Thinking out loud on agentic AI and enterprise operations.

Why Your SKU Portfolio Is Quietly Killing Margin

Manual, spreadsheet-driven SKU rationalisation fails because it treats a data problem as a one-off cutting exercise. A governed, AI-assisted scoring and what-if simulation approach turns it into an ongoing, cross-functional decision process instead.

Read on LinkedIn →

The Next Cost Takeout Opportunity in BFSI: AI for Back-Office Case Operations

BFSI's AI focus has skewed toward customer-facing chatbots, leaving the bigger cost opportunity — back-office case operations — largely untouched. AI should prepare the case; humans should still apply judgement.

Read on LinkedIn →

Agentic AI Will Fail Without Human-in-the-Loop Governance

The real question isn't whether to deploy autonomous AI agents, but how to govern them. A staged model — detect, explain, recommend, approve, execute, audit — keeps humans in control of consequential decisions while still letting AI accelerate routine ones.

Read on LinkedIn →

Document + Image + Video Intelligence: The Missing Layer in Customer Operations

Text-only chatbots break down the moment a case depends on a photo, a scanned document, or a video, not just a ticket description. Real operational efficiency requires extracting and cross-referencing evidence across every format before a case is routed.

Read on LinkedIn →

AI in Dispute and Chargeback Operations: The Unsexy Use Case That Can Save Millions

Dispute and chargeback handling is manual, back-office, and gets little executive attention — which is exactly why it's a large, underexploited AI opportunity. An intelligence layer that reads cases and prepares structured decision cards cuts handling time and revenue leakage without removing human accountability.

Read on LinkedIn →

David vs. CPG Goliaths: How AI Empowers Mid-Market Food Manufacturers to Compete

AI is levelling the field between mid-market food manufacturers and enterprise-scale CPG competitors — through better demand forecasting, computer-vision quality control, faster product development, and the agility to deploy faster than legacy-bound giants.

Read on LinkedIn →

Why Customer Support Automation Is No Longer About Chatbots

The real value in support automation isn't a better front-end chatbot — it's automating the back-office workflow behind it: document reading, evidence validation, routing and approvals. That's what actually moves resolution time, cost and compliance.

Read on LinkedIn →

LLMs Will Not Replace Data Science Models. They Will Make Them Smarter.

LLMs aren't a replacement for classical ML models — they're a way to accelerate feature engineering, turning unstructured enterprise data into structured signal, inside a human-supervised feedback loop that keeps governance intact.

Read on LinkedIn →

Your Working Capital Problem Isn't a Finance Problem. It's a Data Problem.

Manufacturing CFOs typically manage working capital through siloed teams and disconnected tools rather than as one integrated data problem. Applying AI across receivables, inventory, procurement and dispatch — in that sequence — builds real visibility into the cash conversion cycle.

Read on LinkedIn →

From AI Pilots to AI P&L: How CXOs Should Measure AI Transformation

Most AI initiatives don't fail on technology — they fail because no one connects the pilot to a P&L line. AI transformation should be measured in working capital impact, turnaround time, exception rates and fraud prevention, not pilot counts.

Read on LinkedIn →

Predictive Maintenance Is Old. Agentic Maintenance Is the New Opportunity

Predictive maintenance stops at flagging risk — it doesn't act on it. Agentic maintenance converts a machine alert into a structured, actionable maintenance card that coordinates inventory, production and approvals, turning prediction into governed execution.

Read on LinkedIn →

From Dashboards to Decisions: How Agentic AI Can Transform Manufacturing Operations

Manufacturers already have plenty of data — what they lack is speed of decision, since choices stay fragmented across systems and teams. Agentic AI works best as an execution layer on top of existing ERP, starting with one bounded workflow and a human-in-the-loop approval gate.

Read on LinkedIn →

Agentic Orchestration for Data Science Operations

AI agents should enhance, not replace, human oversight of the ML model lifecycle. Automating data validation, model evaluation and evidence compilation — while keeping human approval gates on critical decisions — improves both model quality and governance.

Read on LinkedIn →

From SAP as System of Record to SAP as System of Action: The Agentic AI Opportunity for Manufacturers

Manufacturers get more from embedding agentic AI directly inside SAP than from building a parallel AI platform. Four use cases — finance, procurement, planning and maintenance — show how, grounded in Clean Core principles, SAP can move from passive reporting to active decision-making.

Read on LinkedIn →

When the Kiln Already Knows It's About to Fail

Plants already generate the data needed for predictive maintenance and quality improvement — in historians, MES and SCADA systems — without needing new sensors. Physics-aware anomaly detection, remaining-useful-life models, computer vision and time-series forecasting turn that existing data into decisions inside the workflows planners already use.

Read on LinkedIn →

Why SAP Joule Is Worth Evaluating: From ERP Copilot to Agentic Execution Layer

SAP Joule is a meaningful step beyond a generic AI chatbot — it can understand business context and respect existing governance to detect exceptions, recommend actions and execute approved tasks across finance, procurement, planning and maintenance. Evaluate it on measurable operational improvement and Clean Core fit, not on AI novelty.

Read on LinkedIn →