Remember the yoga analogy from my first newsletter? A core principle in yoga is not to force the posture. Flow happens when you release some parts and tighten others, letting the pose advance.
The same applies to change management. When we moved 200 long-tenured salespeople (some with 40 years in the company) to a new digital app, we had to “tighten” by requiring account creation: no anonymous logins. But we also had to “release” by keeping mandatory fields to the absolute minimum, so onboarding was frictionless. Push too hard, and the posture (the organization) collapses.
“Here are the 20 jobs AI will replace this year.” Anyone who sees their role on that list will stiffen with resistance – and rigidity is exactly what makes programs fail.
We’ve seen this before in countless digital transformations. Progress always requires knowing where to tighten and where to release.
Common Problems with Today’s AI
- Sycophancy: reinforces what the user wants to hear instead of offering critical thinking. Result: biased program decisions that simply confirm existing assumptions.
- Model drift: models degrade over time when exposed to real-world data. Result: outputs that once were accurate gradually become unreliable but are still used for planning.
- Unscalable PoCs: poor data, weak infrastructure, high costs, lack of governance. Result: frustrated stakeholders and IT budgets burned on projects with no tangible value.
The Upside of Well-Implemented AI
- Productivity gains: developers using AI can deliver faster. (Though Andrew McAfee‘s “AI Revolution, Meet the Credibility Revolution” rightly questions inflated ROI predictions.)
- Task automation: tools like Zapier connect 8,000+ apps and automate routine workflows. Add-ons in Atlassian, GSuite, and Microsoft365 eliminate repetitive manual tasks.
- Consistency: humans can be tired, limited, or inexperienced. A well-trained AI has broad knowledge, fewer biases (ideally corrected during training), and steady outputs.
The challenge is capturing the benefits without ignoring the risks.
Why Human-in-the-Loop Matters
Google, one of the leaders in AI, emphasizes this repeatedly in its trainings: “Don’t forget to include human judgment.”
Human-in-the-loop means blending machine intelligence with human oversight to train, validate, and refine AI models.
Example: a support agent correcting how AI filters customer requests, or a user fine-tuning generative AI outputs through prompts.
Wouldn’t you think Google would be the first to leave the human out if possible? Still, they insist on human-in-the-loop approach, and for right causes.
Think Washing Machine, Not Magic Wand
If you own a washing machine, would you still wash clothes by hand? Hopefully not: your time is worth more. But does a washing machine always deliver perfect results? Also no. Sometimes you need prewash, special detergent, or a delicate cycle depending on the material.
AI works the same way: it’s powerful, but not automatic. Human intervention is always required.
Three Things Only Humans Can Provide
- Governance: define where AI adds real value, enforce responsible use, correct biases, and run regular checks.
- Supervision: update models to reduce knowledge cuts, challenge hallucinations, and retrain when drift sets in.
- Feedback: the more feedback AI gets, the faster it improves. Think of it as an assistant: untrained, it creates more work; trained, it frees you to focus on what matters.
What AI Can Do for You Today
- Reduce routine, repetitive work.
- Lower human error.
- Cut time spent on research.
- Spark more creative ideas.
The real question isn’t whether AI will replace us—it’s how we integrate it to amplify our work and deliver more value. Will you try AI for your next status report? Sprint retrospective? Or even for birthday party ideas?
Subscribe and share to keep exploring how AI and human judgment work best together.
I am a Senior Program Manager with 12+ years of experience leading large-scale multicultural programs. Every week I will write here about program and project management. If this sparks your interest, hit subscribe and stay tuned.