Microsoft, automate my monthly expense report.
What even is a Co-Pilot?
Microsofts Co-Pilot is an AI-companion that allows us to interact with the various tools of Microsoft365 suite via natural language in a chat interface.
The companion will be able to assist, perform tasks and support you in your daily work. Here are some examples of things that Microsoft Co-Pilot could assist you with:
As per October 2023, the new Windows 11 Update incorporates more than 150 co-pilot features already, including features in Windows, Notepad, Snipping tool, Outlook, File Explorer and many more.
Here are some scenarios, in which Microsofts Co-Pilot could support you:
- You want to catch up with emails
What was the latest information about the license price we received from the supplier? I can not find the email right now. - You want to write an article on the intranet or in teams to inform about a new project rollout (project moonshot)
Please draft a one-pager project summary for ‘Project Moonshot’ using the information in file1, file2 and file3. Ensure it is easy to understand and sound friendly but formal. - You are planning a workshop
Create an agenda for a full-day workshop to rethink our team strategy. The goal is to prioritze our activities based on the corporate strategy. We have 15 attendees and it should be possible to attend digitally
RPA-Companions
Automation Companions are already here. Microsofts Power Platform already allows to generate flows in Power Automate from natural language prompts in preview mode, i.e.
- Create a task in Planner whenever I receive a mail from xy
- Download attachments from sender ‘xy’ and save them to a OneDrive folder
While there are no AI-companions for RPA / desktop-automation yet, it is a matter of when not if they will arrive.
So what could this mean for RPA-development?
Personally, I consider mainly two scenarios on how AI-companions will impact RPA:
Scenario A: RPA-Software AI-companions
In this scenario AI-companions are integrated into RPA development software like Power Automate Desktop, UiPath, AutomationAnywhere, etc. Essentially, this extends the features Power Automate already provides to desktop applications.
These companions will allow RPA-developers to develop solutions faster, more efficiently and more robust by using a natural language interface. The extent of this capability depends heavily on the maturity of the AI-models and could range from basic project structuring to entirely auto-generated flows.
Sample prompts could look like this:
- CREATE: Create a flow that downloads all .pdf-documents from application x, extracts the information and then publishes the data in application y based on the input field names.
- IMPROVE: Structure my project in subflows. Eliminate all unrequired steps and re-name my variables, so that they are easily understandable
- DEBUG: Check my flow for potential error-causes and provide some ideas for improvement, i.e timeouts, try-catch or retry scopes
This scenario brings low-code closer to no-code, further reducing the level of technical understanding and skills required for automation and switch the focus of RPA developers more towards process understanding and improvement. After all, the developer will still need to understand the process in all detail to be able to draft a solution and generate a proper prompt.
Scenario B: Software-native companions
With growing maturity and popularity AI-companions will eventually also find their way into enterprise tools themselves, meaning that the tools will offer their own integrated functionalities to automate tasks. In a way, this phenomenon has been around for a while already. If you think of Salesforce or SAP, most tasks can be automated without the need external tools such as RPA or development.
This eliminates big parts of the need for RPA in modern enterprise applications. The major advantage lies in the likely utilizion of the backend and API of the involved applications, ensuring robust automations without requiring frontend interaction.
However, this type of automation may be more suited for personal productivity enhancement rather than enterprise-solutions, which require high-level access- and system rights.
Moreover, it still requires bridges or connectors (e.g. RPA) to enable interactions between the various systems and their integrated solutions. Legacy systems, prevalent today and likely in the future, will also futher require external solutions for process automation.
Conclusion
I think, Co-pilots are here to stay and both scenarios are bound to unfold one way or another sometime in the near future. The impact however is highly dependent on the model-maturity. Everyone who has used recorders and similar tools in the past would probably agree, that there is substantial work remaining afterwards.
Ultimately, these co-pilots are designed to enhance not replace RPA. Although the focus may shift, I am confident, that there will continue to be opportunities and the need for RPA. By increasing the accessibility and popularity of automation, this could even serve as a catalyst, propelling RPA and automation to new heights
