Feeling embarrassed to work on Otto. What to do :(
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
3 weeks ago - last edited 2 weeks ago
In my current company, I have been asked to work on Otto products.
After looking at the OOTB AI agents/Agentic workflows, I feel like banging the person who allowed to publish these agents. I feel like switching my career to some other better platform/technology.
----------------See these examples (just a few listed here ok)---------------
1. Explain SLA use case - we can ask question like 'will this incident breach?' - AI agent sees if business elapsed time has crossed 70%, then it says YES. if not it says NO.
WOW, what a use case!!!!
2. SLA breach predictor - who caused this SLA to breach?
Suppose resolution SLA is 1hr. User A has it for 59 mins, reassigned to user B who has it for 2 mins, AI agent tells User B breached it.
WOW!!!
3. Incident Assist use case - it will give the summary of incident data including the problem and change request attached to it. But what details you know? PRB123 is attached to this incident. CHG123 is attached to this incident. Thats it 😄
Again, WOW use case!!!!!
4. Change outage assistant AI agent - I have a change request where primary ci is Windows server 'X'.
This AI agent helps me to create a outage record for this windows server X.
Awesomeeeeeeeee AI agent!!!! best AI agent...
5. Wrap up and resolve use case - one agent will generate resolution notes. another agent will create knowledge article. We already have respective gen ai skills. I was wondering what is the difference between those skills and these AI agents.
Then I see that the AI agent is using those skills only as inputs.
Again, WOW use case!!!!!
6. Schedule a change use case - it will check the Maintenance window tagged to the CI and then tell give some suggestions around planned start and end date.
Ex - For a change request, I have mentioned planned start date as 9am, end date as 6pm on Friday,
The maintenance window attached to the CI is> Only Sunday, 6am-7am
This AI agent did not even check the day (which is Sunday). It told me that the allowed duration is 1 hour.
So I am proposing the end time is 10 am.
Again, the best use case I have come across. WOW!!!
7. Generate change request plans
It looks into the short description and description of current change and finds similar changes and curates planning section details.
Just short description and description???? No need to consider the CI, Service, offering, and other important fields?
Tested this in multiple customer instances. the AI agent could hardly find right/relevant past changes which are of similar types. Jesus!
How did this even under Agentic evaluation and passed?
Sh** use cases.................... the list goes on and on and on and on.......
I have been asked to help customers to implement these use cases. but when I start to see the use cases, my bloodd boils like anything.
I want to meet the great people gave these use cases and approved these use cases and give them great awards.
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
2 weeks ago
I can understand your frustration, especially when you're expected to implement these use cases for customers and the results don't match the expected business outcome.
I think the important distinction here is that having an AI agent execute a workflow doesn't necessarily mean the workflow is intelligent. Your examples raise valid questions around context awareness, data quality, decision-making and how these agents are evaluated.
For example, an SLA prediction should consider the complete task history and reassignment timeline, not just the current assignee. Similarly, change planning should consider CI, service, maintenance windows, risk and historical changes rather than relying only on short description and description.
Also, using existing GenAI skills inside an AI agent isn't necessarily a problem. Skills perform specific tasks, while agents can orchestrate multiple tools, gather context and decide what to do next. The real question is whether that orchestration adds meaningful value.
ServiceNow itself allows you to duplicate and customize preconfigured agentic workflows to meet customer requirements.
I'd be interested to know whether others have tested these OOTB Otto agents in real customer environments. Are you seeing similar limitations, or have you managed to get better results through customization?
References: