Experience with implementing AI automations

July 08, 2024

by a professional from University of Groningen in Munich, Germany

I was wondering if anyone on this platform has implemented AI automation solutions. What has your experience been with them? Did they work as expected? Do they occasionally make mistakes?

I'm asking because we are trying out some AI solutions ourselves. We've noticed that AI can be super helpful with certain processes, but it tends to be error prone in others.

We are researching which types of processes are suitable for AI automation and which are not. We're also trying to understand why some processes are harder to automate with AI.

If you have any insights or experiences to share, I'd greatly appreciate it. Thank you!

0 Likes

4 Replies

20 Views

Replies

Reply by a professional
in Saskatoon, SK, Canada

Hi Dirk,

I have some interesting experience with AI implementation. Some applications that I am currently involved in are:

1) AI Document interpretation & insertion into CRM's and Quoting tools
2) AI contract referencing for the purpose of maintaining contract guidelines in quotes & services
3) AI messaging (Chat bots, SMS, Emails. Great for appointment setting and informing prospects)
4) AI Callers (For appointment setting. Best used in marketing funnels)
5) Then of course the classic OpenAI text prompts. Very useful for on the fly document review and creation. I personally use it often to reiterate existing email campaigns and messaging. For this space, it makes a great tool for discovery and due diligence questions to ask for an industry.
6) Lastly, I have seen the power of RPA automation. (Such as Microsoft Power Automate, Robomotion, etc.)

The main reason why a lot of these are hard to implement is 2 things. Integration & AI training. The other factor is that there are MANY people who will say they can build it. But few who have already built it. Since AI is so new, every business that is trying to adopt it is also taking on the standard risks of being an early adopter. But, with AI, the early bird will certainly get the worm of profitability & opportunity.

Reply by a searcher
from Colgate University in Charlottesville, VA, USA

Yes, I've built custom software that leverages LLMs for some processes. The current cutting edge of AI is inherently error prone. I'd say it's correct ~85% of the time, but you cannot rely on it 100%. As such, it is best used for processes where you have human review afterwards, and/or it doesn't need to be 100% correct. Some examples: "write a draft email/post based on the following context..." Then have a human review the output. That kind of thing can really save time in some workflows. Another: "Review the notes in these 10,000 documents, and call out some areas we should review more deeply". This sort of thing can allow you to focus in on areas of interest that you may have just never caught otherwise. Overall, it can definitely be useful, but always implement it with the assumption it cannot be correct every time.

+2 more replies.

Join the discussion