Your field has stopped waiting. So have your customers.

Recently, I was in a room in Europe with more than 1,000 field leaders from lots of different companies. What struck me wasn't the energy (there's always energy). It was that nobody was waiting for permission.

They were working. Muse, Claude, Gemini and ChatGPT, all put to work on their business. They're uploading your comp plan, your incentives and your product data, probably while you're sat in a meeting about your AI strategy.

That's how I opened on the Direct Selling University (DSU) main stage in Dallas this week, and it's the line I keep coming back to. They're not waiting. Most of us are.

The field is already building

I've worked with direct selling companies and field leaders for over a decade, and one pattern holds. Product innovation almost always comes from the company. How we work, and how we get it out there, gets forced by the field, because they jump on trends faster than head office ever has.

That's truer now than it's ever been. In the last six months I've seen field leaders build their own chat assistants trained on the comp plan and product data, then hand them to their teams. Accurate? Compliant? Who knows. They're doing it anyway. I've seen AI avatars and voice clones turned into training resources, and agents that log into the back office, pull the data and push it into whichever model the leader prefers. Some of it is scarily accurate.

Gartner says 57% of workers use personal AI accounts for work, and 33% have put sensitive information into tools nobody approved. Those are employees, with an IT team watching. Your field has no IT team.

Our benchmark with DSN found the same inside this industry. Every organisation in our active engagement set had AI running outside its official tool stack. Without exception. Shadow AI is demand you haven't supplied yet.

The $100 cartoon

A couple of weeks ago, I watched a field leader explain to someone in their team how Netflix grew on word of mouth. Same as us, except in this profession we pay people for the recommendation. The idea was a cartoon that explains network marketing to people who are sceptical, or who just don't get the model.

The team member scribbled it down, said "leave it with me" and spent the rest of the day on a laptop, talking into it. Higgsfield for the characters and scenes, then video, sound effects and voice, with Claude helping edit it all together in DaVinci.

What came back was a 60-second animated explainer. A mid-market studio typically charges $5,000 to $15,000 for that and takes three to six weeks to deliver it. Premium agencies quote up to $45,000. Our field leader did it in a day, for $100.

We're not going to stop these tools developing, and last time I checked, field leaders aren't going to stop looking for ways to make their role easier. That video is one of hundreds of examples I've seen this year.

Your customers' agents are coming

And it isn't only your field that's stopped waiting. Your customers have too.

In early September 2026, Meta launched Muse in the US, an agent that shops, pays, reads your email and runs your calendar. It went to number one in both app stores and Meta shares closed up over 11%.

On 20 September, Amazon blocked it. Shoppers got a pop-up telling them an unauthorised AI agent broke Amazon's conditions of use. A day later Shopify went the other way and switched on agentic checkout through Shop Pay on every Shopify store.

Two of the biggest names in commerce made opposite bets in the same week, and the market didn't take long to pick a side. From the close on Friday 18 September, Shopify was up 15% by Tuesday. Amazon was up half a percent. The Nasdaq rose 2.7% over the same days, so Shopify beat the market more than five times over and Amazon trailed it.

They're fighting this hard because AI shoppers are the best shoppers. Adobe tracked over a trillion US retail visits and found visitors arriving from AI convert 60% better and spend 53% more per visit. Twelve months earlier it was the other way round. McKinsey reckons agentic commerce could drive up to $1 trillion of US retail revenue by 2030.

Set that against our own number. Global direct selling retail sales were $163.9bn in 2024, essentially flat.

When a customer's agent goes looking for your product tonight, can it find you and can it buy from you? And does the distributor who introduced that customer get paid?

If you didn't call an all-hands the week this happened, you're missing the point.

Why the P&L isn't moving

On stage I asked the room to keep a hand up for each of these they had live today:

• A field assistant trained on the comp plan, incentives and product data

• On-demand AI reporting

• Hyper-personalised AI onboarding

• Predictive churn and retention

• Answer engine optimisation (AEO) on autopilot

• Model Context Protocol (MCP) connectors that let ChatGPT, Claude, Grok or Gemini into your back office

• An agent-ready catalogue and checkout

• AI compliance checks on content the field makes

Most hands dropped at two. That matches what we see. Every company I work with has some of these, and none has more than three or four. Our benchmark with DSN covered 24 organisations: 17 foundational, 7 developing and not one maturing yet.

The gaps cost real money. Rallyware's field data says a distributor who places a first order in their first 7 days generates roughly twice the lifetime value of one who orders between day 8 and day 90, and up to six times one who waits past day 90. That window is exactly what hyper-personalised onboarding and predictive churn are built for.

So if everyone's doing something, why isn't it showing up on the P&L? You're in good company if it isn't. McKinsey's State of AI 2026 surveyed 1,719 leaders. Around 80% report individual productivity gains, yet only 37% attribute any EBIT impact to AI and just 6% get 5% or more of EBIT from it. PwC found 56% of CEOs saw neither revenue gains nor cost reductions from AI in the previous 12 months.

I think most companies are aiming at the wrong thing. Big transformation programmes, or lots of small pilots that never add up to much.

Where I've seen problems actually get solved is far more basic. If I had a dollar for every time I found someone on over $100,000 doing a task basic automation could have done 10 years ago, I wouldn't need to work.

The clearest predictor in our benchmark of moving from foundational to developing wasn't strategy or intent. It was the number of completed AI initiatives with measured outcomes. The most mature company in our cohort had shipped the most projects, and the least mature had the loudest ambitions. In other words, maturity comes from repetition.

Get the reps in

Yes, I'm going to say it again. Hackathons.

I nearly didn't. DSU Fall was the third event on the trot where I've banged this drum. Then, the week before, an excited executive came up to me. "Your talk on hackathons was sooo good." Great, what did your team build? "We're doing one really soon."

That talk was at DSU. Last year. I wanted to ask what the heck they'd spent 11 months doing. I didn't, because that's not helpful. But in those same 11 months we've run more than 20 hackathons with companies in this industry.

One person who came out of them is a finance controller at one of our clients. He'd been joining our weekly hackathons and proficiency sessions, where we show people how to get a proof of concept up quickly in Lovable. We helped him build a connector into NetSuite, their ERP, and into Asana, plus a couple of other inputs. For the first time, data that lived in separate spreadsheets and needed several people to analyse was in one place.

Then he pushed it. He pointed it at historic purchase ledgers, supplier by supplier, and built an anomaly tracker in hours (not weeks) to spot price changes over time on specific stock items. Bottles, for example. Hidden in the data was a run of small, incremental price increases across a few suppliers.

In three months he saved the business over $350,000. Every time he runs it he finds a new use for the data, and he now expects to pass $1 million in savings before the end of the year.

He didn't need an engineer, the IT team or budget approval. All he asked for was protected time each week.

A project like that would never have survived a budget meeting a few years ago, because nobody could have written the business case for the build. Now you can build a proof of concept and validate the business case for a fraction of the cost and time. How many times has your team done that in the last month?

Two things this month

Every use case I've seen bring real business value back has one thing in common. AI stopped happening to people and started happening for them. Create hundreds of those small aha moments across the business and the P&L follows.

So, two things before the month's out:

1. Count how many items on that list you have live, pick the next one and give it an owner.

2. Put a hackathon date in the diary. A date, not "really soon".

And next time you tell me you liked this, I'm going to ask what your team built.