New platform capabilities allow businesses to build, own, and continually improve their own AI models rather than renting intelligence from external APIs.
Empromptu AI, the company leading enterprises through the transition from static SaaS to self-improving AI-native applications, today announced Alchemy Models, a new feature that enables enterprises to create, train, and deploy their own production AI models by building on the internal and external AI applications they already create, without requiring model training expertise. This release addresses a growing gap in the current AI tools ecosystem. Companies share their most valuable data with large model providers to compete in the agent market. Companies are already hiring the subject matter expertise needed to train models, but they lack the AI talent to capture it. To avoid interruptions by model providers, subject matter expertise is key to creating an effective data mote. Alchemy bridges that gap by adding custom model ownership on top of your application stack.
Companies spend billions of dollars hiring subject matter experts to document the workflows for selling training data back to model providers. Companies already have the subject matter experts they need. What they lack is the infrastructure and AI expertise to capture it and use it effectively. Alchemy Models encodes that expertise into an easy-to-use infrastructure.
“Most companies today rent intelligence,” said Shanea Leven, CEO of Empromptu. “They send their own data to other people’s models and hope that economic conditions and policies remain good. Alchemy gives them another option. They can build and own the intelligence behind their products. The provider of the model is Amazon, and in this scenario, the rest of us are willfully Toys R Us. Except now, we know exactly what’s going on.”
Also read: AiThority Interview with Glenn Jocher, Ultralytics Founder and CEO
Empromptu’s platform takes a step-by-step approach to custom model development. Companies start by building AI-driven applications using natural language interfaces. This automatically collects high-quality training data through real-world usage. Once the workflow produces output, subject matter experts label edge cases to validate the results and create accurate training data for fine-tuning. This removes traditional barriers to training custom models. This means no need for manually curated datasets or armies of data annotators. Instead, companies leverage existing workflows to continuously generate training data.
Alchemy simplifies tasks that previously required a full machine learning team. Users define their tasks through natural language or the Empromptu builder interface, and the platform automatically handles the rest:
- Generating synthetic and/or real data using Empromptu’s Golden Data Pipelines
- Select and prepare training datasets based on model performance
- Enable subject matter experts to score and modify output from real product workflows
- Evaluate model output using automated evaluation framework
- Fine-tune the base model for a specific task or application
- Deploying Empromptu in the cloud or on customer’s own infrastructure
- Continuously improve models using agent-driven training loops
The result is a production-ready model that learns from real-world usage and improves over time. No machine learning expertise required.
As AI adoption expands, companies face three major challenges. Many AI coding tools generate working code, but lack the infrastructure needed to run AI in production. Without structured data pipelines, evaluation frameworks, and governance controls, applications that work in demos often fail when real users or messy data enter the system. Additionally, most AI applications today are completely dependent on external model providers, raising concerns about data control, vendor lock-in, and long-term cost burden. Additionally, API-based models can become expensive as usage increases. Fine-tuned models that are optimized for specific tasks can often achieve higher accuracy at significantly lower cost.
Early corporate adopters are already seeing measurable results. In internal benchmarks, custom models built using Alchemy reduced inference costs by 40-80% and increased accuracy rates by 25-30%.
Ascent Health increased the accuracy rate of the learning application by 30% on the first run.
Organizations across financial services, healthcare, legal technology, retail, and more use Alchemy to build industry-specific models and train them on their own datasets for risk analysis, compliance monitoring, diagnostics, contract review, and demand forecasting.
Many companies remain cautious about implementing AI due to a lack of governance processes or concerns about exposing their proprietary data to external providers. Empromptu was designed to directly address these concerns. The platform includes governance policies, audit logs, environmental controls, evaluation pipelines, model drift monitoring, and rollback paths to enable enterprises to securely deploy AI systems within regulated environments.
“Right now, a lot of companies are saying, ‘I don’t need AI,’ because they don’t know how to control it,” Leven said. “The moment you can run AI on your own infrastructure, with your own data and governance policies, the conversation completely changes.”
The release of Alchemy follows Empromptu’s recent platform enhancements that introduced Golden Pipelines and AI Policies to bring data preparation and governance directly into the AI application development process. Together, these capabilities enable the platform to extend across the entire lifecycle of an AI system, from data preparation and building applications to applying controls and training models, all within a single environment.
Alchemy is immediately available to enterprise customers using the Empromptu platform. Organizations interested in early access can sign up at empromptu.ai.
Also read: The infrastructure war behind the AI boom
[To share your insights with us, please write to psen@itechseries.com ]
