1
00:00:02,279 --> 00:00:05,720
Speaker 1: Since you're a subscriber to this Bloomberg podcast, we thought

2
00:00:05,760 --> 00:00:09,240
Speaker 1: you'd be interested in a new four episode sponsored podcast

3
00:00:09,360 --> 00:00:14,400
Speaker 1: called The ROI Rules of Ai, produced by IBM and

4
00:00:14,440 --> 00:00:18,520
Speaker 1: Bloomberg Media Studios. It explores how business leaders are thinking

5
00:00:18,560 --> 00:00:23,759
Speaker 1: about the return on investment of artificial intelligence projects. You

6
00:00:23,800 --> 00:00:28,200
Speaker 1: can subscribe wherever you listen to your favorite podcasts. Here's

7
00:00:28,320 --> 00:00:33,280
Speaker 1: a recent episode. If your company is like most, after

8
00:00:33,320 --> 00:00:37,800
Speaker 1: a contract is negotiated and signed, congratulations, it gets filed

9
00:00:37,800 --> 00:00:41,720
Speaker 1: away in a lawyer's office, not to be consulted again

10
00:00:41,840 --> 00:00:44,479
Speaker 1: until it has run its course and needs to be renewed.

11
00:00:47,920 --> 00:00:50,280
Speaker 1: But what if your contracts could be a roadmap to

12
00:00:50,360 --> 00:00:54,200
Speaker 1: running your business the first step in managing your relationships

13
00:00:54,200 --> 00:00:58,480
Speaker 1: with your suppliers and clients. That's the function of Syrion,

14
00:00:59,160 --> 00:01:03,320
Speaker 1: a contract life cycle management software for in house legal departments.

15
00:01:04,080 --> 00:01:08,160
Speaker 1: It's recently been upgraded with generative AI functionality designed to

16
00:01:08,200 --> 00:01:11,319
Speaker 1: make it something the whole company can use, not just

17
00:01:11,520 --> 00:01:12,480
Speaker 1: its legal staff.

18
00:01:12,959 --> 00:01:15,360
Speaker 2: It gives you real time information on how to run

19
00:01:15,400 --> 00:01:17,760
Speaker 2: your business and gives you the insights you need to

20
00:01:17,760 --> 00:01:20,399
Speaker 2: pivot so that you can make the best decisions with

21
00:01:20,480 --> 00:01:22,160
Speaker 2: the best data at the right time.

22
00:01:22,640 --> 00:01:26,560
Speaker 1: That's Gordon Thompson, Syrian's executive vice president of pre sales

23
00:01:26,600 --> 00:01:29,800
Speaker 1: and business strategy. He and his colleagues are planning a

24
00:01:29,880 --> 00:01:33,600
Speaker 1: future where contracts are integrated into a company's day to

25
00:01:33,680 --> 00:01:40,000
Speaker 1: day operations. From IBM and Bloomberg Media Studios, this is

26
00:01:40,040 --> 00:01:44,320
Speaker 1: the ROI Rules of AI and I'm your host Edward Adams.

27
00:01:45,200 --> 00:01:48,400
Speaker 1: On this podcast, we're exploring how companies of all sizes

28
00:01:48,680 --> 00:01:52,720
Speaker 1: are using AI to remake their operations, increasing their return

29
00:01:52,760 --> 00:01:58,880
Speaker 1: on investment and that of their customers. Syrian has been

30
00:01:58,880 --> 00:02:02,680
Speaker 1: in business for a decade. It enables companies to digitize

31
00:02:02,720 --> 00:02:07,200
Speaker 1: their universe of contracts into a single database, extracting metadata

32
00:02:07,320 --> 00:02:12,400
Speaker 1: like specific clauses and obligations to make them searchable. It

33
00:02:12,440 --> 00:02:16,520
Speaker 1: also helps lawyers draft new contracts, making sure that the

34
00:02:16,639 --> 00:02:21,000
Speaker 1: terms limit their company's risk, and it enables a general

35
00:02:21,040 --> 00:02:25,760
Speaker 1: counsel's office to measure the company's performance against those contracts.

36
00:02:26,560 --> 00:02:29,880
Speaker 1: Unlike a lot of companies, Syrion has a long history

37
00:02:29,960 --> 00:02:32,240
Speaker 1: of working with artificial intelligence.

38
00:02:32,840 --> 00:02:36,680
Speaker 2: From our inception, we were using natural language processing. As

39
00:02:36,840 --> 00:02:41,360
Speaker 2: the technology evolved over the years, we started using machine learning.

40
00:02:41,800 --> 00:02:45,799
Speaker 2: We've built over six hundred small and medium sized language

41
00:02:45,880 --> 00:02:49,040
Speaker 2: models that are purpose built to extract very specific pieces

42
00:02:49,040 --> 00:02:52,480
Speaker 2: of data out of contracts at a very high precision rate.

43
00:02:53,280 --> 00:02:56,640
Speaker 1: But when it wanted to improve its contract review function

44
00:02:56,840 --> 00:02:59,520
Speaker 1: and turn contracts into a device to run a business,

45
00:03:00,160 --> 00:03:04,359
Speaker 1: needed to employ generative AI. It had to move from

46
00:03:04,360 --> 00:03:08,120
Speaker 1: simply copying and pasting an approved clause in place of

47
00:03:08,160 --> 00:03:12,480
Speaker 1: a riskier one to rewriting the proposed contract as a

48
00:03:12,560 --> 00:03:16,560
Speaker 1: lawyer might do. That meant using a large language model

49
00:03:16,760 --> 00:03:19,600
Speaker 1: or LM. The problem is.

50
00:03:20,560 --> 00:03:23,919
Speaker 2: Lms are expensive to build and maintain, so we wanted

51
00:03:23,960 --> 00:03:28,640
Speaker 2: to create a solution that was scalable, that was cost effective.

52
00:03:34,920 --> 00:03:39,440
Speaker 1: To understand Syrian solution, it helps to first understand what

53
00:03:39,560 --> 00:03:45,120
Speaker 1: contract lawyers actually do. Lawyers call reviewing a contract redlining,

54
00:03:45,680 --> 00:03:47,680
Speaker 1: and like a lot of things in the world of law,

55
00:03:48,040 --> 00:03:50,920
Speaker 1: it's got some history to it. For those of the

56
00:03:50,960 --> 00:03:53,960
Speaker 1: audience who are not lawyers, what does redlining a document mean?

57
00:03:54,360 --> 00:03:56,480
Speaker 2: So redlining got its name from when we used to

58
00:03:56,520 --> 00:03:58,760
Speaker 2: do paper contracts, right, so you would send a paper

59
00:03:58,800 --> 00:04:01,920
Speaker 2: contract over to your counter party. They would literally take

60
00:04:01,960 --> 00:04:03,800
Speaker 2: a red pen or red pencil, and that they would

61
00:04:03,800 --> 00:04:06,600
Speaker 2: review it and strike through it and make changes. That

62
00:04:06,720 --> 00:04:10,960
Speaker 2: name has really persisted into the digital age. Ultimately, redlining

63
00:04:11,080 --> 00:04:16,560
Speaker 2: it boils down to making sure that you identify issues, risk, compliance,

64
00:04:17,160 --> 00:04:20,760
Speaker 2: terms and conditions that meet your ultimate goals as an organization.

65
00:04:21,680 --> 00:04:25,520
Speaker 1: When lawyers redline by hand, it can be a slow process.

66
00:04:26,080 --> 00:04:28,760
Speaker 2: On average, it takes about thirteen minutes to review a

67
00:04:28,800 --> 00:04:31,680
Speaker 2: page of a contract. Some of these contracts can be

68
00:04:31,760 --> 00:04:35,080
Speaker 2: several hundred pages long. When you automate it using jenerative

69
00:04:35,120 --> 00:04:38,840
Speaker 2: AI that goes from days weeks down to minutes and seconds.

70
00:04:39,680 --> 00:04:43,360
Speaker 1: To speed up the process, Syrian used an open source

71
00:04:43,839 --> 00:04:48,600
Speaker 1: LLM and a series of proprietary playbooks which spell out

72
00:04:48,640 --> 00:04:52,400
Speaker 1: a company's default positions on a variety of legal issues.

73
00:04:53,080 --> 00:04:55,279
Speaker 1: Can you give me an example of what you mean

74
00:04:55,320 --> 00:04:56,120
Speaker 1: by a playbook?

75
00:04:56,440 --> 00:04:59,440
Speaker 2: Your preferred position might be I want payment in thirty days.

76
00:05:00,120 --> 00:05:03,520
Speaker 2: The counterparty says that payment terms might be sixty days,

77
00:05:03,880 --> 00:05:05,640
Speaker 2: So you want to be able to identify that very

78
00:05:05,680 --> 00:05:08,360
Speaker 2: quickly and make sure that you put in your preferred language.

79
00:05:08,680 --> 00:05:12,200
Speaker 1: Syrian can redline even some of the longest contracts in

80
00:05:12,240 --> 00:05:15,839
Speaker 1: a matter of minutes, and while human lawyers still review

81
00:05:15,880 --> 00:05:19,440
Speaker 1: the software as changes. Companies have found they can go

82
00:05:19,520 --> 00:05:24,239
Speaker 1: from first draft to final agreement thirty to eighty percent faster.

83
00:05:24,640 --> 00:05:28,920
Speaker 1: Thompson says, once those contracts are signed, it took another

84
00:05:29,040 --> 00:05:31,719
Speaker 1: form of AI to open up a company's database of

85
00:05:31,800 --> 00:05:38,160
Speaker 1: contracts to the wider organization. IBM's conversational search technology allows

86
00:05:38,240 --> 00:05:41,679
Speaker 1: designated people in a company's lines of business to ask

87
00:05:41,800 --> 00:05:46,200
Speaker 1: questions of the Syrian database and they get answers in

88
00:05:46,360 --> 00:05:47,480
Speaker 1: plain English.

89
00:05:47,920 --> 00:05:50,280
Speaker 2: If you're a chief procurement officer and you want to

90
00:05:50,360 --> 00:05:54,520
Speaker 2: understand a certain piece of data around your supplier contracts,

91
00:05:54,880 --> 00:05:57,320
Speaker 2: it empowers you to go out and get that information

92
00:05:57,480 --> 00:06:00,039
Speaker 2: versus having to rely on legal to go get that

93
00:06:00,080 --> 00:06:01,000
Speaker 2: information for you.

94
00:06:01,880 --> 00:06:05,240
Speaker 1: That makes contracts into a tool to help manage the business.

95
00:06:05,279 --> 00:06:14,960
Speaker 1: Thompson says the ability to use and open source LM,

96
00:06:15,520 --> 00:06:19,080
Speaker 1: something not every AI platform enables, was one of the

97
00:06:19,120 --> 00:06:24,839
Speaker 1: reasons Sirion chows IBM, Thompson says, So was its strength

98
00:06:25,040 --> 00:06:28,000
Speaker 1: in conversational search and the fact.

99
00:06:27,720 --> 00:06:31,119
Speaker 2: That IBM is a pioneer in the I space, which

100
00:06:31,279 --> 00:06:33,760
Speaker 2: is a big plus for us. They bring a lot

101
00:06:33,800 --> 00:06:37,840
Speaker 2: of thought leadership to AI and generative AI. We were

102
00:06:37,880 --> 00:06:40,960
Speaker 2: looking at tools that we could implement on We decided

103
00:06:41,000 --> 00:06:42,880
Speaker 2: that Watson X was one of those tools that we

104
00:06:42,920 --> 00:06:45,760
Speaker 2: could really go to market with and build a relationship

105
00:06:45,760 --> 00:06:49,359
Speaker 2: with IBM. They are open architecture, so we can use

106
00:06:49,560 --> 00:06:53,200
Speaker 2: other types of technologies inside of the Watson platform and

107
00:06:53,240 --> 00:06:57,400
Speaker 2: governance with using their Watson X platform.

108
00:06:57,920 --> 00:07:01,360
Speaker 1: For its part, IBM is not just supplier to Syrion,

109
00:07:01,760 --> 00:07:04,560
Speaker 1: but a client of the company too, using its enormous

110
00:07:04,640 --> 00:07:08,039
Speaker 1: database of contracts to help run its operations.

111
00:07:08,760 --> 00:07:11,640
Speaker 2: IBM they have over half a million contracts that we've

112
00:07:11,800 --> 00:07:15,320
Speaker 2: ingested into the solution. So from a scale standpoint, making

113
00:07:15,320 --> 00:07:17,840
Speaker 2: sure that our generative AI solution can stand up to

114
00:07:17,880 --> 00:07:21,280
Speaker 2: that scale has been a great journey that we've gone

115
00:07:21,280 --> 00:07:22,520
Speaker 2: on with IBM.

116
00:07:22,920 --> 00:07:26,480
Speaker 1: The lesson for other companies like Syrion companies which are

117
00:07:26,520 --> 00:07:30,680
Speaker 1: neither startups nor enormous enterprises, but part of the vast

118
00:07:30,760 --> 00:07:34,360
Speaker 1: middle of American businesses, is that they probably have the

119
00:07:34,440 --> 00:07:38,440
Speaker 1: data they need to get started with generative AI, says

120
00:07:38,520 --> 00:07:43,880
Speaker 1: Garghie Dasgupta, director of Product Management and AI leader at IBM.

121
00:07:44,320 --> 00:07:48,000
Speaker 3: The lesson that we can learn from Syrians experience is

122
00:07:48,040 --> 00:07:52,320
Speaker 3: that there is data that already exists, and it's super

123
00:07:52,440 --> 00:07:58,400
Speaker 3: critical to democratize access to that data. What because it

124
00:07:58,640 --> 00:08:04,320
Speaker 3: unlocks business value and business insights that were not possible before.

125
00:08:05,080 --> 00:08:08,920
Speaker 1: While small companies tend to outsource AI projects, and large

126
00:08:09,040 --> 00:08:11,840
Speaker 1: enterprises have teams of engineers who can take them on.

127
00:08:12,400 --> 00:08:16,680
Speaker 1: Mid sized companies generally need to upscill their employees to

128
00:08:16,800 --> 00:08:18,200
Speaker 1: take on an AI project.

129
00:08:18,320 --> 00:08:22,080
Speaker 3: Desk Grouped says, I would say Syrian is an exception

130
00:08:22,240 --> 00:08:25,720
Speaker 3: to that rule because they've been native AI for yours.

131
00:08:26,640 --> 00:08:29,480
Speaker 1: When it comes to advising other companies on what to

132
00:08:29,560 --> 00:08:34,520
Speaker 1: look for in AI partner, Thompson says, depth of experience

133
00:08:35,040 --> 00:08:36,319
Speaker 1: is a vital consideration.

134
00:08:36,880 --> 00:08:39,600
Speaker 2: AI is not just one type of technology, right. There's

135
00:08:39,679 --> 00:08:43,800
Speaker 2: lots of different ways to implement AI. Organizations that don't

136
00:08:43,800 --> 00:08:47,680
Speaker 2: have that historical background typically have a myopic view of

137
00:08:47,720 --> 00:08:50,280
Speaker 2: how to leverage AI. You want to make sure that

138
00:08:50,360 --> 00:08:53,160
Speaker 2: they understand all the options available to you and apply

139
00:08:53,280 --> 00:08:57,280
Speaker 2: the appropriate technology or sub discipline of AI to your

140
00:08:57,280 --> 00:08:57,920
Speaker 2: business case.

141
00:08:58,760 --> 00:09:02,920
Speaker 1: But don't let the challenge of deploying generative AI convince

142
00:09:03,000 --> 00:09:04,360
Speaker 1: you to put off the work.

143
00:09:04,640 --> 00:09:10,200
Speaker 2: Thompson warns, start small, think big, go fast. My advice

144
00:09:10,240 --> 00:09:12,840
Speaker 2: would be, don't be afraid of the technology. Make sure

145
00:09:12,880 --> 00:09:16,360
Speaker 2: that you're investigating it, doing your homework, getting smart around

146
00:09:16,400 --> 00:09:18,840
Speaker 2: what the use cases are that generative AI can deliver

147
00:09:18,960 --> 00:09:21,640
Speaker 2: to you. But stick your toe in the water. Generative

148
00:09:21,640 --> 00:09:24,200
Speaker 2: AI is moving literally at the speed of light. We

149
00:09:24,240 --> 00:09:28,079
Speaker 2: see new innovations coming out literally every day. So if

150
00:09:28,080 --> 00:09:30,280
Speaker 2: you're a laggard and you don't start looking at how

151
00:09:30,320 --> 00:09:32,760
Speaker 2: you can use generative AI, you're going to fall so

152
00:09:32,840 --> 00:09:34,880
Speaker 2: far behind that it'll be tough to catch up.

153
00:09:38,400 --> 00:09:42,199
Speaker 1: This has been the ROI Rules of Ai, a podcast

154
00:09:42,240 --> 00:09:46,360
Speaker 1: from IBM and Bloomberg Media Studios. If you like what

155
00:09:46,400 --> 00:09:50,920
Speaker 1: you hear, subscribe and leave us a review. I'm Edward Adams,

156
00:09:51,400 --> 00:09:52,240
Speaker 1: Thanks for listening.

157
00:10:00,679 --> 00:10:02,360
Speaker 2: Why You'll love me