Accepting final enrollments. Cohort starts in:

days

hours

minutes

seconds

View outline Apply now

A 7-Day Leadership Intensive

Build Your Agent

Agentic AI for leaders, taught by building.

The most practical way for leaders to understand agentic AI: build your own AI Chief of Staff, with an expert beside you every step.

See the full course outline →

Duration

7 Days

September 2 - 9

Live faculty

4 sessions

Including HBS lecture

1:1 support

2 sessions

1:1 with an AI consultant

Investment

$1,650

Accepting final enrollments

No coding requiredPaid Claude plan neededCertificate of Completion included

Study the OpenClaw moment Build in Claude Co-work 2 × 1:1 expert sessions Your own AI copilot

View course outline

Taught by Dirk Hofmann · Stephanie Dick · Suraj Srinivasan Harvard Data Science Review board · Harvard Business School Plus 1:1 build support from expert AI consultants

Trusted by leaders at Amazon · Google · Nike · American Express · Workday

Accepting final enrollments

See exactly what you'll build

Get the full course outline: the seven-day schedule, the three workflows, the support structure, and what you walk away with.

Only 24 places left for this cohort. Starts September 2.

"The best executive course I have ever attended."

Franco Denari, Chief Strategy & Growth Officer

Ready to enrol? Apply directly

"The era of reading about AI transformation, commissioning strategy decks, and attending keynotes that do not translate is ending. What replaces it is experimentation – real pilots, real workflows, real outcomes."

Dirk Hofmann, "The OpenClaw Moment," under review by Harvard Data Science Review

Why now

Autonomous AI agents arrived. Most leaders have read about them. Few have built one.

Research finds only a small fraction of organizations capture value from AI at scale. What separates them is not the platform they pick – it is whether they understand the technology from the inside, and redesign how work gets done.

The shift

AI stopped waiting to be asked. It now acts.

The last generation answered questions. This one monitors, plans, and executes across multiple steps on your behalf. It is the difference between a tool you prompt and a colleague you brief. Most leaders have read about this shift. Few have felt it by building one.

The moment

The agent layer is real, contested, and moving fast.

OpenClaw crossed 350,000 GitHub stars faster than any project in history. OpenAI hired its creator; Anthropic launched a rival and removed OpenClaw from Claude subscriptions. The security questions are real and unresolved. Leaders need to understand this landscape, not just pick a vendor. TechCrunch, 2026

The trap

40% of agentic AI projects will be canceled by 2027.

The gap is not technology – it is organizational redesign. Organizations that redesign workflows around agents are three times more likely to see real financial impact. Most bolt AI onto broken processes and wonder why returns disappoint. BCG/MIT, 2026; Gartner, 2025

Bernard Marr, writing in Forbes: every company will need a strategy for doing business in an environment where AI agents are transforming every job, every organization and every industry. For most, the biggest risk is being unprepared.

View course outline|Apply now

"I left with a clear roadmap for where and how agentic AI can drive meaningful impact in my business. In typical Harvard fashion, they over-delivered."

Jill Tarallo, AI-Driven COO, Rollick · Harvard Data Science Initiative participant

What you will learn

A guided build – with the thinking that makes it transfer.

The agent is the vehicle. What you keep is a framework, a vocabulary, and a clear view of where this is heading – transferable to any platform, long after the course.

The shift

From reactive to proactive AI

Why agentic AI is a genuine break from ChatGPT and Copilot, not an incremental upgrade. You move from prompting a tool to issuing standing instructions – defining an automated role into being. Understanding this distinction is the foundation everything else builds on.

The moment

What the OpenClaw moment signals

The platform war, the adoption curve, and why "every CEO needs an OpenClaw strategy" is the wrong framing. The real lesson is organizational: the leaders who capture value are not those who pick the right tool, but those who redesign how work gets done.

The method

The AGENT framework

A repeatable way to design any agentic workflow: Audit how the work is done today, Gauge the outcome you want, Engineer the rebuild, Navigate the boundaries and governance, Track whether it worked. The transferable skill you keep regardless of which platform wins.

The craft

How an agent is actually built

Brief, write, check, error-handling. Skills, memory, guardrails, connectors. And the most useful lesson of all: you do not get it right the first time. You start simple and iterate in minutes, not days – faculty member Dirk Hofmann walks through his own agent reaching version eight.

The hard part

Security, trust, and the limits of autonomy

Treated honestly, not glossed over. What autonomous agents can and cannot be trusted with, where to keep a human in the loop, and why building in Claude rather than self-hosting is a deliberate safety decision. The question the first audience asked, answered properly.

The horizon

From Chief of Staff to AI Twin

Harvard Business School's Suraj Srinivasan shows where your personal agent leads: from a Chief of Staff that knows your context, to a digital twin of your judgment, to the future of how organizations make decisions. The strategic payoff that reframes the whole week.

View course outline|Apply now

How the course works

A prescriptive, supported build. You ship something real.

Not a webinar you watch. A guided build with a clear method: you apply the AGENT framework to a workflow from your own role, configure it in Claude, and iterate fast. By day seven you have something that works.

What you build

Your AI Chief of Staff, in Claude

An AI Chief of Staff that handles email triage, meeting preparation, and research – scoped to your role and built in Claude Co-work, so you can reach it from your phone or desktop. You configure it, you don't code it: no terminal, no installation. Requires a paid Claude plan.

You are not building alone

1:1 support, so you actually finish.

Harvard Data Science Review board members teach the live sessions. The build is different: two one-on-one sessions with an AI consultant – one to get you over the first hump, one to unstick you and ship. Plus an AI copilot between sessions, and a peer cohort to build alongside.

View course outline|Apply now

What you walk away with

Three things. Each reinforces the others.

Understand

A clear view of the OpenClaw moment

Understand what the OpenClaw moment signals, the security vulnerabilities and constraints that come with autonomous agents, and what responsible adoption requires. Leave able to speak about it from experience, not from a headline.

Build

A working AI Chief of Staff

An agent that handles email triage, meeting preparation, and research – configured in Claude, scoped by you, for your own role. Not a prototype. Something you use the morning after the course ends.

Manifesto

Your agentic AI manifesto

A one-page declaration: what your agent does and does not do, what building it revealed about your organization, and the governance questions you now know need answering. The bridge from a personal agent to an organizational view.

The Chief of Staff is the starting point. Harvard Business School’s Suraj Srinivasan closes the week by showing where it leads: from a personal agent to the AI twin, and what that means for how your whole organization makes decisions.

View course outline|Apply now

The seven days

Four faculty sessions. Two 1:1 consultations. One agent, shipped.

Faculty open and close the week. The AI copilot and your two 1:1 sessions carry you through the build – at exactly the points where most people stall.

Wed, Sept 2

The Spark

Live: copilot to colleague

Thu, Sept 3

The Build Briefing

Live: the A.G.E.N.T. framework

Fri, Sept 4

1:1 consultant: get started

Over the first hump

Sat Sept 5 – Sun Sept 6

Build your agent

AI copilot guided

Mon, Sept 7

1:1 consultant: refine

Unstick and ship

Tue, Sept 8

From Chief of Staff to AI Twin

Live: HBS governance lecture

Tue, Sept 9

Showcase and close

Demo your agent

Gold marks your hands-on build days and 1:1 consulting. Full session detail and build methodology in the course outline.

View course outline|Apply now

Program details

Dates	September 2 - 9 (7 days)
Format	4 live faculty sessions + 2 one-on-one AI consultant sessionsPersonal AI tutor throughout the build
1:1 support	2 sessions with an AI consultantScope your build, then refine and ship
Investment	$1,650 Accepting final enrollments
Pre-requisite	Paid Claude planRequired for hands-on build. Setup guidance in pre-work.
Certificate	Certificate of CompletionFrom Harvard Data Science Initiative
Pre-work	"The OpenClaw Moment" by Dirk HofmannUnder review by Harvard Data Science Review · plus demo videos and AI tutor onboarding

Ready to build your agent?

$1,650 · Includes 2 one-on-one consulting sessions

Apply now

Faculty

Learn from the people who wrote the research

Participants read the thesis before the first session. The author teaches the course.

Dirk Hofmann

Co-Founder & CEO, DAIN Studios. Harvard Data Science Review board member.

Author of "The OpenClaw Moment." Leads opening lecture, deep dive, and co-closes.

Prof. Stephanie Dick

Simon Fraser University · Harvard Data Science Review board member

Historian of AI, computing, and mathematics. The human and ethical dimension of the agentic moment.

Prof. Suraj Srinivasan

Harvard Business School · Chair, Digital Value Lab (D³ Institute)

Corporate governance, AI-driven transformation, and board oversight of agentic systems.

View course outline|Apply now

Who this is for

For leaders who want to build something real – and finish it

You have read and watched enough

You have done the webinars and read the articles. You do not need another overview. You want to build a working agent and you want expert support while you do it.

You want to lead from experience

A leader who has never configured an autonomous agent cannot credibly lead an organization through this transformation. You want the conviction that only comes from having built one yourself.

You want support, not a sink-or-swim

Most people stall on one small technical hurdle and quietly give up. You would rather have two 1:1 sessions with an AI consultant, an AI copilot between sessions, and Harvard Data Science Review board members teaching live – and actually ship something in seven days.

You are senior enough to act on it

Director, VP, C-suite, founder. You can take what you build back to your organization and make decisions about where autonomous agents fit – with evidence, not speculation.

View course outline|Apply now

In their words

From past Harvard Data Science Initiative participants

"I left with a clear roadmap for where and how agentic AI can drive meaningful impact in my business. In typical Harvard fashion, they over-delivered."

Jill Tarallo

AI-Driven COO, Rollick

"The best executive course I have ever attended."

Franco Denari

Chief Strategy & Growth Officer

"This program has been truly transformative. The use of actual artificial intelligence is absolutely amazing."

Paul Michael Miles

Past Participant

"As CIO, it's my responsibility to craft our AI strategy and budget. This course gave me a solid foundation for building and scaling our strategy."

David Friedman

Chief Information Officer

Common questions

Before you enroll

You configure, you do not code. No terminal, no local installation, no debugging alone. You make the decisions about what your agent does and where its boundaries are. Your Chief of Staff is built using Claude. This course develops leadership judgment about autonomous systems, not technical implementation skills.

Yes. A paid Claude plan (Pro or Team) is required for the hands-on build component of the course. This is a separate cost from the program fee. If you do not yet have a Claude subscription, we will provide setup guidance as part of the pre-work.

Security is a core thread, not an afterthought. Autonomous agents that can read your email, access files, and act on your behalf raise real risks – and OpenClaw’s own security disclosures make that concrete. You work through the vulnerabilities and constraints directly: what these agents can and cannot be trusted with, where to set boundaries, what to keep a human in the loop for. Building in Claude rather than self-hosting OpenClaw is itself a deliberate, safer starting point.

No. OpenClaw was inspired by Claude Code – the same agentic architecture, open-sourced and self-hosted. Installing and securing OpenClaw responsibly takes longer than seven days, and its own security disclosures make clear why that matters. Instead, you study OpenClaw as a case study – architecture, security model, governance implications – and build your Chief of Staff using Claude, the platform OpenClaw was modeled on.

Seven days from first session to showcase. Four live faculty sessions (60 min each, including an HBS guest lecture), two one-on-one sessions with an AI consultant, and approximately two hours of AI tutor-guided build work per day during the build period. Pre-work takes one to two hours.

Two one-on-one sessions with an AI consultant who has technical capability. The first, midway through, helps you scope your build and troubleshoot your approach. The second, near the end, helps you refine, debug, and prepare to ship. This is the support that makes building a working agent in seven days realistic rather than aspirational.

Yes. A Certificate of Completion from Harvard Data Science Initiative is awarded to participants who attend the live faculty sessions (or watch the recordings) and submit a completed agentic AI manifesto.

Dirk Hofmann's "The OpenClaw Moment: Why Every Leader Needs an Agent Workforce Strategy," currently under review by Harvard Data Science Review. You read the research before the first session. The author teaches the course.

Simple way to decide: if you want a comprehensive strategy framework for evaluating and governing agentic AI across your organization, the 2.5-week Agentic AI Intensive is built for that – it is broader and more business-oriented. If you want to actually build a working agent in seven days with hands-on expert support, this is built for that – it is shorter, more technical, and more hands-on. Strategy-first, choose the Intensive. Build-first, choose this. Some leaders do both.

Details on cancellation and refunds will be provided at checkout. Contact info@hdsrcourses.org with any questions before enrolling.

Build your agent in seven days.

September 2 - 9 · $1,650 · 2 one-on-one consulting sessions included

Ready to enroll? Apply now