The Company Play — A Worked Example

Step One

A job board is an org chart, if you count it

Most people scan a careers page for a title that matches theirs, find nothing, and leave. The useful move is to count the postings by team. Volume tells you where the money went this quarter, and shape tells you what is being built. Here is that board, grouped:

Department	Roles	What the count is telling you
Solutions	58	The largest department in the company — larger than Engineering. This is a company betting that the bottleneck is deployment, not model quality.
Engineering & Infra	34	Building the platform.
Human Resources	15	A recruiting team that size is what hypergrowth looks like from the inside.
Science	13	Notably smaller than Solutions.
Business	13	Sales. Roughly a quarter the size of Solutions — a technical-led motion.
Legal · Marketing · Finance · Compute · Public Affairs	32	Corporate scaffolding.

Now Look Inside Solutions

Two teams account for the bulk of them.

Deployment Strategist — the largest single title on the board. Titled “AI Deployment Strategist,” open in Palo Alto, Paris, London, Munich, Madrid, Amsterdam, Stockholm, Montréal, Singapore, Seoul, Abu Dhabi, Bangalore, Casablanca, Luxembourg, Marseille. Customer-facing, executive-level, strategy through production.

Use-Case — the other half of the same build-out. “Applied AI, Forward Deployed Machine Learning Engineer” and “Technical Lead, Forward Deployed AI Engineer.” The build side of the same motion.

That is a Solutions Architect organization and a Forward Deployed Engineering organization, being stood up in parallel, across four continents, right now. It is the identical structure this client had already built, from zero, at a hyperscaler.

The Conclusion That Matters

The client does not apply for AI Deployment Strategist. That is an individual contributor role, and for a leader already running a North America organization it is a step down.

They go in as the person who owns those hires in the Americas — scaling the team, setting the delivery model, and owning the deployment motion end to end. That role is not posted. It rarely is at this stage: the reqs get opened for the doers first, and the leader gets hired when the volume becomes unmanageable. Twenty-one open strategist roles is what unmanageable looks like about ninety days before anyone admits it.

Step Two

Their words, and the client’s

Lifted verbatim from the AI Deployment Strategist — USA posting in Palo Alto, set against how this client had described their own work on our first call — before either of us had read the posting.

Mistral’s posting

“Lead executive-level workshops to identify business challenges… Co-create AI adoption roadmaps with customers, articulating the ‘art of the possible’ and a clear path to value.”

“Architect end-to-end AI solutions… ensuring seamless transitions from presales to postsales.”

“Act as the bridge between customers and Mistral’s internal teams, synthesizing feedback to influence product and research roadmaps.”

“Develop reusable assets, best practices…”

The client, unprompted

Used the phrase “the art of the possible” for their own method — the same words as the posting, arrived at independently.

Had staffed an AI innovation booth where CIOs and CTOs brought their adoption problems directly.

Was pulled into strategy sessions by design-thinking consultants because, in their words, “they don’t know the tech” — and was the one who could say whether an idea was buildable.

Had partnered with product and engineering to prioritize features and build enablement material.

Why This Is The Whole Argument

“The art of the possible” appears in both columns. You side got it from the other. It is the phrase this category of work uses when it is being done properly — and the client was using it before they had heard of the req.

Mistral’s customers are ASML, BMW, HSBC and CMA CGM. Their named verticals are financial services, manufacturing, energy and public sector. This client had led a global semiconductor solutions practice and taken an automotive customer to 100+ generative AI solutions in production. That is two of Mistral’s four named customers already on the resume, before the conversation starts.

Step Three

Five tracks, running at once

Not one list of people. Five different kinds of conversation, each with its own tone, running in parallel. Mixing them up is how outreach at this level fails — the message that works on a founder is the wrong message for an operator, and vice versa.

Track One — Direct

The C-suite. At a company this age the founders still make the org-design decisions, and a seat that does not exist yet can only be created at their level. All three founders are Paris-based and third-degree. You are not asking them for a job — you are telling them something true about their North America build that they cannot see from Paris.

Arthur Mensch ↗ Co-founder & CEO · Paris

Ex-DeepMind staff research scientist. Two recent posts and a LinkedIn Influencer badge — he is publicly active, which is the whole reason a cold approach at this level is not absurd. His own About line is “CEO of Mistral. Try Vibe.”

Timothée Lacroix ↗ Co-founder & CTO · Paris

Eight years at Facebook AI Research across New York and Paris, PhD done inside FAIR, École Normale Supérieure. One recent post. Three years three months in the seat. The technical founder.

Guillaume Lample ↗ Co-founder & Chief Scientist · Paris

Ex-Facebook AI Research scientist, Carnegie Mellon. Three years four months in role. The least publicly active of the three — lowest-yield of the founders for a cold approach.

Marjorie Janiewicz ↗ Chief Revenue Officer · SF Bay Area

The only one of the four in the United States, and the executive the entire North America commercial organization reports into. Title confirmed through two Bloomberg interviews. If any single message has to land, it is this one.

Track Two — Peer To Peer

The operators. Connection requests sent as an equal. The frame is “I am exploring what you are building, here is where I think I could help, and I would like to hear how you are finding Mistral” — never “I would like to work for you.” That distinction is the entire track. These are people who will remember you and vouch, and who are the wrong door to knock on as a candidate.

Ian Mathew ↗ Head of Deployment Strategy, North America · San Diego

Not a peer — the operator who owns the NA org. States a team of eight scaling to roughly fifteen, and technical lead on Mistral’s first eight-figure North America deal. His entire organization is half the size of the one this client already runs. Came out of neuroscience research at Harvard, not enterprise software. He accepted a peer-framed connection request in under twenty minutes.

Nico Javelaud ↗ AI Solutions Architect · Mistral

The closest true peer in the US. His own headline: “At Mistral, I help scale the company’s North American presence.” He is doing today what this client did at a hyperscaler in 2019. Start here — the background is recognized instantly.

Adrien Perrigault ↗ AI Deployment Strategist Lead · Paris

Closest to the client’s level of anyone below the C-suite. Thirteen-plus years, describes C-level customer management and AI implementation. A Lead, not an individual contributor. If you want to know what the job actually is inside Mistral, this is the conversation.

Tanmay Shishodia ↗ Applied AI · Palo Alto, since July 2025

Ex-Nutanix machine learning engineer, now on the Applied AI team in Palo Alto. Junior to the client, and useful precisely for that — people fourteen months in tell you what the culture is actually like.

Michael Hoff ↗ SVP, Partners and Alliances · Seattle

States a global team of 120+ sellers. Partnerships is how a frontier lab reaches enterprises it cannot cover directly, and managing 50+ global system integration partners was on this client’s own record.

Track Three — Talent

The internal recruiters, and this one is counter-intuitive. A headhunter never approaches a company’s talent team — they are the competition. You are not a headhunter. For a candidate at your level the Head of Talent is a legitimate front door, and the only person in the building whose job is literally to find you. They know which unposted seats are being discussed, and they are measured on filling senior roles.

Brian Cannon ↗ Head of Talent, US · Mistral AI

His profile headline reads “Talent @ Mistral” with Head of Talent, US listed first in his experience — confirm the exact current title on his profile before you use it. He owns US hiring. With fifteen recruiting roles open company-wide, he is under real pressure to fill senior seats and is paid to take this call.

Benjamin Watrinet ↗ Talent Acquisition Lead · since June 2024

Two years three months in seat, Paris. Ran his own technical recruiting firm before Mistral. The right second name if Brian Cannon is slow — and the one closer to the founders.

Track Four — Investors

The track almost nobody runs, and the one with the best odds. These people speak to Arthur Mensch regularly, they are trusted on senior hiring, and they are measured on Mistral succeeding — so a serious operator is a gift to them, not an interruption. A firm name is not a contact; each of these is the named individual on this deal.

Tone: thesis-level. Why this company, why now, what you see that others do not. Never a resume.

Quentin Clark ↗ Managing Director, General Catalyst

The single highest-value name on this page. Former CTO of SAP, former SVP Engineering at Dropbox, before that Microsoft. He is an enterprise technical operator who became an investor — he understands an operator’s record in about ninety seconds, which is not true of most partners. Named by General Catalyst themselves on the Mistral Series B announcement.

Jeannette zu Fürstenberg ↗ Managing Director, General Catalyst · Europe

Co-led the €600M Series B and named first on General Catalyst’s own Mistral announcement. Leads their European practice, which is where Mistral’s center of gravity sits.

Anjney Midha ↗ General Partner, Andreessen Horowitz — Mistral board member

a16z led the Series A and Midha sits on Mistral’s board, confirmed across independent reporting. A board member’s introduction to a CEO is not outreach — it is a different category of event.

Why This Track Is Worth More Than The Other Four

One relationship here opens a portfolio, not a company. General Catalyst alone also backs Applied Intuition, PhysicsX, Helsing, Hippocratic AI, Adept, Grammarly and Shift Technology — and named every one of them in the same Mistral announcement. Applied Intuition and PhysicsX are physical-AI companies squarely in this client’s other lane.

Nvidia, a16z and General Catalyst each hold dozens of positions across exactly the domains on this client’s map. If a partner decides someone is worth placing, they are not thinking about one seat. That is the difference between applying to a company and being known to the people who fund it.

And ASML led the Series C. They are a Mistral investor and a Mistral customer, and their CEO posted about the round naming all three founders. You were client had led a global semiconductor solutions practice. There is no warmer factual bridge in the file.

Track Five — Engagement

Anyone at Mistral who posts. Right now that is Arthur Mensch with two recent posts and Timothée Lacroix with one. This costs nothing, needs no introduction, and is the only track that works before anyone has accepted anything.

It is also what makes the other two tracks land. A connection request from a stranger is a coin flip. A connection request from someone whose comment they read last week is a conversation.

Step Four

How those four names were found

Nothing here is proprietary. It is Sales Navigator with the filters set properly — and one detail most people get wrong.

Sales Navigator — Lead Search

Current company Mistral AI
Geography San Francisco Bay Area (then re-run with United States)
Seniority level Director · VP · CXO
Job title "Solutions" OR "Deployment" OR "Applied AI" OR "Field Engineering" OR "Customer Engineering"

The Detail Almost Everyone Gets Wrong

Put the Boolean in the Job Title field, never the Keyword field. Keyword searches the whole profile, so “Solutions” matches anyone who ever wrote the word in a job description — you get four hundred results and no signal. Title-field search returns the people who actually hold the job. That single change is the difference between a list and a target.

Then sort by tenure, ascending. Someone eight months into building a function has open headcount and an unsolved problem. Someone six years in does not.

One Caution, Because You Are Employed

Mistral competes with your employer. Keep every message about their problem and your general experience — never about Google’s customers, roadmap, numbers or accounts. Your profile-update broadcasting is already switched off. The same discipline applies here: nothing in writing that you would not want read aloud at work.

Step Five

The sequence — five tracks, fourteen days

This is the part that separates people who get replies from people who send resumes into a portal. None of it is volume. All of it is patience with a calendar.

1

Day 1 — Track Three only. Engage, do not reach. One substantive comment on Arthur Mensch’s most recent post, and one on Timothée Lacroix’s. Not agreement — a point only someone who has deployed this technology inside a Fortune 100 could make. Nothing is sent to anyone this day.

2

Day 3 — Track Two opens. Connection requests to Ian Mathew and Michael Hoff, peer framing, no ask. These are the lowest-risk messages in the whole plan and the most likely to be accepted, because nothing is being requested.

3

Day 5 — Track One, the one that matters. Marjorie Janiewicz. She is the only US-based executive of the four and North America rolls up to her. The message is about her coverage problem, not your search: twenty-one deployment strategist reqs open across four continents, and a North America org still in single digits. One paragraph. No resume.

4

Day 8 — Arthur Mensch. Only after the comment on Day 1 has been live for a week. Founder-level and therefore founder-length: three sentences on what he cannot see from Paris about building a US enterprise deployment organization, and what it costs to get the first year of it wrong. Never send to Mensch and Janiewicz in the same week.

5

Day 11 — Lacroix, and the second wave of Track Two. Lacroix if the technical angle is the stronger one for you; Joel Davidson to round out the operators. Guillaume Lample is the lowest-yield of the four and I would hold him unless something opens up.

6

Day 14 — Track Three, Brian Cannon. Deliberately last, and deliberately different in tone: with a recruiter you can be direct about what you are looking for, because that is the transaction. By this point you have commented publicly, connected with two of his operators, and messaged his CRO. You arrive as a name he has already seen, not a cold profile.

7

Any time — Track Four, and it does not wait its turn. Quentin Clark at General Catalyst is the one to open with: ex-CTO of SAP, ex-Dropbox, ex-Microsoft, and they read an operator’s record correctly on sight. A partner introduction jumps this entire sequence, so if a warm path to any of the three exists, take it on day one and let everything else run underneath it.

8

Everything gets logged. Company, person, track, date sent, date accepted, date replied, next action. Sixteen companies at three to four people each is roughly seventy threads across five tracks. The follow-up is where this is won or lost, and no one holds fifty threads in their head. We build the tracker together and I keep it honest.

Fifteen companies to go. The next one we do live, together, from a blank screen — so you can see it is a method and not a document.