Chemical AI Should Take the Busywork. People Keep the Hard Cases.
Every chemical distributor is getting the same AI tools at once. The edge left is judgment, and people only build it on the work AI shouldn't take.
The numbers here
Everyone gets the same agent at once.
On September 2, ERP maker SYSPRO launched an AI platform whose agents spot problems across orders, inventory, and suppliers, recommend a response, and carry out the actions a company approves, Distribution Strategy reported. Chemical AI is arriving the way ERP updates do, to everyone on the platform at once.
- 1
Spots the problem
Reads orders, inventory, suppliers, and operating schedules together.
- 2
Recommends a response
Works out what could be done about it.
- 3
Carries out the action
Inside the ERP and the systems around it, once it's allowed to.
Waits for a person
The company can require human approval before an agent takes an action. Distributors are already in its controlled rollout, some with more than $2 billion in revenue.
SYSPRO Puts AI Agents to Work Inside Distributor Operations15 Sep 2026
A tool that reaches every distributor in the same quarter makes nobody faster than the competitor down the road. So the useful question is which slow work to hand it.
Busywork and the hard cases.
Every chemical business runs on 2 kinds of slow work, and they need opposite answers. Open any task to see what doing it again actually teaches.
Hand it to AI
Busywork. It stops teaching after the first month.
- Doing it again teaches
Nothing a second time. The answer was always going to arrive.
Keep it with people
The hard cases. Every one teaches the next question.
- Doing it again teaches
Which grades really swap, and for which customers.
Hand the first pile over and the hours come back to the team. That's capacity, and it counts. The line isn't perfectly clean, though. A new hire learns the product line by re-keying orders for a few weeks. After that, it's typing.
A tidy answer, or the next question.
The hard cases are where people learn which question to ask next. A general model hasn't had that training.
A customer calls about haze in a batch of resin.
- Moisture
- Incompatible additives
- Storage temperature
Every item is true. None of it says what to do by Friday.
- Which lot did it come from?
- Which warehouse shipped it?
- How long did it sit on a dock in August?
A model can answer each of these. It won't think to ask them.
An illustration of the difference, not a recorded exchange.
Shortages trained the last round of experts.
People build that instinct on cases that refuse to go smoothly. In a European Coatings interview published September 22, Guild CPO's EMEA director described how Covid shortages pushed technical teams that had resisted changing raw materials into qualifying substitutes.
“Identifying alternatives was of secondary importance until Covid.”
“They experienced a lot of alternative solutions and scientists have been trained in approving them faster... this is now valuable know how.”
Guild CPO: Identifying alternatives was of secondary importance until Covid22 Sep 2026
Nobody would pick a pandemic as a training program. It still shows where the skill comes from. Teams learned substitution because there was no way around the work.
The people who work the hard cases are getting older.
Census Bureau researchers tracked how much employment sits at firms where at least a quarter of workers are over 55. In manufacturing and wholesale trade it more than doubled. The study came out last December, and a shift built over 22 years doesn't reverse in one.
- Utilities35% in 200680% in 2022Utilities: 35% in 2006, 80% in 2022.
- Manufacturing and wholesale tradeIncludes chemicals14% in 2000Over 40% in 2022Manufacturing and wholesale trade: 14% in 2000, Over 40% in 2022.
- Retail trade14% in 2022Retail trade: 14% in 2022.
- Accommodation and food servicesAbout 10% in 2022Accommodation and food services: About 10% in 2022.
The Census gives a start year of 2006 for utilities and 2000 for manufacturing and wholesale. For retail and food service it gives only the 2022 share.
Source: U.S. Census Bureau, Business Dynamics Statistics of Human Capital, published 2 Dec 2025. Read 1 Oct 2026.
22% of the 1.5 million people in US chemicals manufacturing were 55 or older in 2025.4
The Census row groups all of manufacturing and wholesale, so it isn't a chemicals-only number. This one is.
The person who asks about the dock in August is often the one nearest retirement. If the people coming up behind them spend their days approving whatever the model drafts, that question retires too.
Wholesale hasn't decided yet.
There's still time to choose where AI goes. Statistics Canada's spring survey found wholesale near the bottom of every industry it measured, even as adoption across the economy more than tripled in 2 years.
Share of businesses using AI to produce goods or deliver services, past 12 months
Wholesale trade sits second from the bottom, at less than half the rate of businesses overall. In that market at least, most wholesalers haven't decided where AI goes.
Wholesale Trade Trails Nearly Every Industry in Canadian AI Adoption15 Sep 2026
Sort last week's work.
It takes an afternoon, and it's the same sort this page just made, on your own tasks.
- 1
List last week
Write down the 10 tasks that took your team the most time.
You end with10 tasks, in your team's own words
- 2
Ask one question
Next to each, answer it. Would doing this again teach anyone something?
You end withA yes or a no beside every task
- 3
Hand over the no's
The tasks marked no are where AI goes first.
You end withHours back, from work that was teaching nobody
- 4
Pair up on the yes's
Put your newest person beside whoever has done each one longest, for the next 3 cases.
You end withThe next question, learned while the person who asks it is still here
Try it on one task
Type a task from last week, then answer the question. Nothing you type leaves this page.
Where AI goes first
Pair someone new with whoever does it best
Keep the hard cases in the building.
Vailent builds AI-native infrastructure for the chemical industry. It's software that runs above the systems a company already uses, grounded in that company's own data, embedded where the work happens, and governed by the same rules a person would follow.
That's where the busywork belongs. The hard cases stay with your people, and so does the judgment they build.
What AI-native infrastructure isSo which hard case in your building can only one person still work?
Sources.
- SYSPRO Puts AI Agents to Work Inside Distributor Operations Distribution Strategy, reporting the launch.Published 15 Sep 2026 · Vailent checked 1 Oct 2026
- Guild CPO: Identifying alternatives was of secondary importance until Covid European Coatings, an interview by Damir Gagro.Published 22 Sep 2026 · Vailent checked 1 Oct 2026
- Workforce Is Aging, Especially in Some Firms U.S. Census Bureau, Martha Stinson and Sean Wang, from the Business Dynamics Statistics of Human Capital.Published 2 Dec 2025 · Vailent checked 1 Oct 2026
- Employed persons by detailed industry and age, Table 18b U.S. Bureau of Labor Statistics, Current Population Survey, 2025 annual averages.Annual averages for 2025 · Vailent checked 1 Oct 2026
- Wholesale Trade Trails Nearly Every Industry in Canadian AI Adoption Distribution Strategy, reporting Statistics Canada's Canadian Survey on Business Conditions (9,251 responses from a sample of 21,105).Published 15 Sep 2026 · Vailent checked 1 Oct 2026
See what AI-native operations look like
Vailent embeds AI into the daily work of chemical and polymer distributors and producers: quoting, onboarding, inventory, and daily intelligence.
Book a Demo