Episode Transcript
Transcripts are displayed as originally observed. Some content, including advertisements may have changed.
Use Ctrl + F to search
0:03
Aqualars . This
0:06
is the Waterforsight podcast powered
0:08
by the Aqualars group , where we anticipate
0:10
, frame and shape the future of water
0:13
through strategic foresight . Today's guest
0:15
is Hudson Hollister , the founder and
0:17
CEO of Hdata . Hudson
0:20
, welcome to the Waterforsight podcast
0:22
.
0:23
Matt , thank you so much for having me . I'm excited
0:25
to participate in this conversation .
0:27
Well , I am too , and I
0:29
want to understand and get to know Hdata
0:32
. Tell me , what is Hdata ?
0:35
Hdata is a platform for regulation
0:37
. Our platform brings together all the
0:39
information of regulation as structured
0:41
and unstructured data and then allows people
0:44
that work with regulation to get their jobs done
0:46
. If you need to file forms
0:48
, or if you need to compare
0:50
information that comes from those forms , or
0:53
if you need to query the whole sector
0:55
using AI , you can do all that On the
0:57
Hdata platform . We work in
0:59
the energy sector today and I know
1:01
that we're going to have some interesting conversations
1:04
about whether the same technology could
1:06
be applied to water , but in the energy
1:08
sector . Our platform serves
1:10
the energy companies and it also
1:12
serves the energy regulatory agencies
1:15
, like the commissions at every state , and
1:17
it serves anybody else that cares
1:20
about energy regulatory information . So
1:22
I really have three kinds of customers that
1:24
all use the Hdata platform .
1:26
So you got the energy companies , like maybe a Duke
1:28
Energy . You've got a regulator
1:31
like the Federal Energy
1:33
Regulatory Commission or a State Public
1:36
Utility Commission . What
1:39
about customers ? Let's say I'm a big industrial
1:42
customer . Can I be a member or
1:44
subscriber to Hdata and gather
1:46
information ?
1:47
You sure can . We have some potential
1:50
customer conversations going on with a
1:53
lot of large power users . Some
1:55
of those are interested in Hdata because they want to
1:57
use regulatory information to figure out if
1:59
they can save money on what they're spending on
2:01
power . But not just customers , also
2:03
suppliers or those
2:05
that work on transactions with assets
2:08
in the industry , or consultants
2:10
. We have consultants
2:12
that pay for access to Hdata and
2:14
we just signed a contract with a major
2:16
supplier that sells
2:18
to the regulated energy
2:21
sector .
2:22
Wow , so even
2:25
if I am an energy company
2:27
, I can work with you to gather
2:30
even competitive information . What
2:32
are my fellow energy
2:34
companies doing in different states ? Is
2:37
there a competitive aspect to this as well
2:39
?
2:39
There sure is . Think about regulatory
2:41
information . It's all these forms that the
2:44
entities that are regulated have to file
2:46
with the government , and some of those forms end
2:48
up being public because of the
2:50
policy decision that we've made that annual
2:53
and quarterly reports by regulated
2:55
utilities , for instance , ought to be
2:57
published for everyone to use . Whatever
2:59
is public ought to be published
3:02
and ought to be really easy for you to grab
3:04
information that compares your own
3:06
entity to all the other ones using that
3:08
corpus . So , like permits
3:10
, it's reports
3:13
, it's permits , and then also it's anything
3:15
that goes into a docket . I know some of
3:17
the audience probably spends
3:19
time looking at regulatory commissions
3:21
and , as they might know , those regulatory
3:23
commissions publish all of their administrative
3:25
records and announcements and propose the final rules
3:28
on dockets , and so that's another source
3:30
for us . Think about all that as one
3:32
big pile . You have reports
3:35
and permits and docket items
3:38
all in one big mass , and
3:40
some of that information has numbers in it . Some
3:43
of that information is blobs of text
3:45
, but all it can be used
3:47
to better inform and better manage
3:50
whatever it is you're doing . If you
3:52
are a professional in energy regulation
3:54
.
3:55
Well , a lot of that is found
3:57
inside a rate case where
3:59
there's a lot of financial and other information
4:01
, capital projects et cetera and
4:03
you can get that information
4:06
. Is that what I hear ? That's right .
4:08
Some of our customers use each data . We're
4:10
getting a little ahead of ourselves because I want to talk about what our
4:12
functions do , but some of
4:14
our customers use each data to
4:17
grab our rate case and then query
4:19
it .
4:21
Well , let me ask this question what
4:23
inspired you to create
4:25
? Each data Sounds amazing .
4:27
15 years ago , I was a lawyer at the Securities
4:29
and Exchange Commission . The SEC regulates
4:32
public companies . The SEC was
4:34
not doing a very good job of it . When I worked there
4:36
that was during the financial crisis
4:38
I discovered one of the reasons why
4:41
the SEC failed to detect
4:43
some famous frauds , like the one perpetrated
4:45
by Bernie Madoff . The reason why the SEC
4:48
failed to detect indicators
4:50
of systemic risk . And the reason is we didn't
4:52
use any technology . We at
4:54
the agency were manually reviewing
4:57
all those forms from public companies . We
4:59
were verifying the mathematics of
5:01
public company financial statements with calculators
5:04
. This did not make any sense . We
5:06
weren't using even the technology
5:08
of that time . We weren't using the technology of that
5:10
time to analyze numbers
5:13
or to hunt through the text
5:15
for patterns , and so ever since
5:17
then , I've been trying to bring technology to regulation
5:20
. I resigned from the SEC and I went
5:22
to work for Congress trying to pass laws to
5:24
force regulatory agencies to digitize
5:27
all their forms and stuff . And then , in 2012
5:29
, I started the data coalition , which is a trade
5:31
association that lobbies regulatory
5:34
agencies to get them to digitize . One
5:36
of the agencies that we were able to persuade
5:39
to adopt the right
5:41
data format for all the numbers that are
5:43
in the filings is the Federal Energy
5:45
Regulatory Commission .
5:46
Hmm .
5:48
Once I began working with the FERC on digitizing
5:50
all of its forms , that's when I decided to start
5:52
H data and build a platform that
5:54
would use all of the data of energy
5:56
regulation to help the people that work
5:59
in regulated energy solve
6:01
problems .
6:02
So we need to get rid of the
6:04
typewriters , the bag phones and the slide
6:06
rules . Is that fair ?
6:09
Yeah , it's , it's fair to say and it's
6:11
it's no knock on the people that work
6:13
in energy regulation , because naturally the
6:15
folks that manage the whole flow
6:18
of documents , the whether it's the filings
6:20
going up or the dockets coming down their
6:22
subject matter experts in energy , not
6:25
in data , not in analytics , not in technology
6:27
. And so it's quickly . It takes some time
6:30
. We lag behind
6:32
other kinds of industries in
6:34
transforming the corpus
6:36
of regulatory information
6:38
from documents into data
6:40
and then applying technology to it . It just takes longer .
6:43
Yeah , I may be an excellent forensic
6:45
accounting person
6:47
, but I'm not very good at
6:49
the technological aspects
6:52
to how to move from analog to digital
6:54
. Fair enough , two different careers yeah
6:57
, interesting so
6:59
broadly . What kind of
7:01
or what type of future or
7:03
futures did you envision when
7:05
you were building H data ? Was
7:08
it efficiency ? Was it trying to detect
7:10
bad behavior ? What kind of futures
7:12
were you thinking about ?
7:14
It's efficiency and transparency . If we
7:16
have all the information of energy
7:18
regulation , if it's all expressed as data , then
7:21
it becomes possible to automate a lot of the work
7:23
that we used to do . It's
7:25
also transparency . I think both the
7:27
regulatory tours and the regulatory Ted and
7:30
the outside observers like those suppliers
7:32
and customers we talked about Everybody makes
7:34
better decisions if they have access to the right data
7:36
and we even have better cooperation
7:39
between the parties . I can give you one really specific
7:41
example . Earlier this year
7:43
, the Nevada Public Utilities Commission announced
7:45
that it was using H data and it
7:48
had found some unsupported
7:50
costs in
7:53
the annual reports of Nevada Energy
7:55
and Nevada Energy actually
7:57
with some of that . Nevada Energy agreed and
8:00
accepted
8:03
that because there was some costs
8:05
that had been miscategorized by accident that
8:08
would never have been found in mountains of
8:10
manual documents . But
8:12
that became really obvious using H data
8:15
for electronic review . And it wasn't
8:17
acrimony , yes , it was just let's
8:19
use technology to answer
8:22
the questions most effectively , wow
8:24
.
8:26
A couple I'll call this maybe some lightning
8:28
round questions , because I think you've touched on these already
8:30
, but the range of solutions
8:32
or aspects of H
8:34
data first . You're in the energy
8:37
world , you're thinking about water and we'll get to that , but
8:39
it's energy . I hear you . If
8:42
I hear you correctly , you
8:44
can get federal data
8:46
, state data and even local data
8:48
. Is that fair ?
8:51
We've got federal and state , today we
8:54
can pull local . We haven't done it yet .
8:58
And then you've got regulatory
9:00
data . What about statutes or
9:04
even policy linked data
9:06
? That's
9:08
fair game for H data .
9:10
It's fair game , but the way that we focus and
9:13
the way that we do , the way that we've
9:15
set up our platform , really
9:17
has two halves to it . We
9:20
have structured on one side and
9:22
we have unstructured on the other , and
9:24
so everything that we try to collect when
9:26
we crawl through all these sources , we
9:29
figure out whether it's to be structured or unstructured
9:32
. Structured
9:34
includes numbers . So
9:37
for all the forms that get filed with the FERC because
9:40
the FERC switched to a new data format instead
9:42
of the PDF we are able to harvest individual
9:44
numbers , each as their own data field . So
9:47
every schedule , every number on every schedule
9:49
of every annual and every quarterly
9:51
form that gets filed with the FERC , we have individual
9:54
numbers , and that means that that's
9:56
all structured . Every number is
9:58
its own data field , so we can use that to
10:00
build charts and graphs and
10:02
more complex things like whole models
10:04
, and the charts and graphs and models they
10:07
change automatically whenever the numbers do
10:09
. That's the structured side . On
10:11
the unstructured side , we have things that
10:13
aren't ever going to really be fully
10:16
built out as structured
10:18
data . You have blobs
10:20
of text , like a witness testimony in a rape
10:22
case . That's always going to be a bunch of text , and
10:25
so for the unstructured side
10:27
, we have libraries of documents , kind
10:30
of more familiar with libraries of documents
10:32
, and we go through those and we can use
10:34
AI to query them .
10:37
I was going to ask about that . I'm glad you talked about the
10:39
structure and unstructured data . What
10:42
about and I think this is unstructured
10:44
data ? But what about utilities
10:46
that are being discussed
10:49
over social media ? Customers
10:52
are making comments about their utility
10:55
on Facebook or Twitter
10:57
or TikTok or even LinkedIn . Is
11:00
that something that HData can eventually
11:02
go get , if not today ?
11:04
No , our corpus is regulatory
11:06
information , authoritative
11:09
stuff . It's information that's
11:11
either filed officially with the regulatory
11:13
agency , so that's the filing that's going up
11:15
, or information that is
11:17
published officially by a regulatory
11:19
agency that's dockets coming down . The
11:22
reason why we limit ourselves to that is
11:24
that our customers need to refer
11:27
to the authoritative
11:31
official sources . You could have Google
11:33
search or you could have chatGPT make something up
11:35
for you if you broadened
11:37
out your work from that . What
11:40
our customers need is they need the right answer and
11:42
they need it to be linked all the way back to the source . They
11:45
need to be able to go to the official source , the
11:47
official filing that was submitted
11:50
to the regulatory agency , or the
11:52
document wherever it
11:54
was published on a regulatory agency docket
11:56
, and they need to go back to that and verify
11:58
it , and they want to do that automatically . We
12:01
offer them a link that goes all the way back
12:03
to the official source , wherever it lives .
12:08
The example would have to be if I'm
12:10
a disgruntled commercial
12:13
, industrial , residential customer and
12:16
I file an email or letter to
12:19
the consumer advocate or the commission in
12:21
a rape case , that would be attached
12:23
to the docket . You could find those things there
12:25
, but just not random people talking
12:27
about their energy or water service
12:30
on Facebook . It's
12:33
a different idea . Is that fair
12:35
Right ?
12:36
We want to optimize all of the technologies
12:39
we use the analytics tools that build
12:41
the charts and graphs , the large language
12:43
models that answer unstructured queries
12:45
, the AI models and the AI technologies
12:47
. We
12:50
want to maximize their usefulness for this specific
12:53
kind of information regulatory
12:55
information .
12:56
So you are focused
12:58
on a lot of the energy
13:00
or economic regulators . I think we talked
13:02
about FERC or maybe the
13:05
SEC or the Public Service Commissions
13:07
. What about other agencies
13:09
? Are you looking at other agencies ? I don't want
13:12
you to have to tip your hand , but what about EPA
13:14
State
13:16
environmental agencies , a lot of these
13:18
energy companies ? They've got environmental
13:20
issues , not just the economic issues
13:22
. Where do we
13:25
go .
13:26
I think we will go there after we do
13:28
some growth . We're
13:31
pretty full up with the
13:34
solutions that rest
13:36
on top of the energy regulatory data
13:39
and with the customers that
13:41
need help with energy
13:43
regulatory data Interesting .
13:45
Okay , well , I get the
13:47
wink and nod . There I get it . It
13:49
sounds like that would be quite another
13:51
conversation . To talk about
13:54
some of the other data that
13:56
are out there that these companies
13:58
could look at . What
14:02
do you see in terms
14:04
of a future for H data and the water
14:06
sector ? I know we've talked quite a bit now about the
14:09
energy sector as a good foundation
14:11
, but what if we took
14:13
H data and all the things you've talked about and
14:16
pivoted over to the water sector
14:18
? What would that look ?
14:20
like . Well
14:23
, none of the software solutions that we
14:25
have built are hard-coded
14:28
for energy . I
14:30
think everything that we've built could work for water
14:32
. We have a TurboTax-style
14:35
solution that files your forms automatically and
14:37
it checks them automatically . I
14:40
mentioned the charts and graphs . We have solutions
14:43
that pull numbers from wherever they live and
14:46
make comparisons , some visual
14:48
comparisons and some financial models and
14:50
so on , that allow the tracking
14:53
of different entities . We have regulatory
14:55
AI , which hunts through
14:57
millions of documents , millions of pages
14:59
, and it gives you an accurate answer , summaries
15:02
and recommendations , and even what will this witness
15:04
say next time . All of those technologies
15:07
are agnostic . We
15:10
, of course , are connecting all of them to energy
15:12
regulation because we built a data net , a
15:14
data store integrated of energy
15:16
regulation , but all of those technologies
15:19
could also work for other kinds of regulation
15:21
.
15:22
Wow , okay . So
15:24
when I'll use my foresight terms
15:26
here about possible , probable and
15:29
plausible futures ? I hear
15:31
you saying , yeah , I do see an
15:33
H data for water . I mean , it is something
15:36
that's out there that we could see in
15:38
the future . Fair enough , yes , okay
15:41
. What
15:45
about the future ? And I guess this
15:47
would be applicable to water and
15:49
energy utilities . But how
15:51
would this in the future , this
15:53
being your company
15:56
and its tools , how
15:58
would this apply to the
16:00
future of rate making ? I mean , as you mentioned
16:02
earlier , rate cases are a big deal , would
16:07
we see ? We can talk about
16:09
these things separately , but just alternative
16:11
rate making issues , formulaic or
16:13
benchmarking , future test year
16:15
, the speed of rate making . What
16:19
about the issue of
16:22
is there too much transparency
16:24
? That may make you upset . I know
16:26
you want transparency , but is there such a thing as
16:28
too much transparency with regulators
16:30
or even customers ? And then , is this
16:32
a tool
16:35
that can be used in a predictive manner
16:37
versus just simply a reactive
16:39
manner ? So those are just some different
16:41
questions and we could go through them , but your
16:43
thoughts on any one of those ?
16:47
First , an alternative rate making . A
16:49
lot of the complication
16:53
and expensive alternative rate making is just it's
16:56
not complicated work , it's just
16:58
pulling the numbers from wherever they live officially
17:00
and plugging them together in different
17:03
configurations
17:05
, and that's what Hdata excels at . If
17:07
you know which numbers you need , you
17:10
can use our gateway solution to identify
17:13
them and make a
17:15
report , or you can
17:17
use our solutions to make a chart or a graph and
17:19
it will just automatically update whenever the underlying
17:21
numbers do you , you can see the
17:23
rates or whatever the formula
17:26
spits out . You can see that change in real time
17:28
as forms are filed . We
17:30
can save some time there . Most
17:33
of the work of rate making in a
17:35
rate case is creating
17:38
and then submitting and then absorbing
17:40
large quantities
17:42
of paper or virtual
17:45
paper on PDFs , and
17:47
that's where regulatory AI
17:49
excels . We're able to answer
17:52
questions about a whole compilation
17:54
of documents . What did this witness
17:56
say ? What
17:58
are the rate formulas that are
18:00
in this 200 page document ? What
18:03
is the commission likely to ask us
18:06
about this document ? Or what did
18:08
the commission say here ? What's it
18:10
likely to say in the future ? I
18:13
think the substance of that work will not change
18:15
, or it doesn't need to , but
18:17
the whole manual part that
18:20
comes before applying your brain , the
18:22
whole manual part that comes from having
18:25
to source numbers for structured data
18:27
or having to absorb and understand
18:30
documents for unstructured . That will go
18:32
away . Hdata can help that go away
18:34
. It'll have some
18:36
of the same work but it'll go faster to be much cheaper
18:38
.
18:40
If I'm on staff of a commission or
18:42
if I'm a consumer advocate . These
18:45
PDFs or even Excel
18:47
spreadsheets can be uploaded and
18:50
the different things that I'm
18:52
looking for can be pre-programmed so that
18:54
the software or Hdata
18:57
, can tell me where there are issues
18:59
, instead of me trying to find it
19:01
in mounds of paperwork . It
19:04
can very quickly in a matter of seconds
19:06
, I assume can tell me
19:09
here's a big change in this form , here's a change
19:11
in this number that's dramatic , or other
19:14
things , and I can then spend my time not
19:16
looking for things . I can spend time analyzing
19:19
things . Is that fair ?
19:21
Yes , most of what you just talked about was
19:23
reactive , not predictive . I think
19:25
most of the initial work that
19:27
we can assist is reactive
19:29
, not predictive . I want
19:32
to make sure that we cover that waterfront first
19:34
, but there are a few things that are predictive
19:37
, that our solution has been able to , our software
19:39
platform has been able to do . For
19:41
instance , you can pull up a rate
19:43
case witness testimony and you
19:45
can ask
19:48
regulatory AI to
19:50
invent questions that the commission
19:52
could choose to ask of this witness .
19:56
Is it even to the point where I
20:00
am making certain filings and
20:02
, as I go along , it will predict
20:05
if you will , or tell
20:07
me what my authorized
20:09
return on equity might be in a year or two if
20:12
I continue down this path . No
20:14
, no , okay , okay
20:16
, all right .
20:19
I can tell you if you're working
20:21
on a filing . It can tell you if
20:23
the filing is going to get
20:26
kicked back for
20:28
violating any of the mathematical formulas that
20:30
the regulator imposes . As
20:33
you're working on your filing , you'll see error
20:35
messages pop up that show you what you might
20:37
need to correct based on what
20:39
the regulator has said the formulas are . I
20:44
don't want to overpromise that
20:46
. We've already made a way to
20:48
predict how the regulator is
20:50
going to reply or going to react
20:52
to that filing years from now .
20:55
All right , I was joking that I would use the
20:57
phrase minority rate making
20:59
right Like out of minority report . We're
21:02
predicting that you're going to not
21:04
do well in two years , so we're going to make you a troubled
21:06
utility right and put you in receivership
21:09
. You're like , wait a minute , We'll
21:12
save that for the movies . It
21:15
seems that a lot of utilities
21:17
are interested in a
21:20
fully forecasted future test year as
21:22
an innovative rate making methodology
21:25
, and that's been received with some mixed
21:27
responses by different
21:29
parties . But how could
21:31
H data , if we talk about predictive elements
21:34
, could H data help utilities
21:36
and even commissions or consumer advocates
21:39
meaningfully address
21:41
this type of mechanism called
21:43
a future test year ?
21:46
If the future test year is built using
21:48
data from existing , from previous
21:50
ones , then we can
21:54
make the modeling much
21:56
easier and make the
21:58
assumptions much easier to test .
22:02
I wonder , in this world of alternative
22:04
rate making , whether this
22:07
type of a tool that you've developed will
22:09
move us to a different kind
22:11
of rate making environment
22:14
, more formulaic , or
22:18
even a benchmark type
22:20
of rate making where you can
22:22
put in some benchmarks and are
22:24
you achieving them or not , and that drives your
22:27
authorized return on equity
22:29
or the other elements in
22:31
a rate case . Do you think that this technology
22:33
may lead us to that kind
22:35
of a simplistic , formulaic
22:38
type of rate making world
22:40
or not ?
22:42
I'm not expert enough to know , but I do know that
22:44
if there are any regulatory
22:46
operations that we used to have
22:49
to forego because they
22:51
would be really expensive on the staff time
22:53
, maybe requiring a lot of computation
22:56
or maybe requiring a lot of quick
22:58
absorption of big quantities
23:00
of unstructured documents , we
23:02
won't have to forego those anymore because
23:04
the manual labor of
23:07
crunching numbers and the manual labor
23:09
of finding and absorbing
23:11
documents is going away .
23:14
Seems that in the future , regulated utilities
23:16
will have to be very careful about the data
23:19
and reports that they compile
23:21
and submit to regulators , because you
23:24
can't just dump a bunch of documents on them
23:26
and hope for the best . It's now going to be
23:29
very quickly reviewed , and
23:31
it just seems that what's left is
23:33
just the analysis . There's more time for analysis
23:36
and response rather than flipping
23:38
through pages of documents looking for a needle
23:40
in a haystack . Is that kind
23:42
of an outcome ?
23:43
It helps both sides . Both sides have
23:45
to invest incredible resources
23:49
, time and money in order to absorb
23:52
and understand that corpus . Today
23:55
, like all of the minimum information
23:57
requirements and a rate case , and then all the document
24:00
and information requests that are made
24:02
based on those minimum information requirements and the responses
24:05
to those documents and information requests , the
24:08
volume is really expensive
24:10
, not just for the commissions
24:12
or the intervenors , but also for the utilities
24:15
. That volume is probably going to
24:17
stay , unfortunately , but
24:19
it's going to no longer be so expensive .
24:24
Well , here's one of the other questions I had , and maybe
24:27
this is an easy one , but I asked about the speed of
24:29
rate making . Right now , rate cases
24:31
can take 300 days , maybe
24:33
even a year , depending on delays With
24:36
this technology . Is it simply
24:38
well , we have all the data , we've analyzed
24:41
it , we just need to have some
24:43
testimony , and instead of nine
24:45
months now it's three months
24:47
or six months . Your thoughts on that
24:50
?
24:51
Any of the work that is delayed by
24:53
the requiring people to read documents
24:55
will be faster .
24:58
Yeah , interesting . A
25:01
lot of people complain about that takes too long
25:03
. I make these investments
25:05
as a company . I got to wait to
25:07
get my recovery . And
25:10
then the transparency issue
25:12
. That's a very open-ended
25:15
issue . I know that's one of
25:17
the pieces of why you are sitting here
25:19
. You wanted to improve transparency . But does
25:23
this tool ? Does
25:26
it create enough transparency ? Not
25:28
enough , or maybe too much ? Where
25:30
are we going to be in 10 to 20 years ? Are
25:33
we going to say , wow , we've got too
25:35
much transparency ? I
25:38
don't know if that's an oxymoron or not , but your
25:40
thoughts on that ?
25:43
I think the most important point of modernizing
25:46
regulatory filings
25:48
going up and modernizing the
25:50
dockets and regulatory announcements coming down
25:53
is not that any new substance gets
25:55
revealed . It's all the same substance . It's just
25:57
easier to use that
25:59
corpus and so the transparency comes
26:01
from not from anything new being public
26:03
, but from the existing
26:06
stuff being easier to manage and understand
26:08
. And , yes
26:10
, there's more transparency that flows
26:12
from that . It's better transparency
26:15
of the existing public information
26:17
. It's harder to bury
26:19
something in the existing public information , but
26:22
the benefit is that we get faster
26:25
to a shared understanding
26:27
between the regulatory tour and regulatory Ted . The
26:30
regulatory tour and regulatory Ted both can get
26:32
a handle on that humongous corpus
26:34
faster and they can narrow
26:36
down what the actual disagreements might be , and
26:39
that's better for both sides .
26:41
What about the customers ? What about the nonprofits
26:44
that you know ? I'm sure you and I know
26:46
lots of different nonprofits that are representing
26:49
residential , commercial , industrial customers
26:51
. They often intervene in rate cases
26:53
and other proceedings . Do you think this tool
26:55
might help them ? Would they be interested in using
26:58
this tool to find opportunities
27:00
to better represent their concerns
27:03
and issues before different regulatory
27:05
bodies ?
27:06
We've had some really interesting demonstration
27:09
projects with intervenors
27:12
, some industry , some citizen
27:14
representing , and
27:16
we found that we can save them a lot of time . Usually they don't
27:18
have the same resources as
27:21
maybe the utility might
27:23
, and sometimes not even the same resources as the state
27:25
commission might . But
27:28
most of the expense of participating
27:30
in a rate case is all
27:33
of that time that it takes the
27:36
lawyer or consultant to absorb and understand
27:38
this out of matter . First , excuse
27:44
me , we shortened that time . No knock
27:46
on lawyers and consultants . We elevate
27:48
them , we won't get rid of them . The lawyers and consultants
27:50
are going to be able to skip forward past
27:53
all of the manual labor
27:55
and be able to apply their expertise
27:58
and their understanding earlier .
28:01
I can imagine individual customers being
28:03
able to someday partake
28:05
of each data's resources
28:09
and really setting up questions
28:11
or other analytical elements
28:14
that give them insights
28:17
into how a company is operating within their
28:19
community . That might help them as
28:21
they think about testifying or
28:24
in certain rate cases . That's
28:28
a future that I could see , because
28:30
you probably know that these rate
28:32
cases can be just intractable
28:34
Mountains of paperwork , complex
28:38
formulas , and how
28:40
does an individual customer , let alone a
28:42
commercial or industrial customer
28:44
, pour through this and figure out
28:46
what's fair and reasonable ? What's happening
28:48
in my neighborhood and
28:50
I could see in the future this tool
28:52
could be programmed to pretty easily pop
28:55
out a number or help
28:57
a customer understand different
28:59
things . That might give them the ability
29:02
to be more participatory
29:05
in regulatory
29:08
endeavors in the future . Am I
29:10
in the ballpark there or am I
29:12
on Mars ?
29:15
I do not know enough to predict , but I
29:17
do think that you
29:20
can . You can analyze the previous periods in history
29:22
. There was
29:24
a time when rate cases had to be managed
29:26
using literal paper , when
29:29
the intervenors had to go
29:32
to a physical place in order to get access
29:35
to read the minimum information requirements
29:37
, and that
29:39
is no longer true , because now the distributions
29:41
are electronic . You might have 11,000
29:44
document requests
29:46
in a rate case and those document
29:48
requests are now emailed to everybody , so you
29:50
can sit in your home office and do your work on the rate
29:53
case . Think about the
29:55
ways in which intervention
29:57
has changed as a result of that technological
30:00
change . Probably there are more intervenors
30:03
. Probably we
30:05
get to a more efficient result , because
30:07
if the cost of intervention goes down
30:10
, we might get
30:13
to a more efficient
30:16
consensus between
30:18
the regulator or the regulator and the intervenors
30:20
faster . So I
30:23
can try to analogize from that
30:25
to a world
30:27
in which the , the corpus
30:29
of information , can be managed and understood using each
30:31
data faster and better , and
30:34
I can come up with conclusions that way . But
30:37
I'm not a subject matter expert
30:39
. It rates , and so
30:41
I wouldn't venture to go further
30:43
than that .
30:44
Well , you can imagine , you know , 40
30:47
, 50 years ago it was the companies that had
30:49
these computers , right , and
30:52
this idea of a personal computer boy . That
30:54
was absurd . Right , a computer in every home . And
30:56
now it's a computer , multiple computers
30:59
, in the home . You got a computer in your hand
31:01
, you got a computer on your watch and
31:03
I had to think that , as you put it , you know , we've gone
31:06
to . You got to go to the government
31:08
file room to copy the paper to
31:10
. Well , now you can sign
31:12
up for the docket electronically and get PDFs
31:14
emailed to you or you can go dig
31:16
it out . What about
31:18
the next stage , which is not just getting the
31:21
documents but actually being able to
31:23
create algorithms
31:26
or other features
31:30
that give you outcomes or
31:32
insights from the materials
31:35
that have been filed with those regulators
31:37
? You know , I don't want to read 500
31:39
pages , I just want to know this number . Did
31:42
this project get done or where is this ? You know , and
31:45
you know that's
31:47
just a future that might be possible , if
31:50
not probable , from what
31:52
I hear from you . Yeah , my
31:55
question . You've touched on these two words
31:57
and I just wanted to ask it
31:59
and let you respond . But how
32:02
is Hdata helping with reactive
32:04
activities
32:06
or predictive activities . Is
32:11
it a tool that's just reacting , or
32:13
is it a tool to react to the things that
32:15
are filed , or can it be used by your
32:18
clients to
32:21
engage in any sort of predictive activity
32:24
? Now , that could be rate making . It could be and
32:27
I'll have some questions on , say , business development
32:29
or competitive intelligence your
32:32
thoughts on these two words reactive or predictive
32:34
.
32:36
The whole thing is reactive just
32:39
automatically , because Hdata
32:41
allows our customers to hunt
32:44
through over
32:46
10 years worth of filings going up
32:48
and dockets coming down yeah
32:50
, so there's a very rich
32:52
historical record
32:55
there . You can see
32:57
the history of return on equity
32:59
for every utility since 2011
33:02
, or you can instantly
33:05
access any rate case from
33:07
any interstate gas pipeline at
33:10
the FERC just by
33:12
typing in a quick query . So
33:16
it's a rich source of information
33:18
that regulatory
33:20
, tours , regulated companies and markets
33:23
can react to . The
33:26
parts of our platform that
33:28
allow for predictive work include
33:32
the ability to build formulas
33:35
. So if you know that you
33:37
want to use past information
33:39
to predict future information and you already know
33:42
the formula you want to use to make that prediction there's
33:44
always going to be a formula . If you're working with numbers , you
33:46
know the formula then you can put it into
33:48
our tools and you can extrapolate
33:51
forward and automatically
33:53
update that forward extrapolation
33:56
with whatever information is coming in
33:58
. That's on the structured side
34:00
. On the unstructured side , you can just
34:02
ask regulatory AI to give you
34:04
a prediction , and it will . Of course
34:06
, you're going to have to check that prediction
34:08
because it is a prediction , so
34:11
that's why it's important for our
34:14
regulatory AI to do citations . It's
34:16
important to link all the way back to the subject matter
34:18
and say , okay , here's what regulatory AI
34:20
says . Why did it say that ? And
34:22
this is one area where we try to
34:26
demystify AI . We
34:28
are not interested in being like chat GPT
34:30
. We do not want to wow somebody
34:32
with a fully formed answer . Instead , we want
34:34
you to see the components of the answer . We want
34:36
you to see why the AI
34:38
answered your query the way that it did
34:40
, and that's why we give you citations . We'll give you a
34:42
way to link back to the source material
34:45
that was used to create
34:47
the answer to your query in regulatory
34:49
AI and then figure out
34:51
which phrases
34:54
even came from which sources
34:56
, and so you can sort of see how the
34:58
answer came . People who work
35:00
in regulation need
35:02
citations back to the source , because
35:05
even when humans were doing it , that's
35:07
what we needed .
35:10
That's interesting . So you can then
35:13
look at why the answer
35:15
came back a certain way and
35:17
then be able to Same thing with the numbers .
35:18
You can link back to the original numbers there's
35:21
where they are in the form Right and you can say
35:23
hmm , okay that's interesting
35:25
.
35:26
Oh , we need to change that or that's not accurate
35:28
in our . Okay , so you can actually
35:30
go through and figure out where those
35:33
data came from and be able to adjust
35:36
because you have the source material .
35:38
Everything goes back to a source , an authoritative
35:41
source , and the authoritative source might be
35:43
a form that was filed with the regulatory
35:45
, or it might be something
35:47
that was published on a regulatory stock
35:50
.
35:52
Interesting . What are the possibilities
35:55
for Hdata being used
35:57
not by the utilities but by government
35:59
regulators ?
36:01
We've got four using it today .
36:05
Tell me , if you can , what inspired them to
36:07
work with you and what are they using
36:09
it for ?
36:11
Well , I mentioned the Nevada Public Utilities
36:13
Commission a few minutes ago
36:15
. The Nevada Public Utilities Commission is using
36:18
it for the same thing . Here's the thing the work
36:20
that regulators are doing is not that dissimilar
36:23
from the work that regulators are doing . Everyone's got to
36:25
look at this corpus and try to analyze
36:27
it .
36:30
Interesting . And is it just commissions
36:32
? Or what about the consumer advocates
36:35
or even the environmental regulators
36:37
? Have you just started to
36:39
see other regulatory
36:42
folks take an interest ?
36:44
Yes .
36:46
Interesting . All right , it
36:48
sounds like I shouldn't ask the next question
36:50
who . But what can Hdata
36:53
do
36:56
to help a utility with
36:59
the ever-persistent
37:02
challenges of capital investments
37:04
or operational efficiencies ? Have
37:06
you talked to some of your clients about those
37:08
aspects of the tool and what
37:12
can Hdata do to help with those things ?
37:14
Yeah , I can give you an example Plenty
37:17
of utilities in their CFO office
37:19
. They run analyses to
37:22
show their own capital efficiency
37:24
, their return
37:27
on equity or their return on invested capital
37:29
, their own operation maintenance expenses compared
37:31
with the others . And they use
37:33
and they run those analyses in
37:35
order to see and benchmark
37:38
themselves against their peers for
37:40
efficiency across the industry . And if they
37:42
see themselves lagging then they try to figure out
37:44
why and they try to figure out which cost items
37:46
are improving or
37:49
which cost items are helping them improve or
37:51
decline in their capital efficiency . All of
37:53
that comes from numbers that are knowable . All
37:56
of that comes from numbers that you can get from the forms
37:58
. It used to be really hard to
38:00
do this because you would have to hunt through the forms , find
38:03
your numbers and then plug
38:06
them into a spreadsheet and run the spreadsheet . Whenever
38:08
new forms came out , you would have to go and find
38:11
the new numbers and type them in . Or
38:13
whenever you wanted to add a
38:15
new company to your peer group . It
38:18
would be a couple of weeks of work . All
38:21
of that is instantaneous now .
38:23
And of course , that cuts both ways , because
38:26
if you're a utility , you can say , hey
38:28
, we've scanned all
38:30
the other regulated jurisdictions and our
38:33
project , our numbers , our costs
38:35
, they are within reasonable
38:38
standards . You can back that up with
38:41
data . But then commissions and even consumer
38:43
advocates can use the same information
38:45
and say , wait a minute , no , it isn't , let
38:47
me share with you . And
38:49
so it just enables
38:52
different parties to
38:54
use the same data and analyze
38:56
it in different ways in a very efficient
38:59
manner . It's not weeks
39:01
of pouring through boxes . They're calling other jurisdictions
39:04
and deposing people asking
39:06
endless data requests
39:08
. It's all right there and
39:11
now , in a matter of a day
39:14
or two , I can get what I need . And
39:16
now I'm sifting
39:18
through it and really thinking about the
39:20
results of the analyses I'm getting and
39:22
putting together my testimony one way or the other
39:25
. Is that fair ?
39:26
That is fair .
39:27
Okay , okay , that's interesting , all right . Okay
39:30
, how
39:33
does H data
39:35
impact this
39:37
notion of privacy and confidentiality
39:40
In the world of energy and
39:42
water and utilities ? We'll
39:44
talk about this in a minute . But we've got
39:46
a lot of smart , this and that
39:48
. We've got the internet of things . We've got
39:52
historically confidential information
39:54
that gets filed with commissions or environmental
39:56
regulators . Is
39:59
the filing of confidential information
40:01
kind of a bit of a limitation or not
40:04
? And then , how
40:06
does this address other
40:09
privacy issues that you may have , even
40:11
if you're gathering data from international
40:13
sources where there may be different privacy
40:16
or confidentiality regimes ? So
40:19
your thoughts ?
40:21
Our platform only works with what's public we
40:26
do . Our platform does have what
40:28
I mentioned before as the TurboTax of
40:30
utilities . That allows our
40:32
customers , if they have to file forms
40:35
, they can do it in our platform . We protect
40:37
the information that is
40:39
entered in there until
40:41
it is filed . When it's filed , it
40:43
becomes public . So there's a very big difference
40:45
between information that is being
40:48
entered into our platform preparing
40:50
for a filing , that is protected
40:53
in many different ways , and information
40:55
that then , once
40:57
the file button gets hit and it shows up crucial
40:59
point it shows up publicly
41:02
accessible on the regulators website . After
41:04
that it can be used for anything at all .
41:07
So if I am a utility
41:09
using a consult , an engineering
41:11
firm that does work around the
41:13
world , if there is public
41:16
information available in Australia
41:18
, canada , the European Union , you
41:20
could get that . Is that
41:22
fair ? If I am a regulator , if I am a utility
41:25
, if I am a consumer , I would say , hey , I got this
41:27
information from one of your consultants , where
41:29
they did work in Sydney
41:31
and this is what they charge . This is
41:33
what they did , and we think
41:35
your prices are a lot higher or what
41:37
have you . Why
41:39
are they doing it this way in this
41:41
jurisdiction in America ? Is that a possibility
41:44
?
41:45
It is a possibility . We are relying here on choices
41:47
that we make as democratic societies
41:49
. The choice that we make in the utility
41:52
sector is that we
41:54
are not going to have the government run power usually
41:57
a lot of municipals and all that
41:59
aside . We are not going to have the government run it . We are going to have
42:01
private companies run it because we think that they might be better
42:03
at it , and we are going to grant them monopolies
42:05
in some cases on predicting your aspects
42:07
of the value chain of moving electrons around
42:09
, and then , in exchange for getting that monopoly
42:12
, they are going to make money off of our rate payers . They
42:15
are also going to need to fulfill transparency
42:17
requirements . They are going to need to publish their
42:20
reports , they are going to need to file their reports and let those
42:22
reports be made public , and that is more transparency
42:24
than a private sector organization might face
42:26
. And when they want to charge more , they are going to have to
42:28
go and ask for it in public and explain the reasons
42:30
in public . We have already made those choices
42:33
as a democratic society that
42:35
publicity is
42:38
going to go along with monopoly , in
42:41
order to provide a restraint on
42:44
the possible self-interest that a monopoly
42:46
in the private sector is inevitably going to exercise
42:48
, and so
42:51
the technology is just going to make that work better . And
42:54
if we discover that the transparency
42:56
is too much , like you suggested earlier , matt
42:58
, if we discover the transparency is too
43:00
much , then I
43:02
am sure there are countervailing forces
43:05
that might try to reduce some of
43:07
the transparency , and the way to do
43:09
that is through the democratic process .
43:11
Yeah , all
43:14
right . What are the future
43:16
? I'll call it reverse opportunities with
43:18
customer data . We have a lot of activity
43:21
on this Internet of Things discussion
43:24
. We have smart
43:27
devices in the home smart meters , smart
43:29
toilets , what else
43:31
? Smart refrigerators , smart
43:34
thermostats . Is there
43:36
the ability in the future
43:38
for Hdata to take
43:42
those data in
43:44
some capacity and work them back into
43:47
a particular utility's
43:50
data analysis
43:52
for filings
43:55
with the commissions or public staff ?
43:58
We are going to maintain our focus on regulatory
44:00
data , so we would not touch those sources unless
44:03
the regulatory tour starts requiring the utility
44:05
to report them .
44:07
Okay .
44:08
Your client you are able to get good
44:10
at it , but by limiting that focus
44:12
. That's how we're able to get good at it . By
44:14
limiting our focus to the filings
44:17
going up and the dockets coming down , we
44:19
are able to develop a
44:21
data store that is pretty comprehensive and we're
44:23
able to build AI technology
44:26
that is effective .
44:28
And we achieve that effectiveness by limiting
44:31
the scope , the tool
44:33
is data source
44:35
agnostic , where a utility could
44:37
gather those data . Let's
44:40
say I had a smart meter and I had 100,000 customers
44:42
. I could gather those data and
44:45
I could put it into the tool and
44:48
have it help me run different analyses
44:50
that might help me going forward
44:52
with a particular regulatory filing . Is
44:54
that fair ? Is that a scenario that is
44:56
out there ?
44:58
No , the structured data part of our platform
45:00
is really aimed at the specific
45:02
data fields that come
45:05
up in the regulatory reports , and
45:07
so there wouldn't be much of a head start there . On
45:09
the unstructured side , our
45:12
customers have a private
45:15
catalog function in our regulatory AI , where
45:17
our customers can upload any document they want
45:19
and they can apply our regulatory
45:21
AI to it . But we have not
45:23
tested the regulatory AI on everything
45:25
. We've only tested it , optimized
45:28
it , on regulatory documents , and
45:30
the reason for that is that we
45:32
want to make sure to effectively
45:35
serve the people that need to find
45:37
summaries or insights
45:40
, analysis or recommendations from
45:42
this specific corpus of regulatory
45:44
information .
45:47
Okay , well , in
45:49
10 to 20 years , if we
45:51
see your
45:53
vision expand
45:56
from energy into water , what does
45:58
H data look like in
46:00
the next 10 to 20 years when it comes to water
46:02
? What are three things in your mind that
46:04
you see for the future of H
46:06
, data and water ?
46:08
If this were to happen , then the
46:11
three things I would see are number one anything
46:14
you have to file with
46:16
a regulator or if you're
46:18
regulated and you're working on a filing , information
46:21
flows automatically from your source systems
46:23
that you've already got to the regulator that
46:25
wants them . You don't have to compile
46:27
a form . The form just happens
46:30
in the background , you sign off on it , but
46:32
you don't have to write it . Number
46:35
two if you are running
46:37
efficiency comparisons or
46:40
if you are trying to predict future
46:42
rates assuming that the rates are formulaic then
46:45
you have all those comparisons and formulas
46:47
in H data . They just exist
46:50
and you can go access them and you can
46:52
watch how they change and you can set alerts
46:54
in case they change a certain way In
46:56
near real time . Well
46:59
, we take 10 minutes to get there .
47:01
Near real time . This
47:03
isn't monthly or weekly
47:06
, this is not even daily . It's you set
47:08
us every 10 minutes .
47:10
We're already doing that . Within
47:13
10 minutes of a form being filed , all the
47:15
numbers go from that form into our data
47:17
set and are reflected in our tools .
47:19
It's like NASDAQ in a sense the ticker
47:21
Wow .
47:24
And then , third , one
47:26
thing I haven't touched on this in this conversation so far
47:28
, Matt that we're excited about is
47:30
that I mentioned how the AI is
47:32
really for the unstructured documents , and
47:36
that is true . It is true for today , but
47:39
within much fewer
47:41
than 10 years , within one year , regulatory
47:43
AI will be able to answer the structured questions as well
47:45
. So you could say to regulatory AI tell
47:48
me the return on equity of every utility in
47:50
the American Southeast with revenue of over
47:52
a billion a year , and it will draw
47:54
you a table and the table will
47:56
be accurate . It will calculate return
47:59
on equity for you using numbers that come from the filings
48:01
, and then you can click on the number and see where
48:04
it came from .
48:05
Wow . So if I'm a consumer
48:07
advocate , I can use the
48:09
function to say tell me what the
48:12
problems are with this utility , or
48:14
et cetera , et cetera . Where
48:17
have things gone wrong ? I'm
48:19
kind of teasing , but
48:22
it really seems to
48:25
shorten and focus
48:27
this whole world of
48:30
utility regulation and
48:33
it changes the world , accelerates
48:35
, adds velocity and transparency
48:38
to this thing that we've
48:40
been dealing with for many decades .
48:45
Other worlds have been shortened and focused already
48:47
Since . Look at how the
48:50
ability to invest in publicly traded companies
48:52
expanded once online
48:54
trading became possible . This
48:57
didn't necessarily make it better
48:59
or worse , just made it faster , and
49:01
I do think it made it more democratic
49:03
. We now don't have to hire a
49:05
stockbroker , so anybody can invest in stocks
49:07
. Like I mentioned , though good or
49:09
bad , because we now have people investing in stocks
49:12
in meme stocks that maybe shouldn't .
49:16
Well , you're right . Anybody
49:18
can be a movie producer now You've
49:21
got YouTube . You can be an influencer
49:24
on TikTok . You
49:27
have apps . I don't have to go through a travel agent
49:29
to reserve
49:31
my seat on a plane . I can do all these
49:33
things through advances in technology , and the
49:36
utility world's a bit behind , but it's catching up . The
49:39
energy side is probably moving faster than the water
49:41
side . Those are
49:43
some great ideas . Your
49:46
first observation , I think , is noteworthy
49:50
in that you have this near
49:53
real-time regulatory system
49:55
where you
49:57
have the economic and perhaps in the
49:59
future , the environmental regulator , knowing
50:02
the challenges before
50:05
you do . Almost they're seeing it real time
50:07
. Uh-oh , there's an unpermitted discharge
50:09
, or uh-oh , the sensor said
50:11
that the drinking water's got a problem
50:13
, and you don't have to wait
50:16
to fill out a report or notify anybody and
50:18
it will go not just to the regulator , it might go to
50:20
the customer as well if those
50:22
things are triggered , the
50:25
reporting and things like that . Lots
50:27
to think about with this
50:29
tool , wow . Well
50:31
, hudson , I want to thank you for being a guest today
50:34
on the WaterForceSide podcast . It has
50:36
been a privilege , a fascinating
50:38
discussion . I probably have 10
50:40
questions right now , probably have another 10
50:42
after we're done , but you
50:45
have really made
50:47
us think about the future of
50:49
water from a very unique perspective
50:52
. How can folks get ahold of you if they want
50:54
to learn a bit more about age data
50:56
?
50:56
You can find us at agedataus and
50:59
it's pretty easy to find me on LinkedIn . There's only
51:01
two Hudson Hollisters and I'm one of them .
51:04
Thank you for listening to the WaterForceSide podcast
51:07
powered by the Aqualaurus Group . For more
51:09
information , please visit us at Aqualauruscom
51:12
or follow us on LinkedIn
51:14
and Twitter .
Podchaser is the ultimate destination for podcast data, search, and discovery. Learn More