Enthalpy is an agentic AI framework & tool designed to accelarate AB testing for SaaS products (software of any kind really). The tool is surfaced as a webapp hosted over enthalpy.cc using off-the-shelf popular LLMs APIs for the underlying AI layer but also self-hosted LLMs specifically fine-tuned to work best with the product. The open-source components here however can be self-hosted & be used independently with any supported LLM provider.
It would be a truism at this point to say that the software industry is changing rapidly. With the advent of agents, established roles are changing as we speak. With agentic AI tools, companies will increasingly hire for and/or transition existing ones to super-ICPs who are responsible for end to end outcome-delivery for specific functional verticals, even merging multiple functional roles into one.
This tool is specifically for such a future where growth / product / marketing teams within SaaS orgs employ a few super-ICPs responsible for the end-to-end revenue outcomes with AB experimentation including designing, implementing & managing execution & rollout of experiments. Some of these responsibilities currently already lie within the charter of these roles but others, especially UI/UX design, analytics and implementation often still require collaboration with other specialised teams.
The tool is specifically for aiding:
- End-to-end design -- in the sense of idea/hypotheses generation, research, UI/UX design
- Implementation (in the sense code changes)
- Execution (in the sense of running, analysis, rollouts)
of AB experiments aimed at whatever objectives an org musters up.
In short, the idea is to allow this super-ICP to be able to perform the whole cycle with minimal involvement with other teams or personnel with all the tools to be able to do so available in a single package.
But that's not all. The tool also aims to achieve a varying degrees of autonomy in being able to deliver on these objectives end-to-end via iteratively autonomously launching AB experiments. This would either be autonomy for specific workflows in the process or for specific ownership criteria (touchpoints, metrics, objectives etc) similar to how a real growth manager would, all of this autonomy user-controlled & managed to the degree of their comfort of-course.