Adept pricing

About

Adept was a San Francisco AI lab founded in 2022 to build action-oriented AI — agents that don’t just answer questions but actually operate software the way a person does. Its flagship, ACT-1, was billed as a “large model for actions”: give it a natural-language request and it would click, type and navigate across real enterprise tools (browsers, CRMs, spreadsheets) to get the job done, coordinating across multiple apps when needed.

The pedigree was elite — co-founders included former OpenAI VP David Luan and several Transformer-paper-adjacent researchers — and the money followed. Adept raised a 65 million dollar Series A (April 2022, led by Addition and Greylock) and a 350 million dollar Series B (March 2023, led by General Catalyst and Spark Capital), reaching unicorn status at roughly a 1 billion dollar valuation and over 415 million dollars raised in total (sources: The Information; TechCrunch; Crunchbase).

Then came the part that defines this page. In June 2024, Amazon hired Luan and the other co-founders — Augustus Odena, Maxwell Nye, Erich Elsen and Kelsey Szot, plus other staff — into its AGI org under Rohit Prasad, and licensed Adept’s agent models and technology (sources: CNBC 2024-06-28; TechCrunch 2024-06-28). Amazon did not formally acquire the company; instead it provided roughly 25 million dollars that flowed through so Adept’s investors could roughly recoup the capital they’d put in (source: Semafor 2024-08-02). The FTC questioned whether the structure amounted to an undisclosed acquisition, comparing it to Microsoft’s parallel hire of the Inflection AI team.

A remnant — roughly one-third of the workforce — stayed on under new CEO Zach Brock and pivoted away from AGI toward enterprise “solutions that enable agentic AI” it could sell today (source: TechCrunch 2024-06-28). But the standalone, founder-led company was effectively wound down. By mid-2026 the pricing page 404s, and even the founder who anchored the Amazon deal, David Luan, had left Amazon’s AGI lab (February 2026; sources: CNBC; GeekWire).

This is therefore a post-mortem, not a live price guide. Adept is a textbook example of a richly-funded frontier-agent startup whose value was realized through a talent-and-license deal rather than a shipped, priced product.

Pricing summary : How Adept’s pricing model worked

There is no public pricing for Adept, and there never was. Across its entire independent life the company published no tiers, no rate card, and no self-serve signup. ACT-1 was offered as limited research / early access, and the only commercial motion was sales-led enterprise inquiry — the homepage’s lone CTA is still “Contact sales” / “Learn more about deploying Adept at work.” The pricing page itself now returns a 404.

That absence is the story. A company that raised over 415 million dollars at a ~1 billion dollar valuation never reached the point of a generally-available, priced product. Pricing was perpetually a future problem, and the future arrived as an Amazon acqui-hire instead.

What makes this different: Adept’s “price” was effectively set once, in private, by a strategic acquirer. Amazon’s roughly 25 million dollar payment (routed so backers could break even on ~415 million dollars invested) plus the value of hiring the founders and licensing the technology was the monetization event. For a buyer, there is no rate card to evaluate — the only way to “buy Adept” today is to license agent technology from Amazon, which now holds it.

Pricing by product

Sales motions across products: sales-led only, throughout. There was no self-serve, no free tier, and no published list price at any point. All figures elsewhere on this page (415M+ raised, ~1B valuation, ~25M Amazon pass-through) are funding/deal facts cited in words with attribution — they are not a customer rate card.

Hidden costs : What Adept buyers actually faced

The honest answer: there is no bill to construct, because there was no published price and no live product to buy. The real “hidden cost” for the market was different — it sat with investors and customers, not with a line-item invoice.

There’s no live Adept bill to model — the Adept pricing calculator reflects that no public rate card ever shipped. To estimate a real usage-based AI bill, compare fully-priced peers in the pricing blueprint.

Pricing evolution : Adept pricing history and changes

Cadence

Tracked range: 2022–present. Adept made zero published pricing changes because it never published a price. Its “pricing evolution” is really a funding-and-exit timeline.

Notable changes

• April 2022 — 65 million dollar Series A; ACT-1 research access, no commercial pricing.
• March 2023 — 350 million dollar Series B at ~1 billion dollar valuation; still no GA product or public price.
• June 2024 — Amazon hires the founders and licenses the tech; ~25 million dollars routed so investors break even; standalone effort wound down.
• July 2024 — New CEO Zach Brock; ~one-third of staff pivot to enterprise agentic-AI solutions, still unpriced publicly.
• February 2026 — David Luan leaves Amazon’s AGI lab. Pricing page 404s; no active standalone product.

What’s unique : Adept’s distinctive (non-)pricing

1. The price was an acquihire, not a rate card. Adept’s monetization happened exactly once, in private: Amazon’s talent hire + technology license, with roughly 25 million dollars routed to make ~415 million dollars of investors whole. The “customer” was a strategic acquirer, not a self-serve user.

2. A unicorn that never published a price. Over 415 million dollars raised and a ~1 billion dollar valuation, yet zero tiers, zero list price, zero self-serve. Adept shows how far frontier-agent capital ran ahead of a shippable, priceable product.

3. Action-as-the-value-metric — never monetized. ACT-1’s whole pitch was actions taken across software, an unusually clean usage metric for an agent. Adept never got to price it, but the metric (actions / tasks completed) became the template later agent vendors actually do bill on.

Strengths & weaknesses

Billing UX : Adept billing controls and transparency

• Billing controls — None publicly. There was no dashboard, no self-serve plan management, and no checkout — the only path was a sales conversation.
• Usage visibility — Not applicable to the public; no metered consumer-facing usage console was ever shipped.
• Payment options — Never disclosed. Enterprise engagements would have been bespoke contracts; today the homepage offers only a “Contact sales” CTA and the pricing page 404s.
• Transparency — price_transparency: sales-only. This is as opaque as it gets — not even an indicative range was published, which is why this page documents funding/deal facts (in words, attributed) rather than a rate card.

Strategic wins : Why Adept’s outcome still “worked” — for some

1. The acqui-hire saved investors from a zero

The deal’s quiet win: Amazon routed roughly 25 million dollars so Adept’s backers could roughly recoup the ~415 million dollars they invested, turning a likely write-off into a break-even. For a frontier lab that never shipped a priced product, that is a meaningfully better outcome than the alternative. See how AI companies’ value capture is shifting in how AI companies structure pricing.

2. Talent + IP monetized without a formal acquisition

Amazon got the founders and a technology license without buying the company — a structure that (controversially) sidestepped a formal M&A process. For the Adept principals, it converted an unmonetized research asset into a senior perch inside Amazon’s AGI org under Rohit Prasad. Related: outcome-based pricing trends.

3. The action metric outlived the company

Adept couldn’t price ACT-1, but its core idea — billing for actions / tasks an agent completes — is now mainstream among agent vendors that did ship. The concept was a strategic contribution even if Adept never captured revenue from it. See choosing the right usage metric.

Areas to improve : Where Adept’s commercialization failed

1. No shippable, priceable product

The fatal gap: despite elite talent and 415M+ in funding, Adept never converted ACT-1 into a generally-available, priced offering. Capability research outran commercialization — a recurring failure mode for frontier labs. See bill shock and cost unpredictability for why a clear, shipped pricing model matters.

2. Zero pricing transparency for buyers

There was never even an indicative price or a self-serve trial, so prospective customers couldn’t evaluate, budget, or adopt without a sales process — a high-friction path for a category that was still being defined.

3. Strategic dependency risk for any early adopter

Any enterprise that built on ACT-1 had no GA path and then watched the founders and core tech leave for Amazon. The lesson for buyers: a richly-funded vendor with no public pricing and no GA product is a continuity risk regardless of valuation.

Key takeaways

1. Funding is not monetization. Adept raised over 415 million dollars at a ~1 billion dollar valuation and still never published a price or shipped a GA product — the two are completely separable.
2. A missing price is itself a signal. “Contact sales” with no indicative range, no tiers, and no self-serve often means the commercial model isn’t figured out yet — which, for Adept, it never was.
3. Acqui-hires are the AI sector’s hidden exit price. Adept’s real “price” was set privately by Amazon (talent hire + tech license + ~25 million dollars routed to investors), not by any rate card.
4. The value metric can outlive the company. Adept never billed for “actions,” but actions/tasks-completed became a standard agent pricing metric for the vendors that shipped.
5. For the category, capability ran ahead of commercialization. Adept is a cautionary tale for frontier-agent startups: build something you can ship and price, not just demo.

UBP implications

1. An “actions completed” meter is the natural UBP unit for software-operating agents — Adept conceived it; later agent vendors operationalized it. If you’re pricing an agent, the action/task is your cleanest value metric.
2. Usage-based pricing only matters once there’s a GA product to meter. Adept shows the prerequisite: you cannot run UBP (or any pricing) without shipping. Pricing strategy is downstream of commercialization — see introduction to usage-based pricing for the foundations Adept never reached.
3. Opaque, sales-only pricing limits adoption in an undefined category. When a market is new, at least an indicative usage rate lowers the barrier to evaluation — its total absence (as with Adept) keeps adoption gated to a slow sales motion.

Sources

• CNBC — Amazon hires execs from Adept and licenses its technology (2024-06-28) (accessed 2026-06-16)
• TechCrunch — Amazon hires founders away from AI startup Adept (2024-06-28) (accessed 2026-06-16)
• Semafor — Investors in Adept AI will be paid back after Amazon hires startup’s top talent (2024-08-02) (accessed 2026-06-16)
• The Information — AI startup Adept raises 350 million dollars at 1 billion+ valuation (accessed 2026-06-16)
• CNBC — Head of Amazon’s AGI lab is leaving the company (2026-02-24) (accessed 2026-06-16)
• Adept official website (accessed 2026-06-16)

Bottom line

Adept is a post-mortem, not a price guide. It built ACT-1, an early and credible “agent that operates software,” raised over 415 million dollars at a roughly 1 billion dollar valuation — and never published a price or shipped a generally-available product. In June 2024 Amazon hired its founders into its AGI org and licensed its technology, routing roughly 25 million dollars so investors could break even, in a structure the FTC probed as a possible disguised acquisition. The remaining team pivoted to enterprise agentic AI under a new CEO, but the standalone company was effectively wound down; by mid-2026 the pricing page 404s. The lesson for pricing teams: funding and valuation are not monetization, and a vendor with no public price and no GA product is a strategic risk no matter how elite its talent.

Want to compare Adept against vendors that actually shipped and priced agentic AI? Browse the pricing blueprint.

Pricing timeline : Major events on a vertical axis

Each milestone below corresponds to a public pricing change, product launch, or material adjustment. Major events use a filled marker; minor adjustments use a faded one.

Pricing page 404s — no active standalone product

Jun 2026

As of mid-2026 adept.ai shows only a homepage and a 'Contact sales' enterprise CTA; the pricing page returns a 404. Post-mortem confirmed: no public rate card, no self-serve, founders long since at Amazon (Luan departed Amazon's AGI lab in Feb 2026).

Remaining team pivots to enterprise agentic AI

Jul 2024

New CEO Zach Brock and roughly one-third of remaining staff refocused on enterprise 'solutions that enable agentic AI' using existing in-house models — still without any public pricing or self-serve product.

Amazon acqui-hire — founders + tech absorbed, company wound down

Jun 2024

Amazon hired CEO David Luan and the co-founders into its AGI org and licensed Adept's models/agent tech. Amazon routed ~25M so investors could break even; FTC probed it as a possible undisclosed acquisition. The standalone AGI product effort was effectively wound down.

Series B at ~1 billion dollar valuation — still no public price

Mar 2023

A 350 million dollar Series B (led by General Catalyst and Spark Capital) made Adept a unicorn (415M+ raised total). Despite the funding, no generally-available product or published pricing ever shipped; the only motion was enterprise inquiry.

Series A — research-stage, no commercial pricing

Apr 2022

Adept raised a 65 million dollar Series A (led by Addition and Greylock) to build ACT-1, a large model for actions. Access was research/early-access only — no plans, no rate card, no self-serve.