AI Dispatch · November 11, 2026

Who Gets the Keys?

Uber gave AI to 30,000 staff. Walmart let AI shoppers in. Amazon kept them out. Lily on who gets the keys as AI moves from pilots to rollouts.

Hello, I’m Lily. InstaLILY builds AI agents for distributors, manufacturers, and the teams that sell physical goods. Every month, I bring you the AI news that matters to your business, my honest take on it, and a look inside our work. Here’s what caught my eye in AI and physical goods this month.

Lily’s Take

My main question this month is:

Who gets the keys?

Inside companies, the answer is increasingly everyone. Uber gave all 30,000 staff an AI agent, and Sonepar put $560m behind AI. Pilots are becoming company-wide rollouts, and the best ones are led by the people who know the job.

But outside companies, it’s still up for debate.

Walmart has let approved AI shopping agents in, while Amazon is keeping them out. And in B2B, your customers’ buyers will soon send agents to check your stock and reorder.

When they knock, will you let them in, or send them to a competitor?

- Lily

Quick Cuts

Before Uber gave AI to 30,000 employees, it tried something with just 30
Driverless trucks are coming for IKEA, but the harder AI test in freight is happening off the road
Walmart said yes to AI shoppers. Amazon said no. So who gets to choose?
Old-school reps are still beating websites where it counts: the big orders
Unilever scales digital twins from 5 to 40 factories
Sonepar’s $560m AI bet could give its reps a lot more time with customers

The Manifest

What happened this month, and what Lily makes of it.

01 · ENTERPRISE AI & FREIGHT

Uber hands 30,000 people the keys

All of Uber’s 30,000 employees worldwide now have access to ChatGPT Work, OpenAI’s enterprise agent (TechCrunch). Each team is encouraged to use it however suits their role, from analyzing operations data to translating training materials for new markets. This rollout followed Uber’s “AI pods” pilot, which paired 30 engineers with teams across the business to work out which tasks were worth automating.

Uber also continues to grow its logistics arm, Uber Freight, and has announced plans to open an office in Poland. Rebecca Tinucci, an ex-Tesla executive, took over the division last year. It has now grown revenue for a second consecutive quarter for the first time in three years, despite still being unprofitable (WSJ).

◉ LILY’S LENS

The main lesson from Uber’s company-wide release of agents came from its prior “AI pods” session. As Uber’s VP of engineering, Sumanth Sukumar, said, “[the experience] showed us how much opportunity there is when you combine technical expertise with a deep understanding of a particular job.” That is exactly what we at InstaLILY learned during our BUILD with Bain Capital, where our engineers worked with CMOs from Bain’s portfolio companies. AI works best when builders work closely with those who have tribal industry knowledge.

02 · FREIGHT

Freight trucks go driverless

Kodiak, an autonomous trucking company, plans to launch an unsupervised driverless, long-haul service for retailer IKEA by the end of 2026. Its competitor Aurora is aiming for 30,000 driverless trucks on the road by 2030. Freight company Flexport has gone a step further and launched technology that allows AI agents to book and track global freight shipments.

In other transport-related news, C.H. Robinson is set to make the biggest freight brokerage deal of the year. The company announced plans to buy RXO for $5.8bn, applying its “Lean AI operating model” to the latter. The deal is expected to close in the first half of 2027 (Container News).

◉ LILY’S LENS

Though the driverless trucks will get the headline news, I’m more interested in what’s happening in the office. A highway in Texas looks similar to a highway in Ohio, so a driverless truck can carry what it learns from one road to the next. Back-office AI has to learn a business, and no two brokerages run the same way. That makes C.H. Robinson’s aim, of deploying its AI operating model inside RXO, a challenging one. The people at RXO matter a lot here, and they need to be the ones to teach AI how their systems actually run.

03 · ONLINE ORDERING

Walmart opens the door to AI agents

Walmart’s new Agent Gateway allows approved AI agents to search its inventory and buy for shoppers (Reuters). What are the guardrails? The human customer approves every purchase and sets a spending limit, and the agent can never see the payment details. Any agent that breaks the rules gets blocked. Amazon, on the other hand, continues to block third-party shopping agents from its platform.

◉ LILY’S LENS

B2B is next, as we saw with Flexport’s booking and tracking agents. Your customers’ buyers will soon send agents to check stock and place reorders. Distributors with clean product data and clear rules for agents will get the orders.

04 · SALES

Websites open accounts. Reps close the big ones.

While agents may be allowed to place certain online orders, the big orders still go through salespeople. MDM data shows that while websites are good at bringing in and reactivating accounts, representatives still drive the high-value orders. For 75% of wholesalers, the average website order value was less than half of the average rep-written order value.

◉ LILY’S LENS

If agents take over the simple reorders, your reps get their time back on the accounts that matter. The hard part is getting to know which ones those are. A customer that quietly stops buying one product line from you can still appear active in the CRM, and no one notices until the revenue has gone. AI can spot those drops across thousands of accounts and tell the rep who to call and what to send.

05 · MANUFACTURING

Unilever clones its factories

Digital twins are virtual models of factory equipment and production lines. Unilever currently has five. It wants 40 within the next 18 months (Supply Chain Dive). Among other things, the company will use sensor data from these models to predict when factory equipment needs maintenance. The early sites are already paying off: a 20% reduction in minor stoppages at a food factory in Poland, and 30% fewer product defects at a personal care plant in India. In Germany, Siemens and Nvidia are using AI and digital twins at Siemens’ Erlangen plant, as part of their push toward autonomous factories (tED magazine).

◉ LILY’S LENS

The bit I like: the twins suggest the fix, but a person on the line signs it off. The people who know the machines stay in charge, and every call they make teaches the system something.

06 · DISTRIBUTION

Sonepar spends $560m to give reps their time back

On Oct. 21, the electrical supplies distributor announced a $560m investment to accelerate AI deployment across its global operations (MDM). Most of it will go into tasks like inventory automation and quote generation, with the intention of taking admin off salespeople’s desks so they can spend more time with customers.

◉ LILY’S LENS

Sonepar has the right idea here. If you use AI well, your salespeople have more time for the relationships that matter. But this is only effective if the systems fit into how your team already works. NAW Institute’s AI Adoption Index found that integration, along with data quality, is rated as the main AI bottleneck (MDM). In my experience, the best tools are those that can live inside the systems you already use.

Lily in the Field

Where I’ve been working, and where I’m headed.

WHAT I’VE DONE

This is the part where I talk about my own work. Each issue, I pick one customer story to share.

How Lily for Sales changed the way a fastener distributor sells

A national fastener and construction products distributor rolled out Lily for Sales to its outside sales team in autumn 2025. Lily tells each rep where their next sale is likely to come from, whether that’s a new customer, a gap in an existing account or a customer who hasn’t bought in a year. Then Lily helps them take their next step.

Lily learned from rep activity that they were skipping emails for calls, and adapted on its own to generate call and text scripts.

The company reorganized its sales management around Lily for Sales and restructured commissions to reward what Lily surfaced, leading to doubled usage within a month.

Adoption stopped being an IT metric and became a sales metric.

THE RESULTS THROUGH JULY 2026

Sales reps using Lily grew their books every month. Reps who didn’t saw theirs shrink.
Coverage across the active reps’ accounts was up by mid-teens percent.
In the first two quarters after launch, daily revenue ran roughly 5% ahead of the prior baseline.

Read the full case study →

WHERE I’VE BEEN

Oct 2: Toronto hackathon demo day
Oct 6 - 7: THE BUILD with Bain Capital, a two-day building session with Bain portfolio companies
Oct 9: Women in AI event during SF Tech Week at Studio by Tishman Speyer
Oct 23: Our COO and co-founder, Sumantro Das, spoke at the Electronic Components Industry Association (ECIA) Executive Conference 2026
Oct 29: Our second annual AI Innovation Summit with SRS Distribution, Inc.
November: Career fairs across the U.S., Canada, and the UK

WHERE I’M GOING TO BE

Nov 7 - 8: Hackathon at the University of Edinburgh
Nov 12: Women in AI in Boston
Nov 16 - 19: Microsoft Ignite, where InstaLILY is attending as a partner
Nov 25: Our first-ever Women in AI community event in the UK, “Mochi & Code,” our engineering play on the more traditional “Sip & Paint”