As Automation Strategies Pivot, Do We Need Better Guidelines?
I have been fascinated by our single-minded focus on AI, in all its forms, as the latest silver bullet.
To be clear, I am a supporter, user and work for a provider of AI capabilities, but it is the HI I struggle with 😉
HI – Human Intelligence
But we all know there is a place for everything, and everything has a place.
In my opinion, there is not enough discussion and debate about WHY and WHERE to apply different technologies.
So, I was encouraged to read the experiences of Jim Swanson of J&J reported by Isabelle Bousquette of the Wall Street Journal titled “Johnson & Johnson Pivots Its AI Strategy” where the company is making a shift to focus on only the highest-value GenAI use cases and shutting down pilots that were redundant or underdelivering.
This is great news.
“We’re prioritizing, we’re scaling, we’re looking at the things that make the most sense”.
Jim observed that “in some cases, too much experimentation can be at odds with actually finding business value”.
It is about time we developed a guide, roadmap, checklist, quadrant or paradigm shift gizmo thingy 😉 that simplifies and accelerates the discussion and decision on what technology should be applied to which problem.
Obviously, we need to start with the “end to end” business process, which can be made up of hundreds of tasks and task groupings. The mapping, if it were to exist, needs to focus at this task level. Because it is the tasks that exhibit the characteristics, not the end-to-end processes.
We know that when we understand an “end to end” process, we need to work out how to streamline and optimize.
Classically (but still not universally applied), we address these four choices with regard to tasks, sometimes in sequence, sometimes just one;
- Eliminate
- Simplify
- Automate
- Delegate (outsource)
But, it is more nuanced than this.
In the age of AI vs HI, I suggest we need a guide to answer the following questions;
- Where do we want Human Intelligence in our processes, tasks and interactions and why?
- Where do we want Human Empathy in our processes, tasks and interactions and why?
- Where do we want Deterministic (binary) Automation in our processes, tasks and interactions and why?
- Where do we want Probabilistic AI (Analytical AI) in our processes, tasks and interactions and why?
- Where do we want Generative AI (Analytical AI) in our processes, tasks and interactions and why?
Thats ruined my plan for a four box quadrant!
If we can have this debate it will save us a lot of time and wasted money.
So, I started to develop a “strawman” to stimulate a discussion, and then asked my friendly LLM for a first stab.
As Automation Strategies Pivot, Do We Need Better Guidelines?
Purely to stimulate some human creativity, not assuming it would be an end result in itself!
So, for starters, here is what it suggested;
🔍 1) Human Intelligence
Best for:
- Abstract reasoning & judgment
- Strategic thinking under ambiguity
- Creativity in undefined problem spaces
- Integrating context and experience
Optimized Business Problems:
- Long-term business strategy
- Cross-functional decision-making
- Innovation/new product development
- Crisis management
❤️ 2) Human Empathy
Best for:
- Emotional and ethical decision-making
- Relationship-building and trust
- Understanding human nuance
- Handling sensitive or moral dilemmas
Optimized Business Problems:
- Patient care in healthcare
- HR leadership and mentoring
- Customer conflict resolution
- Diversity, equity, and inclusion efforts
🤖 3) Probabilistic AI (e.g., Machine Learning, Deep Learning)
Best for:
- Making predictions from uncertain data
- Learning from patterns and behaviors
- Dealing with noisy or incomplete data
- Personalized decision-making
Optimized Business Problems:
- Risk scoring (insurance, loans)
- Predictive maintenance in manufacturing
- Customer churn prediction
- Targeted marketing and ad optimization
✅ 4) Deterministic Logic (Binary / Rule-Based Systems)
Best for:
- Automating repetitive and rule-based tasks
- Performing exact calculations
- Executing predefined workflows
- Ensuring compliance and auditability
Optimized Business Problems:
- Tax or payroll processing
- Compliance reporting
- Inventory reordering systems
- Access and permission control
🎨 5) Generative AI (e.g., GPT, DALL·E, image/audio/video synthesis)
Best for:
- Content creation at scale
- Ideation and creative augmentation
- Natural language understanding and generation
- Automating knowledge work with flexibility
Optimized Business Problems:
- Marketing copy, blog generation, email drafting
- Code generation and documentation
- Product design mockups or prototypes
- Chatbots and conversational agents
Then there is the question of where any one of these options should be used in isolation and where it is best used in augmentation to one of the others?
It is just an incomplete and not well refined first cut, now let’s apply some HI.
I suspect we have some work to do.
Discuss . . .
I suspect this is going to be a big topic over the coming 18 months.
Where are the most effective use-cases for HI vs AI? It is not just a binary choice.
But we do need to define the classes of activity/task and likely best fit for each category of process support.
Thanks for making me think and for buffing up the old HI . . . .
As Automation Strategies Pivot, Do We Need Better Guidelines?
Thanks for reading . . .