記事一覧へ
# AI時代のCEOの役割
2024年、Paul Grahamはすべての創業者が誰かに転送したエッセイを書いた。「Founder Mode」だ。
基本的な主張はこうだ：会社の運営に関する従来の常識——委任し、優秀な人を雇い、邪魔をしない——は間違っていた。最も優秀な創業者は現場に留まった。深く関わった。最高の創業者は執念深いから、ルールを破った。
それは共感を呼んだ。なぜなら真実だったから。
Founder Modeは免許証だった。現場に残り続けたいという創業者の本能は、性格上の欠陥ではなかったと言ってくれた。それが競争上の優位性だったと。
Paul Grahamが見抜けなかったこと：AIがFounder Modeを時代遅れにしようとしている。
Founder Modeで生きた創業者たちはプロのマネージャーたちから抜け出した。今、次に何が来るかを理解した創業者たちは、古いモデルで動き続けている他のすべての創業者を含む全員から抜け出すだろう。
ギャップはすでに開き始めている。それが**Architect Mode**だ。
## なぜ放置するとコストがかかるか
この新しいモードが何であるかに入る前に、無視するとどれほどのコストがかかるかを理解する必要がある。
私が共に働く、すでにArchitect Modeにいる創業者たちは、ほとんど不公平に見えるレバレッジ比率で動いている。より一生懸命働いているわけではない。まったく異なるレイヤーで働いており、競合他社が問題から問題へと走り回る間、継続的に学習するシステムを設計している。
ギャップは来ていない。すでに開いている。そして他のほとんどの競争上のギャップとは異なり、このギャップは毎時間複利で膨らむ。世界は今やデジタルだが、あなたはアナログにとどまっている。
これは5年後の予測ではない。AIはすでに分析、下書き、調整、実行のコストを崩壊させている。毎月、今あなたが人で行っていることの多くがシステムで可能になる。2024年と同じ組織図を持っているなら、すでに遅れている。そしてそのギャップは毎週広がっている。
## モデルは死んだ
あなたは情報の非対称性を背景にビジネスを構築した。顧客が知らないことを知っていた。大手より速く動いた。優秀な人を雇って問題に向けた。デジタルな組み立てラインのように見えた。それがゲームだったし、あなたはそれが得意だった。
そのゲームは終わった。
適切なニッチと少数精鋭チームが意味したラップトップライフスタイル時代は、ほとんどの人が認めようとしているよりも速く閉じている時間の窓の上に構築されていた。
問題は変化するかどうかではない。強制されて閉鎖する前に変化するかどうかだ。
## リーダーシップの3つの時代
どこへ向かうかを理解するには、どこから来たかを見る必要がある。3つの異なるCEOリーダーシップモデルがあり、それぞれが時代への対応だった：
**Manager Mode**：ほとんどの企業が今も動いているモード。情報はリレーノードを通じて上方向に移動する。顧客の洞察はICレベルで始まり、チームリーダーに圧縮され、マネージャーに再フレームされ、ディレクターに要約され、最終的に元のシグナルのフィルタリングされたバージョンとしてCEOに届く。到着する頃には、3人の人間がすでにあなたが知る必要があることを決めている。ボトルネックは承認だ。コストはスピードと正確さだ。
**Founder Mode**：ピラミッドをフラット化した。創業者はすべてに直接触れた——これは官僚主義の問題を解決したが、新しい問題を生み出した。すべての決定、すべての品質チェック、すべての方向性が一人を通る。30人ほどまでスケールして、それから壊れる。ボトルネックはもはや官僚主義ではない。それはあなただ。
**Architect Mode**：どちらでもない。インテリジェンス層が中心に座る。すべての機能ユニットがそれに直接接続する。各ユニットは自己完結している——システム設計を所有するアーキテクトと、それを運用するアセンブラー。顧客チャーンイベント、却下された提出、失敗した営業コールからのシグナルが中心に直接入り、行動する必要がある人に反映される。リレーなし。劣化なし。そして中に入るすべてのシグナルが次のシグナルのためにシステムをよりスマートにする。
Architect Modeが何でないかに注意する。それは「AIに会社を動かせ」ではない。システムであることではなく、システムを構築する規律だ。Founderのビジョン、アーティストの趣味、人間の説明責任をAI支援の分析、機械支援の実行、システム支援のスケールと組み合わせること。
## Architect Modeが実際にどう見えるか
すでにこのように動いている企業の3つの例：
これらの企業はどれもカスタムLLMを構築していない。AIリサーチチームもない。正しい質問をするアーキテクトがいる：「AIをどう使えばより速くなれるか」ではなく「何かが起きるたびにスマートになるシステムをどう構築するか」、そしてそのループを閉じた。
**セールスリーダー2.0**
営業リーダーがかつてしなければならなかったことを考えてみよう。週末に通話録音を聞く。反論のスプレッドシートを作る。各担当者を個別にコーチングする。感覚と不完全なデータで（誰もCRMを更新しないから）どのスクリプトが機能しているか推測する。問題の兆候が明らかになってから6ヶ月後に下位パフォーマーを解雇する、全体像を明確に把握したことが一度もなかったから。
これが仕事だった。戦略ではない。構築でもない。コールで起きていることと経営陣が実際に知っていたことの間の情報ギャップの管理だ。
Architect Modeでは、その仕事はもう存在しない。すべての通話が終了した瞬間に文字起こしされ、スコアリングされ、分析される。システムが機能していることをフラグ立てする——クローズドな取引と相関するフレーズ、異議が上手く処理される瞬間、見込み客を前進させるシーケンス。スクリプトを自動的に更新する。レビュー時ではなくリアルタイムで下位パフォーマーをサーフェスする。
最高の担当者と最悪の担当者のギャップはもはやマネジメントの問題ではない。データの問題だ。
今や天才はシステムの中に宿る。創業者や主要な幹部の中ではなく。
## DorseyモードかArchitect Modeか？
Jack DorseyとBrian Haliganは企業で起きていることを命名した功績があるが（本当はHalliganの天才だ）、Dorseyモード——円形組織、継続的計画、趣味での採用——は真の進化であり研究する価値がある。
しかし誰も声に出したくないことがある：
Dorseyはこのフレームワークの信頼性を、まず自分でその問題を作り出してから構築した。Twitterはイーロンが2000人以下に削減して運営を維持する前に7500人に膨れ上がった。Blockはパンデミック中に4000人から10000人超に成長し、その後40%を一度の発表で削減して先見性があると言った。
彼は自分が与えた病気を診断し、今その治療法を売っている。治療法の功績は認める。ただし誰が患者ゼロだったかを忘れないように。
Dorseyモードは、すでに穴を掘ってしまった後にすることだ。Architect Modeは、掘らないようにするためにすることだ。
## CEOの仕事、3つの言葉に
歴史的に、CEOが委任できない5つのことがあると考えてきた：
1. ビジョン
2. 文化
3. 経営チームの採用
4. 資源配分
5. 会社が現金切れにならないようにすること
そのリストは方向性として正しい。しかしAIは今それを圧縮する。
文化は誰を雇い、何を報いて、何を許容するかの産物だ。資源配分と現金管理は実際には同じ仕事：会社を生かし続けながら賭けを選ぶこと。ビジョンはまだ重要だが、AIは戦略の下に位置していた意味のある分析作業を今や行っている。
5つの責任は3つに凝縮される：
**1️⃣ Aim（狙い）：どこへ向かうか？**
AIは地図を改善できる。しかし山を選ぶことはできない。
AIは市場をマッピングし、シナリオをモデル化し、競合を要約し、パターンを特定し、人間のどのチームよりも速くセカンドオーダー効果を示すことができる。それは重要だ。積極的に使え。
しかし戦略はビジョンと同じではない。ビジョンは市場がどこへ向かうかだけではない。あなたがどこへ行く意志があるかだ。
AIは何が合理的かを教えられる。何が意味があるかは教えられない。インテリジェンスが豊富になるにつれて、確信は希少になる。
**2️⃣ Army（人材）：才能の計算式が変わった**
第二の仕事はArmyだ。ここで多くの創業者が高くつくミスをしようとしている。
旧世界では、頭数を増やすことで問題を解決できることが多かった。AIに対して高い代理性と優れた趣味を持つ一人のオペレーターが、今や平均的なパフォーマーのチーム全体を上回ることができる。スキルだけのために採用しているのではない。乗数のために採用している。
**中間管理職**が2つの方法で失敗する。両方とも探さないと見えない。
最初は溜め込み——9人の直接報告を持ちながら委任しようとしない、自分をボトルネックにした、頭数が自分のステータスであるマネージャー。
2番目は盲目——CEOが先に知るまで2レベル下で何が起きているかを知らないマネージャー。
Architect Modeでは、中間管理職の仕事が完全に変わる。彼らはもはや情報ルーターではない。マイクロアーキテクトだ。
**3️⃣ Assets（資産）：希少なリソースはもはや情報ではない**
資産にはお金が含まれる、そうだ。しかしCEOは資本以上のものを配分している。時間、注意、フォーカス、ブランド、信頼、組織のカロリーはすべて資産だ。すべて有限だ。すべてうまく使うことも無駄にすることもできる。
AIはこの仕事の一部をすでに得意としている。モデルをより速く構築する。シナリオをより速くストレステストする。より早く非効率を検出する。つまりCEOは生の分析にかける時間が減り、どの分析が重要かを決める時間が増える。
希少なリソースはもはや情報ではない。それは確信だ。
## なぜほとんどのCEOは苦労するか
ほとんどのCEOはこの移行ができないだろう。なぜなら間違ったものを最適化しているから。
Founder ModeからArchitect Modeへの移行は、主にテクノロジーの採用ではない。アイデンティティのシフトだ。自分の関与によって価値を定義するのをやめ、設計によって価値を定義し始めなければならない。
Founder Modeは自由に感じた。Architect Modeは異なる種類の規律を要求する：自分なしでも動くものを構築する規律——それが生み出すものに対して深く責任を負いながら。
## 最終フレームワーク
アーキテクトはすべてのレンガを積まない。すべての基礎を注がない。建物を個人的に配線しない。しかしアーキテクトは何を構築するかを決め、構造を支えるものを決め、最終的なものが首尾一貫しているかどうかを決める。
**Aim**
**Army**
**Assets**
それが今CEOの仕事だ。他のすべては（委任できるし、すべきだ）。
Founder Modeは死んだ。Founder Mode万歳。
$100Mでまた会おう。🤝
ai-thinkingbusiness-modelagent-ops
AI時代のCEOの役割：Architect Modeへの移行
♥ 96↻ 14
原文を表示 / Show original
THE CEO'S ROLE IN THE AI ERA
In 2024, Paul Graham wrote an essay that every founder forwarded to someone. "Founder Mode".
The basic argument: conventional wisdom about how to run a company -delegate, hire great people, stay out of their way - was wrong. The BEST founders stayed close. They went deep. They broke the rules because the best founders are obsessed.
It resonated because it was true.
Founder Mode was a permission slip. It told founders that their instinct to stay in the engine room wasn't a character flaw. It was their competitive advantage.
Here's what Paul Graham didn't see coming: AI is about to make Founder Mode obsolete.
The founders who lived in Founder Mode pulled away from the professional managers. Now, the Founders who figure out what comes next will pull away from everyone else including other founders still running on the old model.
The gap is already opening
That next thing is Architect Mode.
Before we get into what this new mode is, you need to understand what it costs you to ignore it.
The founders I work with who are already in Architect Mode are operating at a leverage ratio that looks almost unfair. They're not working harder. They're working on a different layer entirely, designing a system that learns continuously while their competitors are still running from problem to problem.
The gap isn't coming. It's already opening. And unlike most competitive gaps, this one compounds hourly. The world is now digital and you're stuck in analog.
This is not a prediction about five years from now. AI is already collapsing the cost of analysis, drafting, coordination, and execution. Every month, more of what you currently do with people becomes doable with systems. If your org chart looks the same as is it did in 2024, you are already behind. And that gap is widening every week.
The model is dead
You built your business on the back of information asymmetry. You knew things your customers didn't. You moved faster than the big guys. You hired smart people and pointed them at problems. It looks like a digital assembly line. That was the game, and you were good at it.
That game is over.
The laptop-lifestyle era, where the right niche and a lean team meant you could print money indefinitely, was built on a window of time that is closing faster than almost anyone in your position is willing to admit.
The question is not whether to change. The question is whether you change before you're forced to shut down.
Three eras of leadership
To understand where we're going, you need to see where we've been. There have been three distinct models of CEO leadership, each one a response to its moment:
Manager Mode is what most companies still run on. Information travels upward through relay nodes. A customer insight starts at the IC level, gets compressed by a team lead, reframed by a manager, summarized by a director, and finally reaches the CEO as a filtered version of the original signal. By the time it arrives, three people have already decided what you need to know. The bottleneck is approvals. The cost is speed and accuracy.
Founder Mode flattened the pyramid. The founder touches everything directly which solved the bureaucracy problem but created a new one. Every decision, every quality check, every direction flows through one person. It scales to about 30 people and then it breaks. The bottleneck isn't bureaucracy anymore. It's you.
Architect Mode is neither. The intelligence layer sits at the center. Every functional unit connects to it directly. Each unit is self-contained, an architect who owns the system design and assemblers who operate it. The signal from a customer churn event, a rejected submission, a failed sales call goes directly into the center and is reflected back to whoever needs to act on it. No relay. No decay. And every signal that enters makes the system smarter for the next one.
Notice what Architect Mode is not. It is not "let AI run the company." It is the discipline to build the system rather than be the system. To combine Founder vision, Artist taste, Human accountability with AI-assisted analysis, Machine-assisted execution, and System-assisted scale.
What Architect Mode actually looks like
Three examples from companies already running this way:
None of these companies built a custom LLM. None of them have an AI research team. They have architects who asked the right question: not "how do we use AI to go faster" but "how do we build a system that gets smarter every time something happens" and then closed that loop.
Sales Leader 2.0
Think about what a sales leader used to have to do. Listen to call recordings on the weekend. Build a spreadsheet of objections. Coach each rep individually. Guess at which scripts were working based on feel and incomplete data (because nobody updates the CRM). Fire the bottom performers six months after the warning signs were obvious, because they never had a clean enough picture to act sooner.
This was the job. Not strategy. Not building. Managing the information gap between what was happening on calls and what leadership actually knew.
In Architect Mode, that job doesn't exist anymore. Every call is transcribed, scored, and analyzed the moment it ends. The system flags what's working, the phrases that correlate with closed deals, the moments where objections get handled well, the sequences that move prospects forward. It updates the scripts automatically. It surfaces the bottom performers not at review time, but in real time, so you don't waste 6 months of wasted leads and missed quota.
The gap between your best rep and your worst rep is no longer a management problem. It's a data problem.
Now the genius lives within the system, instead with the founder or a key executive.
Dorsey Mode?
Jack Dorsey and Brian Halligan (really this is more Halligan's genius) deserve credit for naming what's happening in the enterprise. Dorsey Mode: circular org, continuous planning, hiring for taste, was a genuine evolution and worth studying.
But here's the thing nobody wants to say out loud:
Dorsey is a habitual overhirer who built his credibility for this framework by first creating the problem it solves. Twitter ballooned to 7,500 people before Elon gutted it to under 2,000 and kept the lights on. Block grew from 4,000 employees to over 10,000 during the pandemic, then cut 40% in a single announcement and called it visionary.
He diagnosed a disease he gave himself and is now selling the cure. Give him credit for the cure. Just don't forget who was Patient Zero.
Dorsey Mode is what you do after you've already hired your way into a hole. Architect Mode is what you do so you never dig it.
For seven and eight-figure Founders who never had the luxury of a giant cash reserve to absorb a decade of bloat, the lesson isn't "blow it up and rebuild." It's "build it right the first time": lean, modular, and designed from day one around human judgment with AI leverage.
The org chart you never built is better than the one you have to destroy.
The full comparison:
Pay attention to the moat row. That's the litmus test.
Manager Mode moat is on borrowed time, built on contracts and switching costs that erode the moment something better shows up.
Founder Mode moat is mortal: it lives inside one person and retires when they do. (See Adam Neumann and WeWork).
Architect Mode moat is the only one that gets harder to beat the longer you run it. Every customer interaction, every transaction, every query trains the system. The data is the differentiation. This is the only moat that compounds automatically.
You're not just building a better company. You're building a system that gets smarter with each interaction.
The CEO's job in three words
Historically, I've believed there were five things a CEO couldn't delegate:
1. Vision
2. Culture
3. Hiring the exec team
4. Resource allocation
5. Making sure the company never runs out of cash.
That list is directionally right. But AI now compresses it.
Culture is largely the output of who you hire, what you reward, and what you tolerate. Resource allocation and cash management are really the same job: choosing bets while keeping the company alive. Vision still matters, but AI now does meaningful analytical work that used to sit underneath strategy.
The five responsibilities condense into three:
1️⃣ Aim: Where Are We Going?
AI can improve the map. It cannot choose your mountain.
AI can map markets, model scenarios, summarize competitors, identify patterns, and show second-order effects faster than any human team in history.
That matters. Use it aggressively.
But strategy is not the same thing as vision. Vision is not just where the market is going. Vision is where you are willing to go.
AI can tell you what is rational. It cannot tell you what is meaningful. As intelligence becomes more abundant, conviction becomes more scarce.
This is why founder-led companies are still pulling away. The difference is not that founders have better spreadsheets. The difference is founder-product fit. The best founders care at a level that professionals cannot fake. They transmit belief. They shape product with intensity. They keep going when the obvious path is to stop.
At AppSumo, the companies that really broke out were often not the prettiest or most polished. They were the ones where the founder was still clearly leading the charge. You could feel it. The customer could feel it too.
The product had a pulse because the founder had one.
Architect Mode does not change that. In fact, it depends on it. AI handles the analysis. The CEO still sets the meaning.
AI is the body and brain. The Founder is the heart.
2️⃣ Army: the talent calculus just changed
The second job is Army, and this is where most founders are about to make an expensive mistake.
In the old world, you could often solve problems by adding headcount. More managers. More analysts. More coordinators. More layers. In the new world, that is a trap.
One high-agency operator with great taste and strong AI fluency can now outperform entire teams of average performers. You are not just hiring for skill anymore. You are hiring for the multiplier.
Culture is not a separate CEO responsibility. Culture is the output of Aim plus Army over time.
It is who gets rewarded. Who gets corrected. Who gets fired. And who is building the machine that builds the machine.
Middle managers
Middle management fails in two ways, and both are invisible until you go looking.
The first is hoarding - the manager with 9 direct reports who won't delegate, who's made themselves the bottleneck, whose headcount is their status.
The second is blindness - the manager who doesn't know what's happening two levels down until the CEO finds out first. Both are happening right now in companies that think they have their org under control.
In Architect Mode, the middle manager's job changes completely. They are no longer information routers. They are micro architects. Their job is to take everything happening at the frontline, every customer signal, every friction point, every pattern, and feed it back into the system so the whole organization gets smarter.
The best ones are actively compounding the moat. The worst ones are quietly starving it. The question isn't whether they're a good manager in the old sense. It's whether they're feeding the machine or blocking it.
The new org chart
In Architect Mode, the org chart stops being about reporting lines and starts being about proximity to the machine. There are three layers that matter.
Anyone who doesn't clearly fit one of the three isn't in the new org they're a relic of the old one.
Are you improving or architecting?
Most founders who think they're in Architect Mode are still just improving. The difference is everything.
Improving is running A/B tests on your landing page. Architecting is building a system where every click feeds the model, the model updates the page, every future visitor gets a more personalized experience than the last and none of that requires anyone to touch it. The input of today becomes the output of tomorrow.
This is where the moat actually gets built. Not in the strategy deck. Not in the org chart. In the loop. Every customer interaction that feeds back into your system is a data point your competitor doesn't have. A competitor can copy your product. They cannot copy your loop because your loop is built from your data, your customers, your history. It is the one thing that is structurally yours.
The honest question isn't whether you're using AI. Almost everyone is. The question is whether your AI knows more today than it did yesterday and whether that gap is widening automatically.
If the answer is no, you're not in Architect Mode. You're in Founder Mode with ChatGPT. And Founder Mode, as we established at the top of this article, is no longer enough. (But Aymannn, I'm using Claude. Same difference).
3️⃣ Assets: the scarce resource is no longer information
Assets includes money, yes. But the CEO is allocating more than capital. Time, attention, focus, brand, trust, and organizational calories are all assets. All of them are finite. All of them can be deployed well or wasted.
AI is already excellent at parts of this job. It builds models faster. Stress tests scenarios faster. Flags inefficiency earlier. That means the CEO spends less time doing raw analysis and more time deciding what analysis matters.
The scarce resource is no longer information. It is conviction.
Every company will soon have more dashboards, more options, more strategies, and more machine-generated ideas than they can possibly act on. Most will drown in possibility.
The CEO's new job, the one that cannot be automated, is to decide with conviction: What are we actually doing? What are we not doing? What bet deserves our best people and our best years?
The best CEOs are allowed to be uncertain. They are not allowed to be unclear.
Before and after
Why most CEOs will struggle
Most CEOs will not make this transition because they're optimizing for the wrong thing.
The shift from Founder Mode to Architect Mode is not primarily a technology adoption. It's an identity shift. You have to stop defining your value by your involvement and start defining it by your design.
Founder Mode felt like freedom. Architect Mode requires a different kind of discipline: the discipline to build the thing that runs without you, while staying deeply accountable for what it produces.
Not with better strategy. With better architecture.
The final framework
An architect does not lay every brick. Does not pour every foundation. Does not wire the building personally. But the architect decides what gets built, what holds the structure together, and whether the final thing is coherent.
Aim
Army
Assets
That is now the CEO's job.
Everything else can (and should) be delegated.
I've spent my career at the intersection of where founders get stuck and where they break through. At AppSumo I lived this transition firsthand, scaling from $3M to $84M in annual revenue. Through Agoge I've walked hundreds of seven and eight figure founders through it, driving over $1B in combined growth.
The full playbook, the specific moves, the sequencing, the places where this breaks down and how to fix it is coming soon. Follow along so you can hear about it when it drops.
Architect Mode is not the death of the founder. It's the founder fully realized.
The founder who stays in Founder Mode forever eventually becomes the ceiling of their own company. The founder who moves to Architect Mode becomes the architect of something that outlasts their involvement in any single decision. The machine runs. The loops compound. The moat deepens.
The builder didn't die. They've just built something bigger than themselves.
Founder Mode is dead. Long live Founder Mode.
See you at $100M. 🤝