WHAT MAKES A GOOD AI STORY?- 5 KEYS FOR STRONG AI NARRATIVES

---

Any good story have certain characteristics, e.g.

• The storyline gets built with a specific targeted audience/ readership, in the minds of the authors.
• The story includes characters that are relatable or at least imaginable, if drawn from the world of Magic for instance (ref. the Harry Potter stories).
• The plotlines/ storylines evolve around the characters. They are personalized/ personified (e.g. Dobby the elf, a magical creature but personified so that s/he can interact and behave with human wizards like Harry)
• The deeper and more impactful the plots are, the more the readers get drawn into it. 
• The endgame is crucial- in terms of significant, visible and perceptible impact. 

Good AI storytelling has to fundamentally embrace these practices of good storytelling. Therefore, any well-developed and well-articulated AI story should check in the all of the following boxes:

1. Audience: Does your story reflect real business challenges for your targeted audience, instead of just talking merely 'good-to-have' generic tech-speak?

2. Reality check: Are there enough relatable persona's/ characters, reflecting the real-world stakeholders covered in the story? (e.g. stakeholders can be customers, suppliers, employees, shareholders, functional teams like HR/ finance)

3. Validity: Do your AI stories build upon the challenges faced by these stakeholders, in personally relatable, evidence-provable, experientially valid or common-sensical, realistic terms?  

E.g. for an AI story of detecting potentially fraudulent auto-insurance claims, the claims processing team are the target audience and the key stakeholders. Now, suppose they see that the problem context in the AI story is inflated with numbers like "more than 70% claims are actually fraudulent, and 92%+ of them are detected/ classified correctly by the AI solution (reflecting high accuracy of the model)". They will find the 70%+ number hard to believe and relate to, based on their decades of process experience on the ground. Now, even if the model accuracy was indeed 92%+ which is very good, it will sound incredulous and rather fictitious, in front of the business teams. Result: The story falls flat :(

4. Deepening/ cascading plot: Do the AI stories accentuate the plotlines in terms of impact, so as to draw the readers into it? 

For example, for a common AI usecase for AIOps/ autonomous IT ops, the start-point may involve just one example client-team leader or CXO role persona. Then, as the plot progresses, the impact metrics can slowly become more integrative, expansive and inclusive, e.g. in terms of the impact of a server crash on to an e-comm/ stock trading portal/ app losing millions of $s of transactions/ business in a few seconds. 

5. Impact articulation & scalability: Do the AI stories cover and articulate the impact in a multi-channel/ multi-dimensional, measurable, scalable, extrapolatable manner? 

E.g. for the above example of an AIOps solution storyline, in terms of expanded scope/ context, the impact metrics can cover integrated tech infrastructure- related incidents  (e.g. across the entire hybrid cloud compute-storage architecture in an enterprise), much beyond just one server stack. It can also cover all the critical business apps and cloud apps etc., using the server crash as a high-impact/ value example instance and then multiplying it with relevant factors of scale.