• 📊 Data
• 💼 Talent
• 🤝 Trust
• 🤔 Other

------------------------------
Rachel Miles Sijacic
Sign up for the User Experience Program: ibm.biz/userexperience
------------------------------

#CloudPakforDataGroup

Trust

------------------------------
Rohit Goyal
------------------------------

Can you elaborate Rohit? Is it trust in the output being accurate enough to be used directly in decisioning making or used to automate a task.

Do you have any ideas of how to overcome this hurdle? 

------------------------------
Marc-James Abi-Jaoude
------------------------------

Original Message:
Sent: Thu May 06, 2021 02:21 PM
From: Rohit Goyal
Subject: 🎟 Biggest challenges with AI adoption

Trust

------------------------------
Rohit Goyal
------------------------------

Probably a mix of Trust and Data.
The AI model need to be re-trained to know when and why it is the delicate job. You cannot just leave it there and you need to be proactive if you do not want surprises.

------------------------------
Chiara Baldan
------------------------------

Thanks for your thoughtful response! That makes a lot of sense and is really helpful.

------------------------------
Rachel Miles Sijacic
Sign up for the User Experience Program: ibm.biz/userexperience
------------------------------

Original Message:
Sent: Thu May 06, 2021 02:26 PM
From: Chiara Baldan
Subject: 🎟 Biggest challenges with AI adoption

Probably a mix of Trust and Data.
The AI model need to be re-trained to know when and why it is the delicate job. You cannot just leave it there and you need to be proactive if you do not want surprises.

------------------------------
Chiara Baldan
------------------------------

one of the biggest challenges is getting access to the "right and accurate" data and providing the appropriate functionality to help with auto curation.  Seems to be a consistent theme with many of our clients  in fact data marketplaces are becoming a very hot ticket item

------------------------------
Beth Ackerman
------------------------------

Thanks for your thoughtful response! I can definitely see where that would be challenging.

------------------------------
Rachel Miles Sijacic
Sign up for the User Experience Program: ibm.biz/userexperience
------------------------------

Original Message:
Sent: Thu May 06, 2021 02:27 PM
From: Beth Ackerman
Subject: 🎟 Biggest challenges with AI adoption

one of the biggest challenges is getting access to the "right and accurate" data and providing the appropriate functionality to help with auto curation.  Seems to be a consistent theme with many of our clients  in fact data marketplaces are becoming a very hot ticket item

------------------------------
Beth Ackerman
------------------------------

Trust - Client's don't trust algorithms that can't explain why it reached a conclusion.
Other - Client's fear it will take away their work and they will lose their jobs. Sensitive area during the pandemic.

------------------------------
Vincent Tran
------------------------------

Thanks, Vincent! The pandemic definitely exacerbates fear for sure.

------------------------------
Rachel Miles Sijacic
Sign up for the User Experience Program: ibm.biz/userexperience
------------------------------

Original Message:
Sent: Thu May 06, 2021 02:32 PM
From: Vincent Tran
Subject: 🎟 Biggest challenges with AI adoption

Trust - Client's don't trust algorithms that can't explain why it reached a conclusion.
Other - Client's fear it will take away their work and they will lose their jobs. Sensitive area during the pandemic.

------------------------------
Vincent Tran
------------------------------

Ha, ha, I'd say the biggest challenge is to have trust in the talent of the analyst based on the low quality of the other input data.

Well the biggest challenge seems to be to get unbiased data.

------------------------------
Roland Schock
Distinguished Engineer
ARS Computer und Consulting GmbH
Muenchen
------------------------------

Kudos to you for making a sentence out of all that! That makes a lot of sense.

------------------------------
Rachel Miles Sijacic
Sign up for the User Experience Program: ibm.biz/userexperience
------------------------------

Original Message:
Sent: Thu May 06, 2021 02:36 PM
From: Roland Schock
Subject: 🎟 Biggest challenges with AI adoption

Ha, ha, I'd say the biggest challenge is to have trust in the talent of the analyst based on the low quality of the other input data.

Well the biggest challenge seems to be to get unbiased data.

------------------------------
Roland Schock
Distinguished Engineer
ARS Computer und Consulting GmbH
Muenchen
------------------------------

Data + Trust. Specifically data lineage and understanding where the data is sourced from.

------------------------------
Shawn Lunny
------------------------------

Yes! Can definitely see that.

------------------------------
Rachel Miles Sijacic
Sign up for the User Experience Program: ibm.biz/userexperience
------------------------------

Original Message:
Sent: Thu May 06, 2021 02:37 PM
From: Shawn Lunny
Subject: 🎟 Biggest challenges with AI adoption

Data + Trust. Specifically data lineage and understanding where the data is sourced from.

------------------------------
Shawn Lunny
------------------------------

Customers need more sophisticated tools that can be easily navigated to explain the outcome of AI models. If models can be explained in a convincing way, customers tend to more go along with the predictions. Current tools do not enable lower levels drilling into reasons behind the outcomes, in particular for end users audience

------------------------------
Hisham Ghanem
Architect
IBM
Vienna VA
571-4216350
------------------------------

Yes! It's such a balance between not overwhelming the users with technical details, but also being transparent. Any ideas on how to meet those needs would be appreciated!

------------------------------
Rachel Miles Sijacic
Sign up for the User Experience Program: ibm.biz/userexperience
------------------------------

Original Message:
Sent: Thu May 06, 2021 02:50 PM
From: Hisham Ghanem
Subject: 🎟 Biggest challenges with AI adoption

Customers need more sophisticated tools that can be easily navigated to explain the outcome of AI models. If models can be explained in a convincing way, customers tend to more go along with the predictions. Current tools do not enable lower levels drilling into reasons behind the outcomes, in particular for end users audience

------------------------------
Hisham Ghanem
Architect
IBM
Vienna VA
571-4216350
------------------------------

We need to close the gaps between AI technical details that can be interpreted by a data scientist (off course with some effort), and the business analyst / data steward that reviews and approves the model behavior to the customer's business needs. Complex Technical Model --> Clear Business Insights. Enhancing the tools will help data scientist make this leap.

------------------------------
Hisham Ghanem
Architect
IBM
Vienna VA
571-4216350
------------------------------

Original Message:
Sent: Thu May 06, 2021 02:56 PM
From: Rachel Miles Sijacic
Subject: 🎟 Biggest challenges with AI adoption

Yes! It's such a balance between not overwhelming the users with technical details, but also being transparent. Any ideas on how to meet those needs would be appreciated!

------------------------------
Rachel Miles Sijacic
Sign up for the User Experience Program: ibm.biz/userexperience
------------------------------


Original Message:
Sent: Thu May 06, 2021 02:50 PM
From: Hisham Ghanem
Subject: 🎟 Biggest challenges with AI adoption

Customers need more sophisticated tools that can be easily navigated to explain the outcome of AI models. If models can be explained in a convincing way, customers tend to more go along with the predictions. Current tools do not enable lower levels drilling into reasons behind the outcomes, in particular for end users audience

------------------------------
Hisham Ghanem
Architect
IBM
Vienna VA
571-4216350

I'm curious to know what you think about the Watson Explainable AI tool! Do you think it helps address some of those issues?

------------------------------
Rachel Miles Sijacic
Sign up for the User Experience Program: ibm.biz/userexperience
------------------------------

Original Message:
Sent: Thu May 06, 2021 03:16 PM
From: Hisham Ghanem
Subject: 🎟 Biggest challenges with AI adoption

We need to close the gaps between AI technical details that can be interpreted by a data scientist (off course with some effort), and the business analyst / data steward that reviews and approves the model behavior to the customer's business needs. Complex Technical Model --> Clear Business Insights. Enhancing the tools will help data scientist make this leap.

------------------------------
Hisham Ghanem
Architect
IBM
Vienna VA
571-4216350
------------------------------


Original Message:
Sent: Thu May 06, 2021 02:56 PM
From: Rachel Miles Sijacic
Subject: 🎟 Biggest challenges with AI adoption

Yes! It's such a balance between not overwhelming the users with technical details, but also being transparent. Any ideas on how to meet those needs would be appreciated!

------------------------------
Rachel Miles Sijacic
Sign up for the User Experience Program: ibm.biz/userexperience


Original Message:
Sent: Thu May 06, 2021 02:50 PM
From: Hisham Ghanem
Subject: 🎟 Biggest challenges with AI adoption

Customers need more sophisticated tools that can be easily navigated to explain the outcome of AI models. If models can be explained in a convincing way, customers tend to more go along with the predictions. Current tools do not enable lower levels drilling into reasons behind the outcomes, in particular for end users audience

------------------------------
Hisham Ghanem
Architect
IBM
Vienna VA
571-4216350

For me it's the talent and the data; most "business" users on the customer side don't possess enough skills to do the data analysis on their own without our support; they often don't understand the methodology and cannot interpret the results of a ML model; they also often misinterpret the results or are focusing on edge cases; I think the most crucial part is having good training data and also good test data to correctly analyse how well a model performs; Most customers find it rather difficult to decide what counts as good training data and which data will only confuse the system or it teach it incorrect things; also the necessity to add negative examples to your training data is mainly overlooked by customers

------------------------------
Dorothee Reinhard
Data & AI Scientist
IBM Switzerland
Zürich
0041 79 565 74 56
------------------------------

Yes! Your analysis is only as good as your data.

------------------------------
Rachel Miles Sijacic
Sign up for the User Experience Program: ibm.biz/userexperience
------------------------------

Original Message:
Sent: Thu May 06, 2021 03:02 PM
From: Dorothee Reinhard
Subject: 🎟 Biggest challenges with AI adoption

For me it's the talent and the data; most "business" users on the customer side don't possess enough skills to do the data analysis on their own without our support; they often don't understand the methodology and cannot interpret the results of a ML model; they also often misinterpret the results or are focusing on edge cases; I think the most crucial part is having good training data and also good test data to correctly analyse how well a model performs; Most customers find it rather difficult to decide what counts as good training data and which data will only confuse the system or it teach it incorrect things; also the necessity to add negative examples to your training data is mainly overlooked by customers

------------------------------
Dorothee Reinhard
Data & AI Scientist
IBM Switzerland
Zürich
0041 79 565 74 56
------------------------------