How to make AI work for you
The promise of Data and AI has been talked about to the core. But cutting through the marketing talk and hype, most organizations around the world are still struggling to realize the full potential of AI. And the concern for us AI practitioners, is that this should not lead to another AI winter.
The reasons for this stalling are multiple. And in the series of posts, I plan to walk through the challenges and how you can successfully navigate them.
In this first post I will be covering Business Buy-In
Other posts in this series : How to ensure business impact of ML Models
From an applied ML perspective, the best models are the ones which are used in real world and which improves business metrics continuously. Thousands of models with high accuracy and recall are stuck for eternity in POC sandboxes and never see the light of the day. So its critical to partner with business effectively and ensure that the models do what they were built for.
Great. But how do we do this?
Business Perspective
Key is to understand business perspective on why they are not fully committed in partnering with data science.
Business thinks AI is hype and have a trust deficiency
As a data science practitioner its critical that trust is built with business. And how to do this? Start with a real high impact problem for business which can be solved using data science in a short duration. And agree on a metric to measure the success. Key here is that it has to be solved quick and you have a clear way to measure success. So it need not be the optimal solution and shortcuts are ok. But this proves team’s credibility, and that data science is not hype and can solve real problems within a short duration. This should set you up for continued and sustained success.
Insecurity that AI is going to take away their jobs
Lets be real. Any repeatable work where the person-in-the-middle does not add value is at risk from automation. But the truth is many of the business functions bring in strong domain knowledge, perspective and expertise which cannot be replaced.
Yes. The low end repeatable jobs can be automated. But other value adding jobs can be enriched. And more importantly AI can open up the potential for niche high end work which business might not even have imagined.
So its important as a data science leader you level with business and open up their minds for possibilities
Businesses find it hard to define clear actionable AI use cases
Business responsibility is to improve their org’s metrics. They are not responsible for coming up with AI use cases.
So as a data science leader you have to understand the business roadmap and identify opportunities. So you should work as a tag team with business where business roadmap drives AI use cases. Once trust and credibility is built, Data science should get it’s rightful place in the table at business strategy and roadmap preparation meetings.
Blame Game
Who gets the credit if data science works (Yes. Sometimes it does !!). And more importantly who gets the blame if it does not.
As we know that best models are the ones used in real world. So typically business and data science should co-lead the initiatives. It’s ok for data science to take a back seat and be the enabler of business success.
Data science team success is measured by the number of businesses waiting to partner with them
Conclusion
A model in production and used by business effectively is better than hundreds of models in a POC sandbox.
Partnering with business effectively is the first step to success
So finally, Data science and business now see eye-eye and have got each other’s back.
But now its time to walk the talk. How do you build and deploy models at scale to solve real world problems ?
We will see in further posts about challenges in operationalizing ML and how to address them
