Execution at the Edge of Intelligence

Each quarter brings new evidence of how quickly the bar is rising in CPG execution. What felt like advanced capability eighteen months ago is now a baseline expectation in the rooms where commercial leaders make investment decisions.

Three themes have defined recent work with customers:

Taken together, what we are describing is agentic execution across the full route to market, not just within field sales, but across logistics, distribution, and the data flows that connect them. That is the direction. That is where we are building.

Building from where we left off in the last issue, recent product work has concentrated on three areas where the gap between what CPG teams need and what legacy systems provide is largest: temperature-controlled distribution, agentic field execution, and distributor-side signal detection.

Cold Chain

We completed the integration of real-time IoT temperature monitoring into the Lighthouse logistics layer. Continuous sensor data now feeds a machine learning model running inference throughout every shipment journey. When a temperature breach is detected or forecast, the system classifies the deviation, estimates product loss risk, and pushes an alert with a recommended action, without waiting for manual escalation.

Agentic SFA

Beyond surfacing the right signal to a rep, we've started letting the system act on one. Our first agentic capability targets the most common leak in field sales, the visit that ends without an order because the owner is absent or unavailable. Rather than record a no-sale and move on, the rep can hand the outlet to an autonomous agent that engages the retailer directly and completes the order on the rep's behalf.

We've built it as a governed extension of the rep, not a replacement:

• Sales rep initiated: the agent acts only on a visit a person chose to hand off, it never reaches out on its own.
• Human-in-the-loop: an unusually large or risky order is held for a person to approve, reduce, or reject before it ships.
• Consent-first and bounded: nothing happens without a confirmed contact and consent, and the field app stays the rep's cockpit to dispatch, track, and resolve every handoff.

This is an early look. The field-app side is live and device verified, and the end-to-end agent is in controlled pilot. It is our first step toward agentic execution at the last mile, systems that detect, decide, and act, with a human owning the moments that carry risk.

Signals for DMS

We extended our predictive risk-scoring layer into the DMS. Distributor-side signals now include:

• Secondary sales velocity anomalies
• Stock depletion curves that diverge from model predictions
• Fill rate drops the system can identify as structurally different from seasonal variation

Real-time inference runs on distributor data streams. The output is an agentic nudge surfaced at the right level of the commercial hierarchy, before the anomaly compounds.

Three capabilities are moving into the field this quarter, in full production or active pilot, each targeting an execution gap that has persisted because underlying data was unavailable in real time, too unstructured for conventional systems, or siloed from the rest of the commercial stack.

Across markets, categories, and organisation sizes, a pattern keeps showing up in our operational audits. The bottleneck is rarely the technology, and it is rarely the people. It is the structural layer in between: the workflows that were never connected, the reporting that still depends on someone manually pulling a number from one system into another, the visibility gap everyone acknowledges but nobody has a mandate to fix.

Looking closer at where that gap actually comes from, we keep running into the same three failure modes.

The anatomy of the disconnect

• The translation chasm: strategic intent, "we need better visibility into secondary sales," gets converted into rigid functional requirements, and what comes out the other end is a platform that records data instead of driving decisions.
• Dashboards vs. control towers: teams are flooded with passive reporting that shows what already happened, with no equivalent system built for active, in-the-moment intervention.
• The rigidity trap: generic platforms built for standard workflows cannot accommodate the multi-tier reality of complex distribution networks, so teams end up building manual workarounds just to get through the day.

How we close it

• Design for the edge, not the centre: build around the daily reality of the person doing the work on the ground, not just the executive summary.
• Prioritise actionable signals over data volume: replace sprawling analytics interfaces with systems that surface exceptions and trigger the next action.
• Synchronise architecture with operational flow: make ERP, DMS, and logistics integrations mirror the actual cadence of the business, not a manual, swivel-chair report.

When that structural layer is resolved this way, field teams move faster than planned, not because they lacked capability or commitment, but because the friction that was slowing them down is finally gone. That is exactly what we saw in our most recent rollout.

212 users active on day one is not a training outcome. It means the people closest to the work recognised immediately that the platform solved something real. That signal matters more to us than any benchmark we track internally. The case studies below go further into what these rollouts looked like on the ground.

The way CPG leadership teams talk about execution technology has shifted noticeably over the past few quarters. A year ago, the conversation was about whether to invest. Now it is about what the next layer of investment is actually supposed to do. That is a meaningful change, and it reflects a broader shift in what the market is rewarding.

Quick commerce has moved from an emerging channel to a structural force. India's q-commerce market crossed Rs 40,000 crore in 2026, an eightfold increase in five years, and now accounts for 60 to 75 percent of all online FMCG sales. Traditional distribution still moves 90 percent of volume, but the speed and visibility expectations being set by 10-minute delivery platforms are being felt all the way back to the distributor depot. Fill rates, replenishment cycles, and stock assurance requirements are all being compressed. That pressure does not stay in the quick commerce lane.