Last week I attended THE BANG, an event Growth Warrior Capital ran in San Francisco to get the three sides of the AI table in one room: backers, builders, and buyers. It is not a configuration I see often. Most days I am building. Some days I am pitching. Rarely do I get to sit in a room where the people writing the checks, the people shipping the products, and the people deciding whether to buy them are all walking the same hallway.
Sushila and I closed out the day on the mid-market panel with Matthew Swanson from Staff AI, moderated by Rachel Route from GWC. It was honestly one of the better hours I have had on stage. Below are five things I am still chewing on a week later, in the order I keep coming back to them.
1. The mid-market is the frontier almost nobody is building for
There are roughly 350,000 US companies in the $10M to $500M revenue band. Enterprise software was not built for them, and the consumer SaaS wave largely passed them by. They run on spreadsheets, on tribal knowledge, on the people who know how the system works because they have been there for eighteen years. When AI lands here, it lands in places where there is no incumbent vendor to displace and no integration fight to win, just real work being done by hand.
This is the band BettrData has been focused on from day one. It is also where Staff AI is selling on commission, and where the food supply chain panelists were turning $1M-a-year food importers into businesses that found $400K in new opportunities and $200K in savings within two months. The companies in this band do not need a platform play. They need someone to show up and do the work with them.
2. The workforce crisis is the real “why now”
The opening keynote put a number up that has stuck with me: 11,400 people turn 65 in the US every day. Manufacturing alone is projected to have 2.1 million open roles next year. In the “4D” jobs (dull, dirty, dangerous, demanding) the replacement rate is roughly five people leaving for every two coming in.
I have been telling people for two years that the AI opportunity in operations is about throughput. After last Tuesday I think the framing is wider than that. The opportunity is about continuity. The institutional knowledge that lives in a 58-year-old plant manager’s head is walking out the door, and there is no junior version coming in behind them. If AI is anything in this decade, it is the system that captures and runs the knowledge that used to live in people.
3. Technology is the easy part, the business model is the moat
Three different panels said a version of this, and it is the line I keep repeating. AMP Robotics stopped trying to sell robots to recycling facilities because the savings the robots produced were captured by municipalities, and the facility operator never saw the $200K. AMP now leases the facilities themselves. Staff AI does not sell AI seats; they sell quota attainment as a 1099 contractor that happens to be powered by AI. We sell BettrData as data operations outcomes that an operator drives, with the AI underneath.
In every case the technology shift is real. The company that wins is the one who figured out how to package the value so the buyer can actually pay for it. The pricing model is the product.
4. Robots work, humans rule
This came from the robotics panel, and an audience member played it back at the end of the day. The framing one of the panelists used with their customers was that AI is an iron suit for the employee, a tool that makes them more consistent and faster. The frontline worker becomes more important: more consistent, faster, with the parts of the job that wear them down absorbed by the tool.
I think this is right, and I think it is more right for data ops than people realize. The customers who have gotten the most out of BettrData are the ones who bought it to let their data ops lead spend ten hours a week on actual analysis instead of fifty hours a week chasing bad files. The iron suit framing is the right one. It is also the easier sale.
5. Post-sale becomes more critical, not less
This was the part of the panel I was most excited to hear Sushila talk about. The line she dropped that I have been quoting all week, verbatim: “Outputs are not the same as outcomes.” We can hand a customer a beautifully working pipeline and call it a success, and at renewal they can tell us it never moved the P&L number that mattered to them. The pipeline was the output. The number was the outcome.
The customer success team of 2028 is going to look different. The role gets more consultative, and the administrative work shifts to AI. The forward-deployed engineer model, where someone sits inside the customer’s office and imagines new use cases with them, becomes the gold standard. AI handles the CRM updates and the meeting summaries. The humans do the part that was always the actual job: understanding what the customer is trying to do, and making sure the system gets them there.
The thread underneath all five is the same. The teams that win the next decade are the ones who figured out how to put AI inside an operating system that real businesses can buy, run, and renew on. That is the work, and it is happening in the mid-market right now, in companies most of the AI conversation has been ignoring.
Grateful to the GWC team for putting the day together, and to Sushila, Matthew, and Rachel for the panel. The mid-market work is the work I keep coming back to, and last Tuesday was a good reminder of why.
– Aaron Dix, CEO
