- Published
- Updated
- 10 min read

TL;DR: Amazon cut about 30,000 roles across October 2025 and January 2026, after asking for fewer managers and flatter teams. Big companies are stripping out coordination work. A small business never had that layer: the owner is it. Start by automating the routing, the reminders and the status checks, and keep the judgment with people.
Coordination work is the part of the job nobody puts on a business card. Chasing the follow-up. Checking whether the invoice went out. Moving a meeting because two people can't make Tuesday. Telling the crew what the customer changed. It eats a morning before anything gets built, fixed or sold.
Amazon has spent two years removing layers of people whose job was mostly that. You can't copy Amazon. But you can take the same idea down to a five or fifteen person shop and get some of your week back. First, though, let's get the facts straight, because the first version of this note got the headline wrong.
What did Amazon actually cut?
Roles across the company, mostly corporate, not "16,000 managers." In October 2025, Amazon told employees the changes would "mean an overall reduction in our corporate workforce of approximately 14,000 roles." (Amazon (opens in a new tab)) On January 28, 2026, Beth Galetti, Amazon's senior vice president of people experience and technology, wrote that "The reductions we are making today will impact approximately 16,000 roles across Amazon." (NBC News (opens in a new tab))
That is about 30,000 roles in two rounds. Both messages used the same language: the company wanted to keep "reducing layers, increasing ownership" and cut bureaucracy. That is a description of coordination work being removed.
One more correction. Our earlier version said an internal AI project filled the vacancies. We could not find a primary source for that, so it is gone. NBC News reported that "AI was not explicitly cited" in the January note. (NBC News (opens in a new tab))
What did Amazon say about managers and AI?
Two memos from CEO Andy Jassy set the direction. In September 2024 he asked "each s-team organization to increase the ratio of individual contributors to managers by at least 15% by the end of Q1 2025," adding that "Having fewer managers will remove layers and flatten organizations more than they are today." (Amazon (opens in a new tab))
In June 2025 he wrote about generative AI and agents: "in the next few years, we expect that this will reduce our total corporate workforce as we get efficiency gains from using AI extensively across the company." (Amazon (opens in a new tab))
So the honest version is this: Amazon flattened management on purpose and expects AI to shrink its corporate workforce over time. It has not published a list of which coordination work the software now does.
What is coordination work, and why is it the first target?
It is the routing. Information comes in one door and has to reach the right person, in the right order, on time. The Bureau of Labor Statistics describes the job of a project management specialist in one line: they "coordinate the budget, schedule, staffing, and other details of a project." The median annual wage for that job was $102,320 in May 2025. (U.S. Bureau of Labor Statistics (opens in a new tab))
Put those two facts side by side. Coordination work is valuable enough that big companies pay six figures for people who do it full time. It is also repetitive enough that executives think software can take a large share of it.
Gartner made that bet in public. One of its predictions for 2025 and beyond reads: "Through 2026, 20% of organizations will use AI to flatten their organizational structure, eliminating more than half of current middle management positions." (Network World (opens in a new tab)) That is a prediction, not a measurement, and it is worth reading that way. It also says nothing about who picks up the coordination work afterward.
How are other companies handling coordination work?
Shopify's CEO Tobi Lütke put it into policy in 2025: "Before asking for more headcount and resources, teams must demonstrate why they cannot get what they want done using AI." (TechCrunch (opens in a new tab))
Employees feel the change. In Korn Ferry's 2025 Workforce Survey, "41% of employees say their organizations have cut management layers," and Fortune reported that senior executives are now "managing tasks once owned by middle managers." (Fortune (opens in a new tab)) Read that second part twice. When a company removes the people doing coordination work without replacing the work, it lands on whoever is left.
Who does the coordination work in a small business?
You do. Big companies built a coordination layer because they had to and are now tearing parts of it out. A small business never had one. The owner is the coordination layer, along with whoever answers the phone.
That is the opening. The same tools large companies use to shrink coordination work can give a small team a version of the layer it never could afford. Not a new hire. A set of reminders, handoffs and status checks that run without you holding them in your head.
How do you measure your own coordination work in one week?
Before buying anything, count. Here is the exercise I give owners:
- Keep a tally sheet for five working days. Paper by the keyboard works fine.
- Make a mark every time you chase, remind, reschedule or ask "where is this?" One mark per event. Every mark is coordination work.
- Note roughly how many minutes each one took. Round to five.
- At the end of the week, group the marks. Follow-ups with leads. Scheduling. Status questions from customers. Invoice reminders. Handoffs between you and the crew.
- Circle the two biggest groups. Those are your first automation candidates.
Most owners are surprised by the total. The point is not the exact number. It is knowing which coordination work actually costs you, so you automate that and not whatever a vendor demo made look exciting.
What does AI handle well, and where does it go wrong?
Klarna is the example everyone cites, so here it is with its sources. In February 2024 Klarna said its AI assistant had handled 2.3 million conversations, two thirds of its customer service chats, and was "doing the equivalent work of 700 full-time agents." (Klarna (opens in a new tab))
In 2025 CEO Sebastian Siemiatkowski told Bloomberg that "As cost unfortunately seems to have been a too predominant evaluation factor when organising this, what you end up having is lower quality." A Klarna spokesperson later said he meant the cost of outsourced human agents, not the AI. (Silicon Republic (opens in a new tab)) Our first version said Klarna's quality "collapsed" because of AI. The record is more mixed than that, so we corrected it.
The lesson still holds. If cost is the only score you keep, quality slips, whoever or whatever does the work. Measure the customer's experience, not only the hours saved.
Are small businesses actually getting results from AI?
Many say yes, but few have gone deep. A Goldman Sachs 10,000 Small Businesses Voices survey of 1,256 owners, released in March 2026, found 76% currently use AI. Of those, 93% report a positive impact. "Only 14% say AI is fully embedded in their core operations," and 87% say AI augments rather than replaces employees. Owners named lack of technical expertise among the main barriers. (Goldman Sachs (opens in a new tab))
That gap between "we use a chatbot" and "it runs part of our week" is exactly where coordination work lives. A chat window you open when you remember is not a coordination layer. Something that watches the queue and nudges the right person is.
Which coordination work should you automate first?
Start with work that follows rules most of the time and costs little when it is wrong:
- Lead follow-up reminders. A new inquiry with no reply in 24 hours pings you or drafts the reply for you to send.
- Scheduling and confirmations. Booking links, confirmation messages and a reminder the day before.
- Status digests. One morning summary of what is open, what is late and who is waiting on whom. Often the cheapest coordination work to hand off.
- Inbox sorting. Messages tagged into quotes, support, billing and spam so they land in the right queue.
- Invoice nudges. A polite reminder at 7 and 14 days past due, with you copied.
Keep pricing, hiring, difficult client conversations and anything that needs judgment with a person. Automate the coordination work around those decisions, not the decisions.
How do you connect the tools so coordination work stops falling through?

Most small businesses already have the pieces: an inbox, a calendar, an accounting app, a spreadsheet, maybe a CRM. The coordination work falls through the gaps between them. Someone copies a name from an email into the spreadsheet, and the copy is where things break.
Fixing that is usually API integration before it is AI. Pick one place where each fact lives. Customer details in one system. Jobs in one system. Then let the other tools read from it instead of keeping their own copy.
Once the tools agree, an AI agent has something reliable to work with. When the job has several steps, split it so each agent owns one step with a clear handoff, the pattern our note on multi-agent orchestration walks through.
What does a safe first agent look like?
Small and boring. One task. The tools it needs and nothing else. A stop condition, so it knows when to quit and hand off. A person who reviews anything it is unsure about. A check on every output before it reaches a customer, which is what AI quality gates are for.
Run it in the background for two weeks while you still do the coordination work by hand. Compare. If it caught what you caught, let it take the job. If not, fix it or drop it. That is cheaper than finding out from an annoyed customer.
What does it cost to automate coordination work?
Our first version said most small businesses spend $50 to $500 a month on AI agent tools. We could not find a reputable source for that range, so we removed it. There is no good published figure we would stand behind.
What drives the cost is easier to pin down: how many tools have to connect, whether your data is clean enough to trust, how much review a person still does, how much coordination work is in scope, and who keeps it running when a vendor changes something. Ask any provider what happens when the automation is wrong, and who fixes it at 3am.
Frequently asked questions
Did Amazon replace 16,000 managers with AI?
No. Amazon cut approximately 16,000 roles in January 2026 and about 14,000 corporate roles in October 2025. It had earlier asked teams for fewer managers, and its CEO expects AI to shrink the corporate workforce over time.
What is coordination work?
The routing around the real work: follow-ups, scheduling, status updates, handoffs and reminders. It keeps a business moving but does not produce the product or service itself.
Can AI do coordination work for a small business?
Much of the rule-based part, yes: reminders, scheduling, status summaries and inbox sorting. A person should still own decisions, pricing and difficult customer conversations.
Will automating coordination work replace my staff?
It should not be the goal. In the Goldman Sachs survey, 87% of small business owners said AI augments rather than replaces employees. Use it to take chasing and tracking off people's plates.
Where should I start?
Track one week of chasing, reminding and rescheduling. Automate the two biggest groups first, run the automation alongside the manual way for two weeks, and compare.
What went wrong at Klarna?
Klarna said its assistant did the work of 700 agents in 2024. In 2025 its CEO said cost had been too dominant a factor and quality suffered; the company said he meant outsourced human agents. Either way, measure quality, not only cost.
Where we fit
Kief Studio does AI implementation for teams with repetitive work that follows rules most of the time. We pick the workflow, define what good output looks like and measure how it runs today. Then we build the automation with retries, error handling and quality gates on every output, keep a human path for anything the model is unsure about, and run it for you on LTFI. We do not build projects meant to replace staff with AI.
When the coordination work has several steps, our multi-agent workflows give each agent one task, the tools it needs and a stop condition, with checks between agents and a person in the loop for decisions. For publishing and back-office routines, content operations automation runs the repeatable steps and sends failures to a person.
First conversation is free. No commitment.




