How can ai help innovation?

I recently attended the Executive Forum on Artificial Intelligence sponsored by the WSJ in New York.  Here are the scribbles from my iPad that I took during the conference.  If you are interested in learning more how AI can impact your Innovation Program please contact me.

My notes fall into the following main categories:

AI Overview

• We are early in the cycle and AI is the new buzz word
  
• In reality, somethings that are not AI are called AI because executive leadership want AI
  
• But given all the noise execs need to understand what is hype and what is reality.
• Behind the hype there are real results that business are achieving
• The key is business and execs need to start focusing on AI
  
• There were lots of questions that asked is this just like e-commerce of 10 years ago?
  
• Yes and no.  Yes: a major shift in technology that you need to be prepared for; No: more complex and disruptive than e-commerce
• Why this is happening now?  The explosion in data and compute power have been driving this since around 2010

Data:

• To be successful with AI you need a deep understanding of your data
  
• Realize that all data will require hard taxing work to clean it
  
• There is a need for data scientist who can relate to your data and create a model
• But there is also a need for a strong business person who understands the context of the data
• This is a dilemma.  First there are not many data scientist to hire and second most don't understand your business
• This is a focus area that you need to build out your skills.  Identify key potential candidates in your business and start to train them
• You want this to be inside your business because this is your IP, your expertise, your value and you don't want your organization to be forward selling for any consultants you hire.
  
• Your data needs a strong and fast feedback loop so that learning and improvements can be incorporated quickly to drive results.
  

Work

• To be successful you will need to re imagine how work is done in conjunction with  AI.  Work needs to augment and provide business value that is relevant to the user.  AI can often obscure this if you think in old how we have done it in the past terms
  
• Example:  a voice response system that authenticates based on caller cid and the users voice and immediately asks which policy "a or b" do you want without telling the user that you just authenticated them and looked up their policies.  This approach can causes anxiety and a feeling that your solution is big brother.  A better re-imagined approach would be to inform the user that based on their voice and caller id you have authenticated them and that you found the following polices for them and  ask them which one are they calling about.  Subtle but critical to success.
  
• Doing good old process modeling work for as-is and to-be work is very important and often undervalued.
  
• Need to focus on what it is you are actually trying to accomplish from a business perspective but looking at it from a users perspective.  Design thinking approaches, Sponsor User, and user focus groups are critical
  
• Though automation is not really AI it plays a big part in the re-imagining of the work
• Need to think of work from user inward not other way around
  
• But in doing this you need to recognize there are some things that machines are better at than humans and somethings that humans are better at than computers.  You need to make sure you understand this in building solutions.  Example: healthcare will still require empathetic individuals that may be assisted by robotic analysis, triaging or diagnosticians
  
• Focus your efforts on high value business areas but take them on in small bites.  Iterate rapidly.  To find best use cases follow the money. Where is money spent in business to drive value.
  
• Work will displace people. It will require a great deal of human redeployment.  Organizations need to plan for this

Audit

• Solutions need to be audit-able
• A key area to consider is is product liability. What happens when your model fails (it will fail).  Will this lead to your organization being litigated?
  
• Need to be able to address privacy and security concerns
  
• All models will have bias - how will this impact you?
  
• All models will fail at some point- how will this impact you?
  
• You need to build into model the ability to record along the way what it has done so that you can go back and audit
  
• Compliance is another front of the audit.  How will your solution comply with ethical, organizational and regulatory components.
  
• A key component to driving audibility and avoiding challenges is to use a very diverse teams to build and validate any AI solution.
  

Technology

• There were some nice technology booths from the following vendors.  Not promoting, but found their solution to be very interesting
  
• Model building. Software from H2O.ai
  
• Cyber security with DARKTRACE
  
• AI for IVR from Interactions
  

Governance

• The organization structure needs to be nimble and effective for AI to work.  This can be different for different orgs. 
  
• Some use Centers of Excellence or Garages.  A centralize approach.  But the discussion focused on it doesn't matter how but that the key ingredients need to be funding that allows the ability to try, fail and learn.  You will fail and this is good.  But need sponsorship to allow quick recovery and improvement.
  
• There is a need for rapid Proof of Concepts )POC) mentality with ability to quickly move to scale once proven. 
  
• The ITR Path Innovation Framework is an ideal fit