Adept
Foundation models for enterprise agents
Overview
Adept builds foundation models for enterprise agents — Action Transformer (ACT) family. Most of the founding team joined Amazon in 2024 under a licensing arrangement; Adept continues to operate with remaining team and IP. Originally targeted at workflow automation use cases.
Use cases
What you can produce with Adept
- Engage Adept's team on an enterprise pilot that automates a repetitive multi-step software workflow, such as re-keying data between a web form and an internal system of record.
- Delegate UI-level tasks to an ACT-family agent that reads a screen, clicks buttons, and fills fields the way a human operator would, instead of writing brittle API integrations.
- Automate back-office processes that live in legacy browser-based tools with no API, using Adept's web-interaction software to drive the interface directly.
- Turn a plain-English instruction like 'pull this quarter's invoices and log them in the tracker' into a sequence of concrete actions executed across business applications.
- Use Adept's agentic data and custom infrastructure as a foundation for building an in-house enterprise agent rather than training an action model from scratch.
- Evaluate the ACT model lineage and published demos as a reference point when comparing modern computer-use agents from other vendors.
ASEAN Perspective
Adept in Southeast Asia
ASEAN-region availability and pricing notes coming soon. Drop the editorial team a note via /contact/ if you can supply local context (Singapore/Malaysia/Indonesia/Thailand/Vietnam).
Adept was an influential AI research company building agents (ACT-1, Fuyu) that could take actions across software interfaces from natural-language instructions. Its work on action-oriented models and multimodal UI understanding was genuinely ahead of its time and shaped the broader agent space.
The critical caveat is organisational: in mid-2024 much of Adept's leadership and key staff moved to Amazon under a licensing arrangement, leaving the standalone product's future direction and availability unclear. There is no broadly available public product or stable API for general users. Of strong historical and research interest, but not something to build a workflow on today.
What people say
Adept is best understood today as a cautionary tale rather than a product you can sign up for. The company raised over 400 million dollars on the promise of its ACT (Action Transformer) models, which were demonstrated navigating real software interfaces from plain-English commands. In June 2024, Amazon hired away CEO David Luan and most of the founding team under a licensing arrangement, paid back Adept's investors, and folded the talent into its AGI division. Engineering head Zach Brock took over as CEO of what remained, roughly twenty employees, and refocused the company on infrastructure for agentic AI.
Because Adept never shipped a broadly available, self-serve product, there is almost no genuine user-review footprint. You will not find meaningful G2 or Capterra ratings, and most discussion on Hacker News and Reddit concerns the Amazon deal itself, which commentators widely read as a hollowing-out acqui-hire rather than a validation of the product. The ACT-1 demos were praised as genuinely impressive at the time, but access remained limited to waitlists and enterprise pilots.
The story kept moving after the deal. By early 2026, David Luan had left Amazon's AGI lab, and four of the five co-founders who joined Amazon had departed, underscoring how unsettled the whole arrangement proved. The remaining Adept entity is still listed as active and continues to position itself around enterprise agent tooling built on its in-house models, agentic training data, and web-interaction software.
Who does this fit? Realistically, only enterprises willing to engage directly with a small team on bespoke workflow-automation pilots, and researchers tracking the history of the agents space. Anyone looking for a supported, off-the-shelf AI agent product should look at more actively developed alternatives, since Adept's long-term roadmap and staying power remain genuinely uncertain.
Summary of public user & expert reviews, compiled by RECATOOLS.
About this listing
This entry was compiled from publicly available data including Adept's official website, press releases, documentation, and reputable third-party publications. RECATOOLS is not affiliated with Adept unless explicitly stated.
Third-party AI tools update their pricing, features, availability, and policies frequently. Information here may be outdated by the time you read this — we make reasonable efforts to keep listings current, but cannot guarantee absolute accuracy.
For the latest details, please refer to Adept directly →
Spotted something out of date? Suggest an update →
Alternatives to Adept
More in Agents & Automation
