As it stands today; just a bit of complexity is all that is required to make AI Agents fail. I expect the gap to narrow over the years of course. But capturing complex business logic and simplifying it will probably be useful and worth paying for a long time into the future.
This means for any "manual" or existing workflow requiring a access to several systems, that requires multiple IT permissions with defined scopes. Even something as simple as a sales rep sending a DocuSign might need:
- CRM access
- DocuSign access
- Possibly access to ERP (if CRM isn't configured to pass signed contract status and value across)
- Possibly access to SharePoint / Power Automate (if finance/legal/someone else has created internal policy or process, e.g. saving a DocuSign PDF to a folder, inputting details for handover to fulfilment or client success, or submitting ticket to finance so invoicing can be set up)
I never understood the evolvement around agents, they just appeared to me as Python scripts initially (Crewai 2-3 years ago).
The question is can people see that agents will evolve? Similar to how software evolves to handle the right depth of granularity.