Point of view
Marketing POV
Marketing for physical AI is a category-education problem before it's a demand problem. Operators need to believe the outcome is real and repeatable in their own building, so the work is proof-first. Most of what follows generalises past that, and all of it has been argued in public or paid for in a quarter.
01
Position on the outcome, not the machine
Buyers own inventory accuracy and cycle-count labor. The drone is how we get there, not what we sell.
02
Architecture beats messaging
You cannot message your way out of a category problem. Gather AI read as a drone company because the product hierarchy said drone. Rebuilding the architecture, one platform with three products and a form factor that happens to fly, retired the old description without ever arguing against it. Gartner picked up the company's own category language afterwards.
03
Evidence compounds
One rigorous customer proof point outperforms a quarter of thought leadership. Build the proof, then scale the channels that carry it furthest.
04
Brand is not discretionary spend
Most hard-tech companies chase pipeline and fund brand later. It runs the other way. Demand gen fills the pipeline you have today; brand decides whether anyone still knows your name when the market shifts and you need a different one. A carousel that made an argument, no product shot and no CTA, hit 70.2% engagement against a 3.6% industry average. People engage with a point of view before they engage with a product.
05
Pipeline quality over lead volume
Sales-accepted opportunity is the only number worth arguing about in a QBR. Everything upstream is a means to it.
06
Not all pipeline behaves the same
A badge scan and someone who found you on their own and booked a demo are not the same person, even when they land in the same CRM stage. Web-sourced leads convert considerably better than event-sourced ones. Treating them identically is how one average conversion rate hides two completely different stories.
07
Messaging discipline is infrastructure
Terminology tiered by audience, a banned-word list, and automated governance checks that content runs through before it publishes. Not a style guide. It is the thing that lets a lean team produce at volume without the positioning drifting a degree at a time.
08
Buyers do their homework inside AI
A year ago AI assistants drove none of our inbound demo requests. Now they drive a measurable and growing share, so it stopped being an experiment. Better to show up in the answer than pay to interrupt it.
09
Show me the brief, not the output
AI makes everything look finished. Two lines of instruction come back as three polished paragraphs and everyone is briefly impressed, including the person who wrote the two lines. The brief is where the judgement lives. The rest is formatting.
10
Marketing is a systems job
Positioning, content, demand, and sales enablement are one machine. Tuning them separately is how teams stay busy and stay flat.
11
Hire people who can build their own teams
Early on I thought good leadership meant a big team. I changed my mind completely. Hire people who can build and lead their own, then get out of the way. At Dataiku that took the regional team from 2 to 20+ while regional ARR grew roughly 10x alongside it.