HBS AI Institute — About the Frontier Firm AI Initiative
TL;DR
The institutional “About” page of the Frontier Firm AI Initiative, run by the HBS AI Institute at Harvard Business School in collaboration with Microsoft. Undated; captured 15 October 2026. Two parts:
- What a Frontier Firm is. “A new class of organization that puts AI at the heart of its business strategy to reinvent how it operates, innovates, and augments human capacity at scale”, from “a one-person startup with a multimillion-dollar annual run rate to a large enterprise.” Member companies run experiments with Harvard and Microsoft researchers on “AI-first work patterns and functions, starting with Sales, IT, and Finance,” and attend C-suite workshops. Frontier Firms “choose to participate in rigorous research instead of waiting for the playbook to be written by someone else.”
- Early insights from an upcoming case by Iavor Bojinov, Raffaella Sadun and Shunyuan Zhang, in four findings below.
Key claims: the early insights
Adoption needs structured experimentation, and employees will not self-discover
“Unlike standard technology, AI has a broader scope and greater depth of collaboration/automation, requiring a first phase of structured, frontline experimentation to discover new value. Do not assume that employees would self-discover uses without specific guidance.” Experimentation is followed by adaptation, then scaling.
Four managerial problems in running the experiments
- What to experiment on: “wide enough for discovery, narrow enough for scaling.”
- Whom to select: people “who can tolerate uncertainty and short-term unproductiveness.”
- How to codify and scale what is learned.
- How to structure incentives for full-scale adoption.
The Skill–Will 2×2
Employees are placed by skill (digital and AI literacy) and will (motivation to adopt):
| High will | Low will | |
|---|---|---|
| High skill | Evangelists — natural early adopters who pioneer new uses | Skeptics — resist; need reframing or incentives. “Their conversion can have outsized influence due to their credibility.” |
| Low skill | Learners — motivated but need structured support (training, shared prompts); “crucial spreaders of established use cases” | Resisters — avoid experiments; poor experiment subjects, but must be brought along once the technology is settled |
Agents make it harder
AI agents “exacerbate adoption challenges.” They need more substantive onboarding, which means codifying tacit knowledge and adds a need for “agent management skills”. They also create a sharper “winners and losers” dynamic, “potentially increasing the performance gap and resistance among employees.”
Dynamic-capabilities reading
digital-seizing/rapid-prototyping— the first adoption phase is structured frontline experimentation, scoped “wide enough for discovery, narrow enough for scaling,” with people chosen for tolerance of short-term unproductiveness.digital-transforming/improving-digital-maturity— the Skill–Will matrix is a way of identifying workforce digital maturity, with structured support for Learners and C-suite programmes as the levers.contextual/internal-barriers— Skeptics and Resisters are the change resistance the cell names, and agents widen the gap between winners and losers, which raises resistance further.
How it connects
- enterprise-ai-adoption — the Skill–Will 2×2 and the experimentation → adaptation → scaling sequence. Sadun’s WorkLab interview is the spoken version: strategic boundaries, the experimentation team, skeptics and enthusiasts “not necessarily correlated with their technical expertise.” That is the will axis here, held apart from skill.
- Miller (WorkLab) — a frontier unit staffed with “weirdos” from every function; this page picks experimenters by tolerance for uncertainty instead.
- agent-fleet-management — “agent management skills” named as a new requirement, with onboarding that has to codify tacit knowledge.
What was actually ingested
The whole page body (~580 words). The case behind the early insights was not yet published at capture, so the findings are known only as these short bullets, with no method, sample or firm names.
Linked entities and concepts
- HBS AI Institute — host of the initiative. Entity created in this ingest.
- Microsoft — collaborator. Updated in this ingest.
- Raffaella Sadun — co-author of the forthcoming case. Updated in this ingest.
- enterprise-ai-adoption, agent-fleet-management — updated in this ingest.
Dangling (single-source mention, deferred): Iavor Bojinov, Shunyuan Zhang.
Scope and reliability
An institutional page promoting a research programme that companies join, run with a vendor partner. The early insights are pre-publication summaries of a single case, without data. The Skill–Will matrix is a long-standing management-training device (situational leadership) applied to AI. It is useful as vocabulary, not as a finding.