Reps start from a drafted order.
An AI agent drafts each order line with a reason and the numbers behind it. The rep edits and submits, so the decision stays with the person who knows the store.
FieldIQ Draft Order Agent →I understand how CPG sells. I build AI that helps commercial teams make better decisions.
I'm Namratha, a Senior Product Manager with 7.5 years across CPG commercial: pricing and revenue growth management, trade promotion, sales execution and B2B commerce, in 22 countries. I build AI that improves commercial decisions, with a person approving every consequential call.
- 7.5 yrs CPG commercial
- 22 countries
- 2024 Silver Stevie
- $250M+ client-reported revenue uplift
- IIM Indore MBA
The agent drafts. The rep approves. The AI never places the order.
Commercial first
Sales execution, shelf compliance, B2B commerce and pricing, across 22 countries.
AI with a brake
A Stevie-winning pricing agent that recommends and never acts alone.
Builder, by choice
A career break to ask "what did we never build, and why?", then FieldIQ, built solo.
Selected work
Writing
All writing →AI autonomy
CPG isn't behind in AI adoption. It's slower to trust it.
Autonomy should match the cost of an error, the reversibility of the decision, and the blast radius when it fails.
(opens in a new tab)EvalsI ran 19 tests. The one that failed was my fault.
I thought I was testing the AI. I was testing my own thinking, and my thinking had the bug.
(opens in a new tab)Sales executionThe rep becomes a calculator with a car.
Why FieldIQ's Draft Order Agent gives selling time back, and the controls I built before trusting it.
(opens in a new tab)RGMToo many RGM insights stop at the dashboard.
What happened in the store should shape the next RGM decision.
(opens in a new tab)AI strategyThe most popular AI use cases in CPG may be the lowest-returning ones.
Judge pilots on decisions changed, not hours saved.
(opens in a new tab)B2B commerceWhen the order places itself, what earns a human's attention?
Lessons from the distributors who never logged in.
(opens in a new tab)My view: how AI can help CPG sell better
AI creates value when it improves the decision, not when it adds intelligence to a task that didn't need to be faster. Here's how I'd connect it across commercial execution, from plan to shelf.
Read the walkthrough
- Plan: AI drafts the demand baseline by account and SKU, and shows the drivers behind each number. The planner adjusts.
- Trade plan: AI scores each planned event on expected return and suggests moving budget from low-return events. The key account manager decides.
- Price and promotion: elasticity and promo history become a specific recommendation with its reasoning, and a person approves it.
- Agents watch the promotion: monitoring agents spot sell-through below plan and a competitor offer, a promo agent drafts an update, and the revenue manager signs off.
- Sales team briefed: the update lands in the rep's visit list as an Act Today task, and the draft order already includes the promotion requirement.
- Rep sells with AI: a shelf photo shows the gap, the gap becomes an order line, the rep edits and submits, and the next visit verifies the gap closed. Store results feed the next plan.
The shelf reports on itself.
A shelf photo read by AI shows every gap against the contract, and each gap becomes a task for the visit. On the platform I owned, about 30% more outlets hit their target product mix.
Shelf compliance, 22 markets →Pricing tools recommend a decision.
Revenue management teams told me they want help deciding what to do next. The product I led recommends price, promotion and mix changes, and a person approves each one. It delivered $250M+ in client-reported revenue uplift for a global food and beverage company.
Agentic RGM platform, 2024 Silver Stevie Award →AI has to pass a test before it ships.
Contract terms, minimum order quantities and promotion windows break most AI demos. Every FieldIQ agent release has to clear a test suite that allows zero critical failures.
The release gate →Simple rules still win some calls.
A rules-based visit score beat a vendor's ML route planner and took reps from 4 to 7 calls a day. Part of the job is knowing when AI is the wrong tool.
The ML pilot I shut down →Try it: how deep should a promotion go?
RGM teams make this call every week. Deeper discounts sell more units, but every unit you'd have sold anyway is sold at the lower price too. Move the slider.
Temporary price reduction, as % off list price.
A display costs a fixed fee but lifts every unit of discount harder, so the best depth moves.
Past 7%, each extra point of discount mostly subsidizes units you would have sold anyway.
+51%Volume lift
66%of units you'd have sold anyway
+2,886Incremental profit vs. no promo (index)
+0.19Profit per 1 of trade spend
Price-to-pocket, per unit (list = 100)
Illustrative model with invented coefficients (baseline 1,000 units, unit cost 50, off-invoice 6, diminishing lift, a 20% post-promotion dip). It shows the shape of the trade-off. No client data or product output is used.
Hiring for a Senior PM or AI PM on a CPG commercial team?
I'm happy to walk you through FieldIQ live, or go deeper on any case study.