ServiceNow AI Agents in Playbooks | Developer Passport Highlight — Transcript

Full transcript

1. 0:00This feature is taking AI agents and
2. 0:03supercharging your business processes by
3. 0:06by bringing AI agents in as specific
4. 0:08steps of your playbook.
5. 0:10The cool thing that we're showcasing as
6. 0:12part of this feature is we now have
7. 0:16this awesome step where we're going to
8. 0:19inject an AI agent into this process.
9. 0:22So, let me just show how someone would
10. 0:24add a new one of these to their process.
11. 0:28You can see here in our activity picker
12. 0:31that we have a new category called AI
13. 0:33activities, and we have a use an AI
14. 0:36agent activity definition.
15. 0:39Now, when I choose this, let's pretend I
16. 0:42did that and I added this step to my
17. 0:44playbook,
18. 0:46what I have to do to get this ready for
19. 0:48primetime
20. 0:50is pick the AI agent that I'm going to
21. 0:52use. So, this simply unlocks the power
22. 0:56of AI agents within a business process.
23. 1:00So, what's going to happen here as we
24. 1:02kick off our complaint playbook, there
25. 1:04will be a few steps that a human may
26. 1:05complete, and then there will be a
27. 1:07specific step in our research stage
28. 1:09where an AI agent is going to take over.
29. 1:11It's going to complete its work and help
30. 1:13us accelerate case resolution in a
31. 1:16complaint scenario. So, we've already
32. 1:18pre-filled our reference to the
33. 1:21complaint case research plan agent. We
34. 1:24can give the agent an objective. In this
35. 1:27case, we are pointing to the complaint
36. 1:30case record itself. So, we we want the
37. 1:32the AI agent to look at information on
38. 1:35the complaint case record. This agent
39. 1:37has been configured with its set of
40. 1:39instructions. So, it knows what it's
41. 1:40going to do when it looks at this
42. 1:42complaint case in runtime. We also have
43. 1:45a couple more configuration things that
44. 1:46I want to show, and then we'll talk
45. 1:48about the runtime experience. So, we
46. 1:51offer the ability to select the run as
47. 1:54strategy. The AI agent needs to run on
48. 1:58behalf of a specific user. And so you
49. 2:01have a couple of options. You can choose
50. 2:03the active user, whoever is currently
51. 2:06viewing the playbook, interacting with
52. 2:08the playbook in runtime. And then the
53. 2:09other option is you can run as the user
54. 2:12triggering the playbook. So this this
55. 2:14allows the AI agent to act on behalf of
56. 2:17a couple of different personas, and
57. 2:19whatever is right for your use case,
58. 2:20you'll be able to choose that. The
59. 2:21second thing that I want to call out is
60. 2:23outputs. So within workflow automation,
61. 2:27we rely heavily on outputs to support
62. 2:30dynamic process execution. So as one
63. 2:33step completes, that step can provide
64. 2:35outputs. And then those outputs can be
65. 2:38referenced in subsequent steps of the
66. 2:39playbook. So if you want to, for
67. 2:42example, have an AI agent draft an email
68. 2:45for you and provide the email draft as
69. 2:48an output, you could then supply that
70. 2:50output as an input to an email activity
71. 2:54where a human's going to essentially
72. 2:55fine-tune and and finalize the AI
73. 2:58agent's draft email before they send it
74. 3:01to the customer.
75. 3:02So here, you are able to provide the
76. 3:06expected outputs that the AI agent
77. 3:08should render as part of its work inside
78. 3:11this playbook. Uh here's my playbook. We
79. 3:14have started this by completing a couple
80. 3:17of the
81. 3:18uh manual steps. So these are forms.
82. 3:21We're essentially defining what the
83. 3:23nature of this complaint is. We've got a
84. 3:25product quality issue. And so there are
85. 3:28a couple steps that the case manager
86. 3:30will complete.
87. 3:31And now it's time for the AI agent to
88. 3:33take over.
89. 3:34Now, this is a helper agent. So what
90. 3:37that means is it is going to work with
91. 3:40human supervision. So you saw there that
92. 3:42we had to click the start button, but
93. 3:44worker agents are also available, and
94. 3:47they will operate autonomously. So what
95. 3:50we're seeing here is now the AI agent
96. 3:52doing its work and it's actually showing
97. 3:54us its progress within the activity
98. 3:57itself. So, I don't have to launch the
99. 3:59now assist panel. I don't have to launch
100. 4:01any other in-product experiences. We're
101. 4:03actually capturing the in-product AI
102. 4:06experience within the activity card
103. 4:08itself and then we're going to wait for
104. 4:09the next step. So, you can see here that
105. 4:11this helper agent asks, are you ready
106. 4:14for us to update the work notes? So, uh
107. 4:18as part of this AI agent's work, it is
108. 4:20going to write something to the work
109. 4:22notes of the case. It is asking if the
110. 4:24human is authorizing the AI agent to
111. 4:27proceed. The The beauty of using an AI
112. 4:29agent is they can surface dynamic
113. 4:32experiences within the activity card
114. 4:34itself. So, if it needs the human to
115. 4:36answer questions like a yes no, it will
116. 4:39surface these dynamic UIs with no
117. 4:42additional configuration uh for the
118. 4:44playbook author when they're designing
119. 4:46this step. It has done some research. It
120. 4:49has looked at other cases. It has uh
121. 4:53identified tasks that have been
122. 4:56generated and uh worked and completed in
123. 4:59order to resolve similar cases. And it
124. 5:03has suggested that there are a list of
125. 5:04tasks that it may need to complete in
126. 5:07order to uh resolve this case. And the
127. 5:11human once again is given the option to
128. 5:14move forward with these or you can see
129. 5:17there I have some freedom to dismiss any
130. 5:20of the the recommended tasks.
131. 5:24And that's it. So, you can see here that
132. 5:26the AI agent has not only done some
133. 5:30research on behalf of the human, it has
134. 5:32proposed case tasks and then has asked
135. 5:35the human for feedback as far as whether
136. 5:37we move forward with these case tasks or
137. 5:39not. So, I want to pause the demo there.
138. 5:42Why is this important? Why does this
139. 5:43matter? Um so,
140. 5:46agents are able to do things that take
141. 5:49away uh some of the burden of the human
142. 5:52agent. So, you can have an AI agent do
143. 5:55research or review
144. 5:57large large volumes of information in a
145. 6:00short amount of time, whereas, you know,
146. 6:03if a human had to do this previously,
147. 6:04this could take hours potentially. I
148. 6:06could be going through complaint cases
149. 6:08for hours to determine what my set of
150. 6:11tasks should be. AI agents can do this
151. 6:14work for me in a couple of moments, and
152. 6:16then I can focus on the the really
153. 6:19important things, like interfacing with
154. 6:21my client, ensuring that their needs are
155. 6:23being met, that they're being heard,
156. 6:25focusing on the human interaction, as
157. 6:27opposed to some of the more menial
158. 6:28tasks, like research.