Founder-led — persona-matched, value-first cold
The founder-led overlay on the three wedges. Same AI-Native spine, but Mickey runs it himself and leads with value, not a pitch. The move that makes cold work: a real builder credential earns the first five seconds, a genuine give earns the reply, and the ask comes last. One person, four credentials — the version of Mickey each persona's world respects.
The cold-approach motion (give-give-ask)
- Touch 0PresenceComment substantively on their post before any DM. Cold becomes lukewarm; your name stops being a stranger's.
- Touch 1The givePersona-matched credential + one sharp insight + a real asset (talk, POV, framework). No meeting ask. Close on the literal words "No ask."
- Touch 2Second giveA second piece of value or a sharp question tied to their specific situation. Still soft.
- Touch 3Soft pullOnly now, low-friction: "worth comparing notes?" Converting a relationship, not a cold contact.
Targeting: stack signals to rank accounts (AI hires + recent launch + stack). No observation, no send.
CTO exception — run more direct. The soft "no ask" is the default for Growth/CX/PM, but technical-exec outbound data says lead the CTO harder: timeline/numbers hooks out-reply pain questions ~2.2×, founder-to-C-level peaks on positive replies, and a low-friction permission ask ("worth me sending a 2-min teardown?") beats a calendar invite ~5× and beats pure "no ask." So the CTO variant opens insight-first with one credential line and a soft permission ask. Never open with a pain question or with "here's what my company does." Reserve the podcast invite as a second-tier wedge for prestige CTOs a normal email won't reach. (Sources: Sales.co 1M-email study; Digital Bloom hook-type benchmarks; Column Content CTO teardown; LangChain State of Agent Engineering 2026.)
The credibility map — one Mickey, four credentials
| Persona | Lead angle | Credential to surface | The give |
|---|---|---|---|
| CTO / VP Eng | AI-Native interface layer | re:Invent talk on building AI-Native (agent loops + interface); built & sold an AI platform (Insightera → Marketo/Adobe) and Gainsight PX | a 2-min architecture teardown (permission-based) + the talk; podcast invite for prestige/unreachable CTOs |
| Growth PM / CPO | AI-First Growth (PLG → AI) | co-author of Mastering Product-Led Growth; coined the Click Tax | the PLG→AI evolution POV |
| CX leader | AI Onboarding (activation / TTV) | co-founded Gainsight PX (the CS/CX brand's product arm) | onboarding that lifts activation without feeling automated |
| CPO / PM | Modernize & True Intent (roadmap) | built Gainsight PX (product analytics); co-wrote the PLG book | the "roadmap you can't predict" POV + intent observability |
Variant A — CTO / VP Eng
Most teams ship the agent in a sprint, then burn two quarters chasing consistency. 57% say agents are in prod; 37% run online evals. The gap isn't orchestration, the labs ship that now. It's the eval and intent layer above it, and that's the part you can't rent.
I gave a re:Invent talk on exactly this, and I built and sold the non-deterministic version of it twice before (Gainsight PX, plus an AI platform Adobe bought).
I put together a 2-minute teardown of where this stalls on a stack like yours. Worth me sending it over?
I gave a re:Invent talk on agent loops and the interface layer, and I want real conversations with engineering leaders actually shipping this. Your work on [specific thing] is exactly what I'd want to dig into: what broke, what you'd change on evals and consistency.
30 minutes, recorded, you keep the clips. Want me to send a couple of times?
Run the podcast invite only if the show is real and the episode is genuinely about their work — zero pitch in the room, max two follow-ups. It backfires the moment it reads as bait.
Variant B — Growth PM / CPO
I co-wrote Mastering Product-Led Growth with Wes Bush and coined the click tax. Spent years arguing self-serve was the answer. Where I've landed since: PLG solved access, it broke on outcomes. Users still have to click their way to value, and the activation curve flattens right there.
The AI shift changes the mechanic. Instead of guiding users through the product, the product does the job they came for, and every interaction becomes intent you can act on.
Wrote up the PLG-to-AI evolution here [link]. Thought it'd resonate given what you own. No ask.
Variant C — CX leader
I co-founded Gainsight PX, so scaling onboarding and activation has been my world for years. The pattern I keep seeing now: teams point AI at the support queue, and the onboarding experience gets worse, not better.
The version I think actually works is the opposite. The product walks the user to the outcome they came for, inside the core workflow, so activation goes up and the experience feels better, not more automated.
Put the thinking here [link]. Thought it'd land given what you own. No ask.
Variant D — CPO / PM
I built Gainsight PX and co-wrote the PLG book, so the product roadmap has been my world for a long time. The shift I think is underrated: for 20 years we shipped deterministic features. Spec it, build it, commit a date, it does exactly that. Now the roadmap is full of AI features you can't fully predict, can't always trust, and can't put a clean date on.
The missing instrument is intent observability: seeing the gap between what users actually asked for and what the product did. That's what makes a non-deterministic roadmap something you can commit to again, and it sharpens what to build next instead of guessing.
Wrote it up here [link]. Thought it'd resonate. No ask.
Shared thesis — the non-deterministic delivery crisis (CTO + CPO)
The strongest cross-persona wedge. For 20 years these teams shipped deterministic software: spec it, build it, commit a date, it does exactly that. Now they ship non-deterministic features and every muscle is wrong — you can't promise a date, can't fully trust the output, can't iterate fast because you can't see what the user meant versus what the model did. Trust (evals) makes it shippable; intent observability makes it iterable. The CTO feels this as a build problem (evals, hallucination); the CPO feels it as a commitment problem (dates, the board). Same thesis, two pains.
Guardrails
- Value-first, cold. No meeting ask on Touch 1. The credential and the give do the work; "No ask" is load-bearing.
- Concede Layer 1, not Layer 2. Orchestration/harness is commoditizing (Anthropic Managed Agents). The interface + the data flywheel is what they can't rent. Never concede the interface.
- Trust Lab + Conversational Analytics are early access — message forward-looking, never "available now." Awareness content is the right place to say "here's where this is going."
- Intent observability is the dividend, not the entry. Conversational Analytics reads the Product Agent's conversations, so the agent ships first; intent then sharpens the roadmap — it never writes it, and it's not a standalone analytics bolt-on.
- Mickey's own bio is fair game; customer names are not. Re:Invent / Gainsight PX / Insightera / the PLG book are his credentials. Optibus/MixMax/etc. stay anonymized externally until cleared.
- Per-angle avoid-lists hold. CTO: no "agentic transformation," "production-ready," "reliable." Growth: no "AI-powered," "AI Copilot," "personalization at scale," "PLG is dead." CX: no "deflection," "support bot," "chatbot" — lead activation/TTV against the product budget. PM: no funnel/activation language; intent sharpens, never writes, the roadmap.
- No em dashes in any sendable copy (the bodies above are clean by design).