Blogs

These are the top 5 things that cought me by surprised me 👇
1️⃣ Users trust results they can validate Transparency matters. When users are a part of the validation loop, adoption grows fast. “Black box” answers create hesitation. Being a part of the validation process and having the tools to validate data builds confidence.
2️⃣ Speed changes priorities When answers take seconds, conversations shift surprisingly. Instead of “Can you send me the report for supplier A's past due PO’s?”, people start asking “What should we do about supplier A's late PO’s?” The focus naturally moves from data retrieval and report building to real time decision making.
3️⃣ Suppliers became more proactive A supplier used the chatbot to check all open POs coming due in the next 30 days with a 12 months on time delivery rate below 70%, and spotted a risk before the buyer did. That moment showed the real power of shared visibility and collaboration between suppliers and buyers instead of the familiar escalation cycle.
4️⃣ Questions evolved At first, users asked what happened (“Which POs are late?”). Within weeks, they started asking why (“Which suppliers are trending toward delay?”) and what’s next (“What PO’s will be late next month?”).
5️⃣ A mash-up feedback loop Integrating user validated questions into the AI feedback loop enhanced model precision by allowing the LLM to refine approved queries and continuously expand the question library for future use.
The below got me thinking about the end goal of AI in operations: It’s not about replacing people, it’s about freeing them to focus on what matters most.
Our next step? 🔮 Predictive Text2SQL: where questions like “What will be late next month if nothing changes?” become part of everyday conversations.
We’re closer than ever to that.