Creating More Efficient Yet Risk Managed Business Processes using AI

AI is everywhere and everyone is talking about AI. By now we all know the limitations of AI and their consequences in real business world. In a world where your competitors, regulators and public vigilante are looking for one mistake from you. Can you live with the 95%+ accuracy of AI? What else can we do? Perhaps modify our business process to suit the information gathered from AI.

While there can be multiple ways of doing so, for this bog's purpose, I am focusing on having an AI assist the human-in-the_loop in most comprehensive way possible. And let the human take most critical decisions.

---

Business Process: Blocking unethical manufacturers from entering the online marketplace.

Problem: If unethical manufacturers or suppliers enter the marketplace of an Online Ecommerce platform. If and when they get caught it may end up becoming bad press, cause of litigation, financial loss in terms of penalties etc.

Although reverse is also true, where an automated or a wrong decision to ban a seller who is actually ethical may end up being seen as anti-business. This may also lead to litigation, and bad press.

Assistance from GenAI: 

1. Collect Summarized Data from different data sources in the company: While there will be existing data which can show violations by the supplier, it needs to be present at right time to enable decision making. GenAI can connect to all those data sources at once.

2. Insights from existing data and external data: GenAI can be trained to combine the internal data with external or public available data to come up with a combined recommendation.

3. Decision or Recommendation: GenAI (LLMs) can recommend a decision or decide based on the above information.

Improvements expected from use of AI:

• As the data collection and building intelligence out of the existing data would be done according to the expertise of few individuals who might have limitations as well or done in silos. Use of AI can break the silos and bring it all together.

• The time taken for analysis of disparate BI and infinite web searches to correlate external data can be reduced by the use of AI.

AI Enhanced Process:

Below can be proposed enhancements to this process.

• The texts that have a credible reference can be highlighted in different color distinguishable from the text which is purely generated by AI.

• The recommendation needs to be presented to the human in the loop who can review all the text or data that has been put together.

• There should be easy ways for the human to verify all the internal information presented by the AI. Some sort of query capability which lets verification of any claimed value on the AI response.

• Similarly for external sources the reference glossary should be present and training should make sure that the AI goes to only verified sources.

• With all this information presented at one place human is more eqquipped to take decision quickly and do additional research if only required.