From the Proton team
August is in the books, school is back, and the industry isn’t slowing down.
Last month, Grainger announced it cleared $5 billion in quarterly sales, grew double digits in both segments, and raised its full-year guidance for the second consecutive quarter. MDM’s 2026 forecast finally caught its breath after five straight upward revisions, though it still edged higher. NAW put numbers to the industry’s AI adoption with its inaugural index of 302 respondents. And Distribution Strategy Group looked one step ahead: what happens when the customer is an AI agent?
We also put Benj Cohen, founder and CEO of Proton.ai, in the guest chair for In the Mind of a Distributor. Dasha Shakov, our VP of marketing, asked about the ride-alongs he grew up on, the four profit levers AI can move, and why the first AI project should pay for the next one.
That conversation sets up this month’s theme: AI agents are moving from answering questions to finishing the job.
Let’s get into it.
In our ears
Benj Cohen on why distributors should redesign workflows around what AI agents can finish
AI pays off when you redesign a workflow around what an agent can finish from end to end.
Benj Cohen explains how AI connects the four profit levers in distribution: increase price, improve product mix, sell more, and reduce operating expense. The bigger opportunity is to rebuild workflows across the business around the same data.
A surprising insight? Seasoned reps may be your best AI users. They already know what a good answer looks like, so they can test an agent’s work and decide where a person needs to review or approve it.
Start with one measurable use case, a clear owner, and data you can reach. Prove it works. Let that first win pay for the next one.
Listen to the full episode
AI, captain!
Could an AI agent buy from you?
Yes, if it can read and trust your product data.
Brian Hopkins, chief operating officer at Distribution Strategy Group, argues that AI agents are starting to reshape how buyers evaluate suppliers. They compare product specs, price, and availability before a person steps in. “The website’s job used to be getting found. Now the job is answering,” he writes.
Try his Monday-morning test: pull up your 10 highest-volume SKUs and read them like a machine would. Are the specs complete? Is availability accurate? Is the price current?
If the answers live in a rep’s head, a supplier PDF, or a spreadsheet nobody syncs, start there.
Proton connects the systems where your product data, customer-specific pricing, and availability live. That gives your reps better answers today and gives AI buyers reliable data tomorrow.
Things SKU should know
August had range
Grainger crossed $5 billion, then went shopping. Second-quarter sales reached $5.021 billion, up 10.3%, and both segments grew double digits. Three weeks later, Grainger paid $210 million in cash for AWM technology assets. Grainger plans to bring the inventory-management technology into High-Touch Solutions and run a commercial pilot over the next several months.
Distribution got two stock-market moments in two days. ADI began trading as an independent company under the ticker ADIG on August 4. Ferguson joined the S&P 500 the next morning.
Hillman found 44,000 ways into master distribution. Hillman agreed to pay $315 million for Kanebridge, a master distributor of commercial and military-grade fasteners that sells exclusively through distributors. The deal would bring Hillman’s industrial addressable market to $3 billion.
The forecast took a breath. Barely. After five substantial upward revisions, MDM’s August update was quiet at the top line. Oil and gas gave back some of its recent gain, while broader upgrades nudged 2026’s forecast higher.
Link(ed) in bio
The only way to make AI load-bearing
Connect the systems it needs to work across.
Benj’s recent LinkedIn article gives you a simple way to grade an AI project: feature, product, or architecture.
Level one adds AI to software you already own. Level two gives AI a specific job, such as entering an order. Level three connects your systems on one data foundation, so an agent can finish work across them.
A collection of useful level-two tools can still leave a person moving data between systems. Start with one use case that pays for itself. Build it so the next agent can reuse the same map of your business.
New on the shelf
What should your first agent do?
Pick the job somebody has to do every Monday.
Proton’s new agent builder lets you describe the job in plain English, choose which tools it can use, and put it on a schedule. We’re already seeing some killer use cases in production. A quote chaser finds unanswered quotes and drafts the follow-up. A meeting booker builds the calendar from wallet share and churn risk.
Those jobs cross systems. The new navigation puts Proton CRM, Order & Quote Entry Automation, Pricing AI, Inventory AI, and agents in one simple platform, so agents can work across the business.
See the new platform in action firsthand
Also on the shelf: Proton PIM. See how Proton PIM standardizes a 50,000-SKU catalog and routes the exceptions to your team.
This month’s fun fact
The first ecommerce shopping cart sold a Sting CD 32 years ago last month.
Rewind to 1994, when a Philadelphia shopper typed his credit card number into NetMarket and bought Sting’s Ten Summoner’s Tales for $12.48. The New York Times said it was the first secure retail purchase of its kind. The team behind the site celebrated and a new era of shopping began.
In a single working lifetime, ecommerce went from one Sting CD to AI agents buying autonomously.
Stay in touch
Just because the Dispatch is monthly doesn’t mean the conversation has to be. Connect with Benj on LinkedIn for what he is learning from distributors each week.
Got a story we should cover next month? Hit reply and send it over. What landed, what didn’t, what you want more of. I read every reply.
Until next time
Shelby KinneySr. Manager, Editorial, Brand, & Content, Proton.ai
