commonsku gathered seventeen distributor CEOs from fourteen states and provinces this year, a group that averaged 42% revenue growth in 2025, well above the industry's 4.2%. When they compared notes on AI, four patterns emerged. One of them is worth its own post.
They called it the middle 80%. Humans handle the first 10% of a workflow, the intent and the context, and the last 10%, the judgment and the validation. AI executes the middle 80%, the volume and the repetition.
Why the split is right
Take a purchase order arriving by email. The first 10% is understanding it: this is a reorder from a customer whose sizes always run across the top, the in-hands date is tight, the art has changed. The last 10% is deciding it is correct before it goes to production and the customer is promised a date.
Everything between those two moments is the middle: opening the ERP, finding the account, keying eleven lines, attaching the art, setting decoration, creating the supplier PO, sending the confirmation. It is the same every time. It is where the minutes go. And it is where a tired person at 4:45 on a Friday makes the transposition that becomes a re-run.
The CEOs' framing puts a person at both ends, where a person is irreplaceable, and takes them out of the part where they add nothing but risk.
Intent and judgment stay human. Repetition does not. That is not a compromise between people and AI. It is a description of where each is good.
The other three patterns reinforce it
The summit's first pattern was "pick one workflow, not the ocean": proof follow-ups, pre-call summaries, supplier rewrites. Small, repetitive, measurable. The third was weekly AI office hours where teams share what worked and what did not. The fourth was that your data is the moat: distributors with structured order, PO and client history outperform those working from scattered knowledge.
Read together, they describe a distributor that hands the middle of one workflow to AI, keeps its people at the ends, learns in the open, and builds up a body of its own rules.
Where the summit's own numbers point next
The same session found that thirteen of fourteen CEOs were using AI for design, and only one for forecasting or operations. The middle 80% of a mockup is a few seconds of rendering. The middle 80% of an order is twenty minutes of keying, repeated a few hundred times a month. The pattern has been proven on the small workflow. The return is on the large one.
How AltOps implements the middle 80%
This is close to a description of how AltOps is designed. The person who does the workflow records it once and narrates the first 10%: why this account is handled this way, what counts as unusual. The agent runs the middle on its own computer across the ERP, the portals and the inbox. The approval points the person set are the last 10%, and when something does not match the rules, it comes back to them with the context attached.
Corrections become rules, so the human share of each run shrinks over time rather than staying a permanent supervision task. The CEOs at the summit found the pattern. The job now is to apply it to the workflows that actually fill the week.
Sources: commonsku, "How Distributors Are Using AI: 4 Patterns from commonsku's CEO Summit," 2026; ASI, "Strong Q4 Lifts Promo Industry to Record $27.7 Billion in 2025 Sales," January 2026.
Written by Kavish Soningra, teaching agents to run the back office at AltOps.
