I joined Dify in January 2025 as the first US-based GTM hire. The product had real momentum — a large open-source community, a global download curve that wouldn't flatten, and inbound conversations from North American companies that nobody on the team was in a time zone to answer.

What didn't exist: a commercial operating playbook tuned for that audience. The Asia and EU business had been built around a different buyer, a different sales cycle, and a different relationship to open source. The NA opportunity wasn't a copy-paste of either.

My remit was deliberately wide. I owned the first definition of what "North America" actually meant for the company — the buyer, the narrative, the partner shape, and the operating cadence behind all three. Eight months later that work shifted shape and I moved into a Chief of Staff – Strategy & Partnerships role; the underlying job is the same.

The framing most people inside the company defaulted to was "we need to build sales in NA." That framing was wrong, or at least incomplete. The real problem was upstream of sales.

The interesting commercial demand in North America was shifting. The original audience — individual developers using the open-source build — had given the company its global brand. But the people writing the checks in North America were no longer that audience. They were centralized IT leads, AI platform owners, automation buyers inside large companies. Risk managers, not enthusiasts.

The existing positioning, pricing, and motion were all tuned for the first audience. They actively undersold the company to the second. A buyer evaluating an internal AI platform for a regulated org doesn't want to hear about velocity or autonomy. They want to know whether their own job survives the rollout.

So the real problem wasn't "build sales." It was figure out which buyer is actually paying, and rewrite the GTM for that buyer — before we hired against the wrong assumption. That reframing was, in retrospect, the most leveraged thing I did all year.

1 — ICP re-segmentation.

I rebuilt the ICP from the buyer side, not the user side. Centralized IT leads inside mid-to-large enterprises — formerly BI or platform engineering, now drafted into "own AI." Understaffed, overloaded with internal requests, measured on absence of failure rather than presence of success. I rewrote the public narrative and the sales-enablement language to land specifically with that profile: governance, traceability, long-term maintainability, the ability to look good internally. Hype was cut. "AI replaces humans" was cut. The shift was uncomfortable inside the company — it meant deliberately under-indexing on the audience that built our brand.

2 — Choosing where PLG ends and SLG begins.

The instinct in NA is to pick one: pure product-led or pure sales-led. I argued for a hybrid, and wrote the seams down explicitly. Open-source momentum remained the top of funnel and the trust-builder; nobody in this market gives you the meeting without proof you ship code. But the conversion from individual user to enterprise contract was a different motion entirely — it needed sales-led handling, design partner conversations, and a security/compliance posture from day one. Treating those as one funnel was the mistake; treating them as two systems sharing a top was the unlock.

3 — Partnership map as the first-30-days trust bridge.

A non-US-headquartered Series A platform doesn't get the first meeting on logo strength. It gets it on who vouches for it. So I mapped the partnership graph as a real artifact, not a deck — cloud ecosystem partners, systems integrators, ecosystem advisors, design-partner candidates — and prioritized which relationships actually shortened the first 30 days of a buyer conversation. By the time I moved into Chief of Staff, that map had grown past 50 mapped contacts across AI, devtools and cloud, with named owners and a real cadence.

4 — Operating cadence: the doc nobody asked for.

The artifact that mattered most wasn't a deck. It was the operating rhythm: a weekly market-intelligence read and a biweekly strategy memo to the founders and to investors, written from scattered customer signals — calls, GitHub issues, partner conversations, competitive moves. The memo became the thing leadership actually used to make decisions. It also became the artifact that made the work transferable when the team scaled.

What I leave behind matters more to me than what I personally close. Five artifacts became durable — they exist independently of whether I'm in the room.

• ICP segmentation map — the buyer profile rewrite, with persona priorities, fears, and the language that lands vs. the language that backfires.
• First-90-days NA partnership playbook — who you talk to, in what order, what you ask for, and how you measure whether the meeting actually shortened a buying cycle.
• Biweekly strategy memo cadence — the artifact format and the rhythm. Originally written for two readers; now read by the whole leadership team and lead investor.
• Internal cloud GTM positioning rewrite — the language layer that landed with risk-managing IT buyers without alienating the developer audience.
• Operating doc that scaled past one author — the part that I'm proudest of. The cadence kept running after I moved out of the seat.

The test of an early operator isn't what they shipped. It's whether the system they left behind still produces decisions without them.

The harder part to write down is the part the numbers don't capture. The biweekly memo became the artifact the founders and the lead investor opened first on a Monday. That's a shift in management trust, not a metric — and it's what unlocked the move into Chief of Staff at month eight.

The other thing the numbers don't say: the operating doc kept running after I rotated out of the seat. Someone else writes it now. That's the outcome I actually care about — the proof that the work wasn't a person, it was a system.

I came in thinking the job was to close deals and build a partner pipeline. Both happened, but neither was the load-bearing part of the year. The load-bearing part was the cadence — sitting down, every other week, in front of a half-empty doc, and forcing scattered customer signals into a decision the founders could actually act on. Nobody briefed me to do that. Nobody graded me on it. It was the work that compounded.

The lesson I'd give the version of me from January 2025, if I could:

Being first-in is mostly about being the operator who keeps writing the doc nobody asked for — until the doc becomes the company's memory, and then your seat becomes something other people can sit in.