly something that states are just now looking at we have a couple of states that
238
00:25:37,360 --> 00:25:43,680
have quantum readiness task forces and looking at that that the the danger with quantum from the
239
00:25:43,680 --> 00:25:50,720
states perspective is the security concerns and the capabilities that that actors using quantum
240
00:25:50,720 --> 00:25:57,920
capabilities can actually bust the encryption levels of many of the encryption tools and the
241
00:25:57,920 --> 00:26:03,040
algorithms that states use today to protect their data so many states are encrypting their data at
242
00:26:03,040 --> 00:26:08,000
rest they're using encryption for transport their websites all have capabilities where they're
243
00:26:08,000 --> 00:26:15,040
encrypted with tokens and quantum has the certainly the capability in the future to break those
244
00:26:15,040 --> 00:26:21,360
encryption and not and do it in a relatively rapid format not hundreds of years that we talked
245
00:26:21,360 --> 00:26:27,600
about before so that's one I think an area that is receiving more attention it kind of was at the
246
00:26:27,600 --> 00:26:33,760
high of the hype cycle and then went down onto the trough for a while's blockchain and so we have
247
00:26:34,480 --> 00:26:39,760
renewed interest in blockchain and a number of states Texas has been very aggressive new blockchain
248
00:26:39,760 --> 00:26:45,680
commission lots of activities with but blockchain but I think the poof is in the pudding to see
249
00:26:45,680 --> 00:26:51,520
exactly what the use cases are they never really materialized except that a fairly nominal level
250
00:26:51,520 --> 00:27:00,160
four or five years ago so that's a another one I think the the one area which I think is really
251
00:27:00,160 --> 00:27:07,120
fascinating from a kind of mixed technology both with data and with AI and with mechanical systems
252
00:27:07,120 --> 00:27:12,800
or drones on manned aerial systems so that is something that's here wrote about almost a decade ago
253
00:27:12,800 --> 00:27:18,480
and we've seen that come to fruition now there's even more advanced capabilities with that
254
00:27:18,480 --> 00:27:23,600
particularly powered by AI solution so we've up the game a little bit but every state today
255
00:27:24,400 --> 00:27:32,080
they are using a small unmanned drones to do a lot of work and they're highly highly productive so
256
00:27:32,080 --> 00:27:37,520
cost effective so I think that's another area we're going to continue to look at and at the Southern
257
00:27:37,520 --> 00:27:43,920
Legislative Conference we have policymakers legislative staff and leaders from across our 15 state
258
00:27:43,920 --> 00:27:50,320
Southern region from your perspective what are Southern states doing well and where do you
259
00:27:50,320 --> 00:27:57,040
see opportunities that exist yeah I think that's a that's an excellent question Lindsey because you're
260
00:27:57,040 --> 00:28:04,400
looking at kind of where are those two intersect and in my certainly career in technology in the
261
00:28:04,400 --> 00:28:11,440
in the public sector space I don't think I've seen previously as much activity and interest
262
00:28:11,440 --> 00:28:17,440
around a kind of what I would call a technology stack from a state legislative bodies and and
263
00:28:17,440 --> 00:28:27,920
deliberators 1500 plus bills introduced in in in 2026 almost over a thousand bills introduced in 2025
264
00:28:27,920 --> 00:28:36,160
and so they certainly kind of latched on with a high degree of interest of what's going on I think
265
00:28:36,160 --> 00:28:42,800
particularly concerned about the private sector use of AI and that's where most of these
266
00:28:42,800 --> 00:28:46,960
bill the majority of these bills are now focused on that but they're still a significant amount several
267
00:28:46,960 --> 00:28:54,080
I think over 300 bills this 2026 session focused on state government use and so when I look at kind of
268
00:28:54,080 --> 00:28:59,840
what they are concerned about a lot of it has to do with governance so they're they're looking at
269
00:28:59,840 --> 00:29:04,160
you know creating or already strengthening task forces and commissions that have been
270
00:29:04,160 --> 00:29:09,520
legislated be mandated to look at that they're concerned about the what I would call the
271
00:29:09,520 --> 00:29:14,880
inappropriate use of AI by state government and a lot of those have to do with topics I already
272
00:29:14,880 --> 00:29:20,640
mentioned things like ethical and responsible use decision making many of them have prohibited very
273
00:29:20,640 --> 00:29:27,600
clearly they prohibited state agencies from using AI to do things like behavioral health therapy
274
00:29:27,600 --> 00:29:33,760
they don't want chatbot therapists providing citizen services they don't want AI making
275
00:29:33,760 --> 00:29:38,880
determinations on eligibility things like Medicaid things like unemployment insurance things like
276
00:29:39,440 --> 00:29:45,760
benefit programs and so I think that's what the legislators have seen a lot of states in the south
277
00:29:45,760 --> 00:29:51,440
have taken the lead for that Georgia has a very excellent governance program and they have some
278
00:29:51,440 --> 00:29:58,240
widespread applications I mentioned what Mr. Cippy is doing north carolina also has that I don't
279
00:29:58,240 --> 00:30:03,760
want to single out all these states because the we don't rate and rank the states so they think
280
00:30:03,760 --> 00:30:08,960
that's a that's a challenge for me to look at all the great things are doing Texas has robust
281
00:30:09,360 --> 00:30:16,000
set of policies they were an early leader in enterprise policies around around AI but I think
282
00:30:16,000 --> 00:30:20,800
they're all focused on predominantly the same thing is making sure they have a strong foundation
283
00:30:20,800 --> 00:30:29,680
on the governance side to make those those policies and you know the legislative body can help by
284
00:30:29,680 --> 00:30:35,440
you know supporting them through the enactments that we've seen because almost all those states have
285
00:30:35,440 --> 00:30:43,040
a number of bills around around AI as we close I'll end with a question we ask all of our guests
286
00:30:43,040 --> 00:30:48,800
what is the best piece of professional advice you have ever received and how does it shape the way
287
00:30:48,800 --> 00:30:53,840
that you approach leadership today or in my career I've received a lot of advice
288
00:30:53,840 --> 00:31:02,720
I would say probably the one a couple things one is one is a quote
289
00:31:04,160 --> 00:31:10,160
that someone gave me 30 plus years ago I still have it on a little frame on my desk
290
00:31:10,160 --> 00:31:19,280
the life you lead is the lesson you teach and so I always think about that from a leadership
291
00:31:19,280 --> 00:31:25,920
perspective is that people are watching people are paying attention whether it's your team in my
292
00:31:25,920 --> 00:31:31,680
case whether it's my my 56 state territorial and district colonial members and they're they're
293
00:31:31,680 --> 00:31:38,400
watching so I always I look at that a lot talk about the fact that that that's a kind of a leadership
294
00:31:38,400 --> 00:31:46,320
mantra about whether and works in life it works in your personal life that works with your children
295
00:31:46,320 --> 00:31:51,920
right you're always you're always watching and so I think you know being you know ethical it if I
296
00:31:51,920 --> 00:32:00,400
high degree of integrity I think the other piece couple with that is something that I think one of
297
00:32:00,400 --> 00:32:08,000
my leaders in I work for some some excellent leaders over the years and and I think one of them always
298
00:32:08,000 --> 00:32:14,720
said you know stay humble be humble and I think that's that's important and then I try to certainly
299
00:32:14,720 --> 00:32:21,840
exercise that as much as I can I'm not very humble about about Nassio and the great work that
300
00:32:21,840 --> 00:32:28,560
that we are doing it's a great association and provide leadership for the states but you know
301
00:32:28,560 --> 00:32:34,960
have be humble about what what I'm doing so that's probably the my best advice for people Doug thank
302
00:32:34,960 --> 00:32:40,160
you for joining us at the Southern Legislative Conference this year and on our real state episode
303
00:32:40,160 --> 00:32:45,200
we are grateful to have an organization like the National Association of Chief Information Officers
304
00:32:45,200 --> 00:32:50,480
that is committed to helping state governments keep up with the ever changing landscape of technology
305
00:32:50,480 --> 00:32:57,440
our missions of supporting state governments strongly align and we deeply appreciate your partnership
306
00:32:57,440 --> 00:33:03,200
thank you Lindsey for being here and to our listeners thank you for tuning in to Real State for more
307
00:33:03,200 --> 00:33:11,360
conversations like this and more about the work of CSG South visit csgself.org I'm Lindsey Gray let's
308
00:33:11,360 --> 00:33:25,760
keep it real thank you for joining us today on real states where we keep it real about what really matters
309
00:33:25,760 --> 00:33:45,000
00:00:00,000 --> 00:00:03,360
What is coming next that state leaders should be watching for?
2
00:00:03,360 --> 00:00:07,040
I don't think there's much doubt that what's coming next is a genetic AI.
3
00:00:07,040 --> 00:00:13,120
This is Real State, a podcast from CSG South.
4
00:00:13,120 --> 00:00:19,680
Join us as we talk with policymakers and public leaders shaping state-level decisions across the
5
00:00:19,680 --> 00:00:25,040
South and explore how ideas move from policy to real impact.
6
00:00:25,040 --> 00:00:30,320
[MUSIC]
7
00:00:30,320 --> 00:00:35,280
Hello and welcome to Real State. I'm your host, Lindsey Gray. Back in July,
8
00:00:35,280 --> 00:00:40,160
we recorded some real-state episodes while at the Southern Legislative Conference,
9
00:00:40,160 --> 00:00:45,600
where state leaders from across our 15 state regions and beyond are discussing the issues
10
00:00:45,600 --> 00:00:51,680
that will shape the future of state government. And few issues are evolving faster than artificial
11
00:00:51,680 --> 00:00:58,800
intelligence. For states, the conversation around AI has quickly moved from beyond the basics.
12
00:00:58,800 --> 00:01:06,240
Leaders are now asking what tools are most effective? What problems can AI help us solve?
13
00:01:06,240 --> 00:01:13,280
How do we manage the risk and how do we prepare for what's coming next? AI, including
14
00:01:13,280 --> 00:01:21,840
generative AI, a genetic AI, and machine learning is now ranking as one of the top priorities for state CIOs.
15
00:01:21,840 --> 00:01:28,320
But successful adoption isn't just about deploying new technology. It requires thoughtful decisions
16
00:01:28,320 --> 00:01:34,320
around governance, security, data, workforce readiness, and public trust.
17
00:01:34,320 --> 00:01:40,080
Joining me today is Doug Robinson, Executive Director of NASCIO,
18
00:01:40,080 --> 00:01:46,560
the National Association of State Chief Information Officers. Doug works with CIOs
19
00:01:46,560 --> 00:01:53,680
across the country and brings a unique perspective on where states are today and where this technology
20
00:01:53,680 --> 00:01:59,280
is heading. He was also gracious enough to lead an awesome session at the Southern Legislative
21
00:01:59,280 --> 00:02:04,720
Conference this year. Doug, welcome to Real State. Pleasure to be here, Lindsey. Thank you for the
22
00:02:04,720 --> 00:02:11,280
invitation. Doug, you have a unique vantage point where you're talking with technology leaders
23
00:02:11,280 --> 00:02:17,680
across the country. When you look at where states are today with AI adoption, how would you describe
24
00:02:17,680 --> 00:02:25,360
this moment? Are we still in the experimentation phase or are states beginning to think about AI as
25
00:02:25,360 --> 00:02:33,600
a long-term transformation strategy? Certainly beyond experimentation. I've characterized
26
00:02:33,600 --> 00:02:41,840
2026 as the year of adoption. So we've moved for the most part, and again, I'm generalizing
27
00:02:41,840 --> 00:02:47,040
from the states, our motto at NASCIO, if you've seen one state, you've seen one state. So I have to
28
00:02:47,040 --> 00:02:54,000
look at kind of unevenness, but for the most part, about 80% of the CIOs report that they already have
29
00:02:54,640 --> 00:03:01,920
AI solutions, generative AI solutions in production. So I would say beyond experimentation and cautiously
30
00:03:01,920 --> 00:03:09,360
moving. So it's kind of a combination of adoption and kind of risk management. What are you hearing
31
00:03:09,360 --> 00:03:17,840
from state CIOs? And are they excited, cautious, or a bit of both? A little bit of both. I think
32
00:03:17,840 --> 00:03:25,600
there's certainly a lot of excitement about the possibilities. We have not seen a large percentage of
33
00:03:25,600 --> 00:03:34,000
these pilot projects move to scale and by that, I mean beyond an agency or beyond a certain line of
34
00:03:34,000 --> 00:03:39,680
business or a project. So when we look at our data, they're fairly specific in what I'll call kind
35
00:03:39,680 --> 00:03:46,080
of the functional domains of government, health and human services, transportation, tax and revenue,
36
00:03:46,080 --> 00:03:52,160
those are the large ones. But in terms of scaling across the entire executive branch, those have
37
00:03:52,160 --> 00:03:58,480
been limited to a lot of the productivity tools, but not something that we're saying. So I think that's
38
00:03:58,480 --> 00:04:05,760
kind of the cautious nature of what their role is, part of their role is risk management, regardless. So
39
00:04:05,760 --> 00:04:13,360
I think they're looking at the exciting opportunities, transforming how states serve their citizens,
40
00:04:13,360 --> 00:04:19,840
but also doing that cautiously because you just can't let the beast loose without those guard rails.
41
00:04:19,840 --> 00:04:26,560
A lot of attention goes to the future potential of AI. But I want to talk about what is available
42
00:04:26,560 --> 00:04:33,680
today? What AI tools or capabilities are already mature enough that states can realistically
43
00:04:33,680 --> 00:04:40,640
begin leveraging them? Yeah, so that's a good question because it's again mixed, but I'll generalize for
44
00:04:40,640 --> 00:04:47,120
the most part what's widely available are the productivity tools that are already built into
45
00:04:47,120 --> 00:04:54,480
collaboration platforms. So things like Microsoft, co-pilot and Google Gemini, those are widespread. Other
46
00:04:54,480 --> 00:05:00,960
states have direct through CIO directives are an executive order by the governor. They've deployed
47
00:05:00,960 --> 00:05:08,000
other tools like chat DPT and Clawed and others throughout the executive branch for employees to
48
00:05:08,000 --> 00:05:14,080
utilize. So those are widespread. Again, what is more limited are the more specific
49
00:05:14,080 --> 00:05:19,920
applications within a line of business. So you have to recognize that the states aren't, they're not
50
00:05:19,920 --> 00:05:26,480
building AI tools. These are commercial products. That's one of the challenges in terms of acquisition
51
00:05:26,480 --> 00:05:34,560
and procurement is all of these products today are you can say putting quotes powered by AI. So every
52
00:05:34,560 --> 00:05:40,320
solution they bring in most likely already has AI embedded in it and that's kind of one of the
53
00:05:40,320 --> 00:05:45,600
challenges is making sure that you have an inventory of what's available when you look at the states that
54
00:05:45,600 --> 00:05:51,760
do have inventories. You'll see it's predominantly commercial products that are focused on document
55
00:05:51,760 --> 00:05:57,840
preparation, generation, chat bots. Those are the kind of leading use cases across all the states.
56
00:05:57,840 --> 00:06:03,040
Where do you see the biggest opportunities for AI to improve how government operates?
57
00:06:03,040 --> 00:06:09,520
Whether that's internal efficiency, citizen services, cybersecurity or decision making.
58
00:06:09,520 --> 00:06:14,720
My speculation and certainly based on the evidence out there is the greatest opportunities are to
59
00:06:14,720 --> 00:06:21,200
enhance citizen services. But that comes with the challenge of trust. And so I think that's the one
60
00:06:21,200 --> 00:06:29,680
thing that CIO's and agency leaders are looking at is they see opportunities to apply beyond things like
61
00:06:30,800 --> 00:06:36,640
responding to questions out of chat bot or doing immediate translation services much more
62
00:06:36,640 --> 00:06:44,640
transformation from end to end cybersecurity. Almost 90% of the states today are using AI in
63
00:06:44,640 --> 00:06:52,160
cybersecurity for threat detection to examine anomalies to scrub logs. Things that take a lot of
64
00:06:52,160 --> 00:07:00,000
human capacity can be done with AI. So we see that as another from cybersecurity ops. Recognizing
65
00:07:00,000 --> 00:07:08,160
that the blade swings both ways. So the bad actors are also using AI extensively to craft
66
00:07:08,160 --> 00:07:14,400
very convincing, fishing emails and to do things in social engineering that weren't available for.
67
00:07:14,400 --> 00:07:21,120
So that's certainly part of it. I think the other area that I'm really looking at, because I think
68
00:07:21,120 --> 00:07:26,880
there's tremendous opportunity there is fraud detection, fraud prevention. So in a states of
69
00:07:26,880 --> 00:07:32,960
classically done kind of what's called the pay and chase model. You've seen a lot of headline news
70
00:07:32,960 --> 00:07:38,000
over the last several years certainly since COVID about the level of fraud and public benefit programs
71
00:07:38,000 --> 00:07:44,960
and just public distribution programs. It is is significant and in the major challenge for that is
72
00:07:44,960 --> 00:07:51,280
identity management and identity verification. So I think the states are testing AI several
73
00:07:51,280 --> 00:07:59,520
state revenue and tax departments are using AI already to look at how to advance that through AI
74
00:07:59,520 --> 00:08:05,280
as opposed to some of the traditional fraud tools. So that's another I think really strong use cases
75
00:08:05,280 --> 00:08:11,120
fraud detection. Are there use cases you were seeing emerge across the states that you think could
76
00:08:11,120 --> 00:08:17,040
become standard practices over the next several years for our states? We're capturing those with our
77
00:08:17,040 --> 00:08:23,600
regular surveys and we're just now analyzing our 2026 data to look across the word kind of to see
78
00:08:23,600 --> 00:08:30,960
where are those were the patterns of of use cases. Again, top of the list is in things like case
79
00:08:30,960 --> 00:08:38,400
management and benefits management with with cautions that ultimately its humans making a decision.
80
00:08:38,400 --> 00:08:45,120
We have a number of state legislative bodies that have already enacted legislation and the past bills
81
00:08:45,120 --> 00:08:51,600
that prohibit AI from making final things like Medicaid eligibility termination or insurance.
82
00:08:51,600 --> 00:08:58,560
However, those tools can be used to obviously streamline the massive amount of rules and
83
00:08:58,560 --> 00:09:04,720
regulations that case workers have to absorb. So we've seen good examples of that good examples
84
00:09:04,720 --> 00:09:12,240
and helping citizens navigate complex bureaucratic structures because they don't understand the rules
85
00:09:12,240 --> 00:09:19,120
of the road and so a lot of those business rules can be embedded in in AI some very fascinating things
86
00:09:19,120 --> 00:09:26,800
and natural resources some very discrete kind of use cases of things like predicting the
87
00:09:26,800 --> 00:09:33,360
ovulation cycles of stripe bass. We got that one. I thought that was very interesting. So a lot of
88
00:09:33,360 --> 00:09:42,000
things California has embarked on a fairly significant effort to look at things like accidents
89
00:09:42,000 --> 00:09:47,440
and to use data historical data that again and previously this wasn't possible with the with the
90
00:09:47,440 --> 00:09:54,320
capabilities and the compute powers to look at 10 years of historical accident data and make
91
00:09:54,320 --> 00:09:59,200
a some type of revelations around the causal factors of that. So that's what they are doing. They
92
00:09:59,200 --> 00:10:05,200
have two major initiatives around vehicular and pedestrian accidents and they are even embarked
93
00:10:05,200 --> 00:10:11,040
on that where they are looking at terabytes and terabytes of data that has been going to put
94
00:10:11,040 --> 00:10:17,040
together for AI analysis. So a lot of good things in those. Which you I think if you look at
95
00:10:17,040 --> 00:10:24,480
and just think about what are the major the functions of state government and the major large
96
00:10:24,480 --> 00:10:28,240
agencies that's where you're going to find a lot of activity because they're trying to reduce
97
00:10:28,240 --> 00:10:36,000
the overall burden but I think certainly the promise in the future is enhanced citizen services
98
00:10:36,640 --> 00:10:44,240
with with AI tools. Many states have launched AI pilots or created task forces to study potential
99
00:10:44,240 --> 00:10:50,400
applications but moving from pilot project to something that truly changes how government works
100
00:10:50,400 --> 00:10:57,360
can be challenging. So what separates states that are successfully implementing AI from those that
101
00:10:57,360 --> 00:11:03,440
are still trying to determine where to start? That's the question on the table today. Certainly as we've
102
00:11:03,440 --> 00:11:11,040
seen the states that have moved that if you look at the data from a year ago 90% of our states reported
103
00:11:11,040 --> 00:11:18,880
that they had pilot projects and only about 30% said they had moved those into production. So fast
104
00:11:18,880 --> 00:11:26,960
forward to 2026 our latest data indicates about 80% of them report that they've moved these pilots
105
00:11:26,960 --> 00:11:33,200
into production. Now some of them will not succeed. They will find issues with those but for the most
106
00:11:33,200 --> 00:11:41,040
part they're reporting that I think strong governance strong oversight, strong measurable impact and so
107
00:11:41,040 --> 00:11:48,080
be basically measuring along the way to see how these things are operating because they advance and
108
00:11:48,080 --> 00:11:53,920
evolve so quickly you have new models that come out all the time and those new models obviously can
109
00:11:53,920 --> 00:12:00,160
impact the direction of the project. So they're looking to stabilize those governance is a big part
110
00:12:00,160 --> 00:12:06,320
of that. The other certainly most defining challenge of all states is data quality and data integrity.
111
00:12:06,320 --> 00:12:14,560
I always remark years ago CIO on our board said to me, "Doug, for a bad data I wouldn't have any at all."
112
00:12:14,560 --> 00:12:21,040
And so that's one of the challenges is how do you make sure that you are ingesting high quality
113
00:12:21,040 --> 00:12:28,240
reliable data into your models that you're using to make those decisions? And I think the third
114
00:12:28,240 --> 00:12:34,560
thing which is a touchstone for all of this is workforce readiness. So we're seeing a lot of states ramp
115
00:12:34,560 --> 00:12:39,920
up their capabilities and in the workforce they're combining it with their cyber security training
116
00:12:39,920 --> 00:12:45,840
perhaps. Some cases they're making it mandatory if you're using an AI tool then you're required to
117
00:12:45,840 --> 00:12:50,800
receive some type of training and that training is going to focus on responsible and ethical use.
118
00:12:50,800 --> 00:12:57,360
And so that's a that can be a major issue. There will be missteps. I am confident that something
119
00:12:57,360 --> 00:13:03,120
will happen and states will learn from that. But and right now it's they're moving ahead but they're
120
00:13:03,120 --> 00:13:09,120
being cautious. Are there foundational investments whether that's cloud modernization, workforce
121
00:13:09,120 --> 00:13:14,240
training or governance that states need before they can fully take advantage of AI?
122
00:13:14,240 --> 00:13:18,880
Yeah well I've mentioned that governance is certainly at the top of the list which is those
123
00:13:18,880 --> 00:13:26,640
is the shared kind of decision model with in my area certainly of focus is the only the executive
124
00:13:26,640 --> 00:13:34,960
branch. So you have to have stakeholder engagement. I think you you bringing in outside experts as
125
00:13:34,960 --> 00:13:38,720
you have many states that are they're not only bringing in industry but they're bringing in the
126
00:13:38,720 --> 00:13:44,480
university. They have coupled their projects with university efforts. Mississippi is a good example
127
00:13:44,480 --> 00:13:50,320
of that. They have a project called Maine which is the Mississippi AI network and it's a combination
128
00:13:50,320 --> 00:13:57,280
of state government and the private sector and industry and higher education that are that are
129
00:13:57,280 --> 00:14:02,240
involved in that. I think that's kind of a good collaborative to the matter of model to look at
130
00:14:02,240 --> 00:14:08,400
other states have done very similar. So that's that's that's the challenge is how to how to kind of
131
00:14:08,400 --> 00:14:16,080
navigate all of that. Procurement is is another piece of the acquisition cycle is another kind of
132
00:14:16,080 --> 00:14:22,000
foundational aspect you have to get right. And those that comes with some challenges because
133
00:14:22,000 --> 00:14:29,920
states today don't have that maturity in how to require things like disclosure from you know all
134
00:14:29,920 --> 00:14:35,680
their vendors during association. So they're going to have to kind of work through you know all of that
135
00:14:35,680 --> 00:14:41,280
because there's tremendous liability if they if they don't workforce readiness as I've already
136
00:14:41,280 --> 00:14:46,320
mentioned in a couple of times is another one of those foundational pieces when you look at some of
137
00:14:46,320 --> 00:14:52,880
the challenges not not is not not not I think a couple of policy areas that have not been given
138
00:14:52,880 --> 00:15:01,280
enough exposure one is things like public records laws right so how do you make sure that you're
139
00:15:01,280 --> 00:15:07,200
aligned when you're using these tools to the existing open records laws that are in the state and
140
00:15:07,200 --> 00:15:12,560
for almost every state it's very similar record as a record regardless of physical form or
141
00:15:12,560 --> 00:15:20,240
characteristics. So for the states AI results AI prompts and AI results used by state agencies are
142
00:15:20,240 --> 00:15:25,440
part of the official public record but the majority of the states today are not handling it in that
143
00:15:25,440 --> 00:15:30,080
way. So they're going to have to advance the maturity of that not surprisingly technology is
144
00:15:30,080 --> 00:15:35,760
always way out in front of the policy framework and in the policy foundations in most states. So
145
00:15:35,760 --> 00:15:41,600
they always have to speed those up and move quickly. So let's stick with the governance theme.
146
00:15:41,600 --> 00:15:48,800
Nassio has highlighted AI governance as a major priority so this technology touches every agency
147
00:15:48,800 --> 00:15:54,960
every part of government so given that from your perspective what does a strong governance model for
148
00:15:54,960 --> 00:15:59,440
state governments look like? Well when you think about you think about governance particularly in a
149
00:15:59,440 --> 00:16:06,880
technology space there are certainly lots of different opinions about architecture essentially the
150
00:16:06,880 --> 00:16:14,800
the the the the blueprint of government as we describe it and so you have to be able to have a
151
00:16:14,800 --> 00:16:23,120
venue, a policy process, a collaborative model to have those shared decision rights and if you
152
00:16:23,120 --> 00:16:30,960
look at the state today, overall majority of them there is a directive of some type of coming from
153
00:16:30,960 --> 00:16:36,800
the executive branch and from the state CIO which outlines what what is their AI framework? What
154
00:16:36,800 --> 00:16:42,160
is the responsible use? What are the ethical guidelines? What are the guardrails that you have to adhere
155
00:16:42,160 --> 00:16:49,520
to? A simple example would be employees are not permitted to use personally identifiable citizen
156
00:16:49,520 --> 00:16:56,160
information right when they're using these tools they can't they can't train the models using state
157
00:16:56,160 --> 00:17:00,720
data that's kind of a guardrail it's across all states so they're all very common they're they're
158
00:17:00,720 --> 00:17:06,400
borrowing from each other and looking out to success and so you know strong governance looks at that
159
00:17:06,400 --> 00:17:11,520
that collaborative approach and shared decision it can't be the big hammer type of authority that
160
00:17:11,520 --> 00:17:17,520
we've seen and it can't be the rigid you know you cannot use these tools it has to recognize that
161
00:17:17,520 --> 00:17:22,800
these tools are in place they're now ubiquitously available and there's got to be a flexibility
162
00:17:22,800 --> 00:17:31,520
around those around those use cases so it has to be taken seriously this is a a transformational
163
00:17:31,520 --> 00:17:38,160
technology in terms of its use but it can also inherit risks and so I think that's the role of
164
00:17:38,160 --> 00:17:45,440
of strong governance is having those conversations and adopting a rubric that works in in that in
165
00:17:45,440 --> 00:17:51,520
that state depending on their operating model so when you talk to state leaders what are the biggest
166
00:17:51,520 --> 00:17:58,560
concerns you hear around AI adoption they don't know what they don't know and I think that's the
167
00:17:58,560 --> 00:18:06,400
the major concern is you know they they they expect surprises and they're trying to mitigate those
168
00:18:06,400 --> 00:18:12,960
and so without understanding the complexities and again you even some of the data scientists and
169
00:18:12,960 --> 00:18:20,320
developers don't understand how these models are evolving and the constant flux in the marketplace
170
00:18:20,320 --> 00:18:25,600
and also the constant changing there's new new models and new capabilities that seems like coming
171
00:18:25,600 --> 00:18:32,080
out every week and so it can be very confusing because the states are trying to kind of put a
172
00:18:32,080 --> 00:18:38,800
stake in the ground at a point in time and it's constantly it's constantly moving and it creates
173
00:18:39,360 --> 00:18:44,480
you know some some friction perhaps and in whether it's because there might be some
174
00:18:44,480 --> 00:18:51,120
agency that really wants to move quickly but there's a level of uncertainty there about
175
00:18:51,120 --> 00:18:58,400
you know what the consequences might be so they've been very careful to test things go to the sandbox
176
00:18:58,400 --> 00:19:04,000
reevaluate and I think that's a big part is constant and continuous evaluation and some states
177
00:19:04,000 --> 00:19:09,440
the others report to me that they're not doing that very well yet because of the magnitude of that
178
00:19:09,440 --> 00:19:15,840
there's certainly as you know winzie a lot of hype around AI right it's at the higher that's kind of
179
00:19:15,840 --> 00:19:20,320
high hype cycle but you've got a temper that with the reality of the role of government
180
00:19:20,320 --> 00:19:28,400
and also the fact that look at the latest results from pew around trust in AI 70% of those
181
00:19:28,400 --> 00:19:35,840
respondents certainly US citizens said they don't trust AI 70% so they don't trust AI and yet we
182
00:19:35,840 --> 00:19:41,760
have state government agencies and leaderships trying to deploy AI across you know all matter
183
00:19:41,760 --> 00:19:47,920
of state services and that creates some some challenges and direct conflicts with what citizens want
184
00:19:47,920 --> 00:19:56,240
to see taking the matter of citizen data protection a step deeper states have a responsibility to
185
00:19:56,240 --> 00:20:03,440
protect citizens information and maintain trust how should leaders think about privacy transparency
186
00:20:03,440 --> 00:20:11,200
and ethical use as AI becomes more integrated in state governments again a broad constituency of
187
00:20:11,200 --> 00:20:17,360
collaborators internally so a specific example of we host the community of practice of the state chief
188
00:20:17,360 --> 00:20:21,760
privacy officers we also heist the state chief information security officers those are
189
00:20:22,640 --> 00:20:29,280
typically part of the CIO organization but not always and we are we we always recommend that the
190
00:20:29,280 --> 00:20:35,520
chief privacy officer be fully engaged in those decisions in fact in a couple of states the CPO
191
00:20:35,520 --> 00:20:43,120
is actually leading their AI effort so we now have over half the states that have a chief AI
192
00:20:43,120 --> 00:20:48,400
officer already appointed and many more that are operating in that role not without equivalent
193
00:20:48,400 --> 00:20:54,480
title but they they operate in support of that so we're seeing the growth of another chief physician
194
00:20:54,480 --> 00:20:59,920
but with tell them that they they have to have communications together and that's a big part of
195
00:20:59,920 --> 00:21:06,960
that to make sure that they are following the appropriate privacy there's no federal privacy rules
196
00:21:06,960 --> 00:21:11,680
there's no federal privacy legislation so each state have crafted their own and some of them
197
00:21:11,680 --> 00:21:17,760
you know are are very restrictive in terms of what the states can do so they have to make sure
198
00:21:17,760 --> 00:21:23,760
that they're aligned with that and that again requires that they they understand the the magnitude
199
00:21:23,760 --> 00:21:31,520
of a state data states have the most amount of personal information about all of us
200
00:21:31,520 --> 00:21:36,880
than any other entity from essentially cradle to grave right they have they issue the births
201
00:21:36,880 --> 00:21:43,520
certificate so they have all that information they host the gold standard for identity which is
202
00:21:43,520 --> 00:21:49,280
the driver's license and then of course they have a whole set of transactions which they're constantly
203
00:21:49,280 --> 00:21:54,960
collecting personally that have all information from citizens and that has to be protected and
204
00:21:54,960 --> 00:22:02,400
certainly fire walled off from a lot of these AI tools AI obviously is moving very very fast
205
00:22:02,400 --> 00:22:09,520
so many leaders are just getting their arms around generative AI now here comes the agents with
206
00:22:09,520 --> 00:22:15,600
more advanced capabilities what is coming next that state leaders should be watching for I don't
207
00:22:15,600 --> 00:22:21,600
think there's much doubt that what's coming next is a genetic AI as you mentioned that is from you
208
00:22:21,600 --> 00:22:28,640
know again our recent data about the next wave what's the the most what will be the most impactful
209
00:22:28,640 --> 00:22:34,560
emerging technology in the next three years it's two thirds of our our state CO said a
210
00:22:34,560 --> 00:22:40,800
genetic AI as they see that actually as surpassing generative AI because of its capabilities
211
00:22:40,800 --> 00:22:48,320
its capabilities to have actually navigate and and and make decisions of complex business rules
212
00:22:48,320 --> 00:22:54,400
that many citizens are not surprisingly are confounded by call it navigating the pirate map
213
00:22:54,400 --> 00:23:00,000
it's very difficult to understand that so a gentigate AI and those agents can actually give them the
214
00:23:00,000 --> 00:23:05,120
tools again comes with a lot of inherent risk a lot of security risks and privacy risks but
215
00:23:05,120 --> 00:23:11,040
they see that as the next capability to use agent technology they're using it a number of states
216
00:23:11,040 --> 00:23:18,720
already using a gentigate AI internally they're using it to make a kind of simple decisions along
217
00:23:18,720 --> 00:23:25,600
along those complex business rules internally I see them opening that up two citizens in the near
218
00:23:25,600 --> 00:23:33,040
future so it is is not widespread is I think the maturity is uneven the adoptions uneven but they
219
00:23:33,040 --> 00:23:40,080
see the promise of what of what that can do and there's certainly plenty more that generative AI
220
00:23:40,080 --> 00:23:47,840
can do when you get down into get down into the large enterprise scale by that I mean cross boundary
221
00:23:47,840 --> 00:23:53,520
interoperability across multiple agencies right now that is a fairly low level and so we're going
222
00:23:53,520 --> 00:23:59,520
to see that we see that build up but I think it's exciting to look at the I guess the innovation
223
00:23:59,520 --> 00:24:08,640
that you're seeing across the states and the fact that old old the legacy modernization topic is
224
00:24:08,640 --> 00:24:14,080
always kind of the top of the list for states see I just modernization and general every state CIO
225
00:24:14,080 --> 00:24:20,160
aspires to have a modern environment these tools can help leak accelerate that but again with
226
00:24:20,160 --> 00:24:26,480
they're doing it very cautiously understanding that things that will certainly could go wrong and
227
00:24:26,480 --> 00:24:30,960
will go wrong at some point in the future is they're trying to mitigate the risk as much as they
228
00:24:30,960 --> 00:24:36,720
can and provide guidance through their policy memorandum and their direction so that'll something
229
00:24:36,720 --> 00:24:42,960
will solidify over the next several years beyond a genetic AI what other emerging tools or
230
00:24:42,960 --> 00:24:49,520
capabilities do you see that could be especially valuable for states over the next five years I
231
00:24:49,520 --> 00:24:57,440
think we're looking at the tools that were originally deployed at a fairly low level during
232
00:24:57,440 --> 00:25:04,880
during COVID one is low code no code development and there's also vibe coding which involves AI
233
00:25:04,880 --> 00:25:11,200
but I think low code no code is really bringing the capabilities of application development out of
234
00:25:11,200 --> 00:25:17,440
the software solution shop and giving it to the lines of business meaning that that state
235
00:25:17,440 --> 00:25:23,840
employees who know the business process can now use modules and kind of plug and play
236
00:25:23,840 --> 00:25:29,760
to develop solutions we saw that happen during COVID that has become certainly more widespread
237
00:25:29,760 --> 00:25:37,360
quantum certainly something that states are just now looking at we have a couple of states that
238
00:25:37,360 --> 00:25:43,680
have quantum readiness task forces and looking at that that the the danger with quantum from the
239
00:25:43,680 --> 00:25:50,720
states perspective is the security concerns and the capabilities that that actors using quantum
240
00:25:50,720 --> 00:25:57,920
capabilities can actually bust the encryption levels of many of the encryption tools and the
241
00:25:57,920 --> 00:26:03,040
algorithms that states use today to protect their data so many states are encrypting their data at
242
00:26:03,040 --> 00:26:08,000
rest they're using encryption for transport their websites all have capabilities where they're
243
00:26:08,000 --> 00:26:15,040
encrypted with tokens and quantum has the certainly the capability in the future to break those
244
00:26:15,040 --> 00:26:21,360
encryption and not and do it in a relatively rapid format not hundreds of years that we talked
245
00:26:21,360 --> 00:26:27,600
about before so that's one I think an area that is receiving more attention it kind of was at the
246
00:26:27,600 --> 00:26:33,760
high of the hype cycle and then went down onto the trough for a while's blockchain and so we have
247
00:26:34,480 --> 00:26:39,760
renewed interest in blockchain and a number of states Texas has been very aggressive new blockchain
248
00:26:39,760 --> 00:26:45,680
commission lots of activities with but blockchain but I think the poof is in the pudding to see
249
00:26:45,680 --> 00:26:51,520
exactly what the use cases are they never really materialized except that a fairly nominal level
250
00:26:51,520 --> 00:27:00,160
four or five years ago so that's a another one I think the the one area which I think is really
251
00:27:00,160 --> 00:27:07,120
fascinating from a kind of mixed technology both with data and with AI and with mechanical systems
252
00:27:07,120 --> 00:27:12,800
or drones on manned aerial systems so that is something that's here wrote about almost a decade ago
253
00:27:12,800 --> 00:27:18,480
and we've seen that come to fruition now there's even more advanced capabilities with that
254
00:27:18,480 --> 00:27:23,600
particularly powered by AI solution so we've up the game a little bit but every state today
255
00:27:24,400 --> 00:27:32,080
they are using a small unmanned drones to do a lot of work and they're highly highly productive so
256
00:27:32,080 --> 00:27:37,520
cost effective so I think that's another area we're going to continue to look at and at the Southern
257
00:27:37,520 --> 00:27:43,920
Legislative Conference we have policymakers legislative staff and leaders from across our 15 state
258
00:27:43,920 --> 00:27:50,320
Southern region from your perspective what are Southern states doing well and where do you
259
00:27:50,320 --> 00:27:57,040
see opportunities that exist yeah I think that's a that's an excellent question Lindsey because you're
260
00:27:57,040 --> 00:28:04,400
looking at kind of where are those two intersect and in my certainly career in technology in the
261
00:28:04,400 --> 00:28:11,440
in the public sector space I don't think I've seen previously as much activity and interest
262
00:28:11,440 --> 00:28:17,440
around a kind of what I would call a technology stack from a state legislative bodies and and
263
00:28:17,440 --> 00:28:27,920
deliberators 1500 plus bills introduced in in in 2026 almost over a thousand bills introduced in 2025
264
00:28:27,920 --> 00:28:36,160
and so they certainly kind of latched on with a high degree of interest of what's going on I think
265
00:28:36,160 --> 00:28:42,800
particularly concerned about the private sector use of AI and that's where most of these
266
00:28:42,800 --> 00:28:46,960
bill the majority of these bills are now focused on that but they're still a significant amount several
267
00:28:46,960 --> 00:28:54,080
I think over 300 bills this 2026 session focused on state government use and so when I look at kind of
268
00:28:54,080 --> 00:28:59,840
what they are concerned about a lot of it has to do with governance so they're they're looking at
269
00:28:59,840 --> 00:29:04,160
you know creating or already strengthening task forces and commissions that have been
270
00:29:04,160 --> 00:29:09,520
legislated be mandated to look at that they're concerned about the what I would call the
271
00:29:09,520 --> 00:29:14,880
inappropriate use of AI by state government and a lot of those have to do with topics I already
272
00:29:14,880 --> 00:29:20,640
mentioned things like ethical and responsible use decision making many of them have prohibited very
273
00:29:20,640 --> 00:29:27,600
clearly they prohibited state agencies from using AI to do things like behavioral health therapy
274
00:29:27,600 --> 00:29:33,760
they don't want chatbot therapists providing citizen services they don't want AI making
275
00:29:33,760 --> 00:29:38,880
determinations on eligibility things like Medicaid things like unemployment insurance things like
276
00:29:39,440 --> 00:29:45,760
benefit programs and so I think that's what the legislators have seen a lot of states in the south
277
00:29:45,760 --> 00:29:51,440
have taken the lead for that Georgia has a very excellent governance program and they have some
278
00:29:51,440 --> 00:29:58,240
widespread applications I mentioned what Mr. Cippy is doing north carolina also has that I don't
279
00:29:58,240 --> 00:30:03,760
want to single out all these states because the we don't rate and rank the states so they think
280
00:30:03,760 --> 00:30:08,960
that's a that's a challenge for me to look at all the great things are doing Texas has robust
281
00:30:09,360 --> 00:30:16,000
set of policies they were an early leader in enterprise policies around around AI but I think
282
00:30:16,000 --> 00:30:20,800
they're all focused on predominantly the same thing is making sure they have a strong foundation
283
00:30:20,800 --> 00:30:29,680
on the governance side to make those those policies and you know the legislative body can help by
284
00:30:29,680 --> 00:30:35,440
you know supporting them through the enactments that we've seen because almost all those states have
285
00:30:35,440 --> 00:30:43,040
a number of bills around around AI as we close I'll end with a question we ask all of our guests
286
00:30:43,040 --> 00:30:48,800
what is the best piece of professional advice you have ever received and how does it shape the way
287
00:30:48,800 --> 00:30:53,840
that you approach leadership today or in my career I've received a lot of advice
288
00:30:53,840 --> 00:31:02,720
I would say probably the one a couple things one is one is a quote
289
00:31:04,160 --> 00:31:10,160
that someone gave me 30 plus years ago I still have it on a little frame on my desk
290
00:31:10,160 --> 00:31:19,280
the life you lead is the lesson you teach and so I always think about that from a leadership
291
00:31:19,280 --> 00:31:25,920
perspective is that people are watching people are paying attention whether it's your team in my
292
00:31:25,920 --> 00:31:31,680
case whether it's my my 56 state territorial and district colonial members and they're they're
293
00:31:31,680 --> 00:31:38,400
watching so I always I look at that a lot talk about the fact that that that's a kind of a leadership
294
00:31:38,400 --> 00:31:46,320
mantra about whether and works in life it works in your personal life that works with your children
295
00:31:46,320 --> 00:31:51,920
right you're always you're always watching and so I think you know being you know ethical it if I
296
00:31:51,920 --> 00:32:00,400
high degree of integrity I think the other piece couple with that is something that I think one of
297
00:32:00,400 --> 00:32:08,000
my leaders in I work for some some excellent leaders over the years and and I think one of them always
298
00:32:08,000 --> 00:32:14,720
said you know stay humble be humble and I think that's that's important and then I try to certainly
299
00:32:14,720 --> 00:32:21,840
exercise that as much as I can I'm not very humble about about Nassio and the great work that
300
00:32:21,840 --> 00:32:28,560
that we are doing it's a great association and provide leadership for the states but you know
301
00:32:28,560 --> 00:32:34,960
have be humble about what what I'm doing so that's probably the my best advice for people Doug thank
302
00:32:34,960 --> 00:32:40,160
you for joining us at the Southern Legislative Conference this year and on our real state episode
303
00:32:40,160 --> 00:32:45,200
we are grateful to have an organization like the National Association of Chief Information Officers
304
00:32:45,200 --> 00:32:50,480
that is committed to helping state governments keep up with the ever changing landscape of technology
305
00:32:50,480 --> 00:32:57,440
our missions of supporting state governments strongly align and we deeply appreciate your partnership
306
00:32:57,440 --> 00:33:03,200
thank you Lindsey for being here and to our listeners thank you for tuning in to Real State for more
307
00:33:03,200 --> 00:33:11,360
conversations like this and more about the work of CSG South visit csgself.org I'm Lindsey Gray let's
308
00:33:11,360 --> 00:33:25,760
keep it real thank you for joining us today on real states where we keep it real about what really matters
309
00:33:25,760 --> 00:33:45,000
