In the race to bring artificial intelligence to the enterprise, a small but well-funded startup is making a bold statement: The problem holding back AI adoption in complex industries has never been the models themselves.
Contextual AIa company created two and a half years ago, supported by investors including Bezos Expeditions And Bain Capital Venturesrevealed Monday Agent Composera platform designed to help engineers in aerospace, semiconductor manufacturing, and other technically demanding fields create AI agents capable of automating the type of knowledge-intensive work that has long resisted automation.
This announcement comes at a pivotal time for enterprise AI. Four years after ChatGPT sparked a frenzy of enterprise AI initiatives, many organizations remain stuck in pilot programs, struggling to move experimental projects into full-scale production. CFOs and business unit leaders are growing increasingly impatient with internal efforts that have consumed millions of dollars but yielded only limited results.
Dear Keela,CEO of Contextual AI, believes the industry has focused on the wrong bottleneck. "The model is almost trivialized at this stage," Kiela said in an interview with VentureBeat. "The bottleneck is context: can AI actually access your proprietary documents, specifications, and institutional knowledge? This is the problem we are solving."
Why enterprise AI continues to fail and what fetch-augmented generation was supposed to solve
To understand what Contextual AI tries, this helps to understand a concept that has become central in the development of modern AI: generation augmented by recovery, or CLOTH.
When major linguistic models like those of OpenAI, GoogleOr Anthropic generate answers, they rely on the knowledge acquired during training. But this knowledge has a deadline and cannot include the proprietary documents, technical specifications and institutional knowledge that are the lifeblood of most businesses.
RAG systems attempt to solve this problem by retrieving relevant documents from the company’s own databases and passing them to the model along with the user’s question. The model can then base its response on actual business data rather than relying solely on its training.
Kiela helped pioneer this approach while he was a research scientist at Facebook AI Research and later as director of research at Cuddly facethe influential open source AI company. He holds a Ph.D. from Cambridge and is an assistant professor of symbolic systems at Stanford University.
But the first RAG systems, Kiela acknowledges, were rudimentary.
"The first RAGs were quite rudimentary: take a commercial retriever, connect it to a generator, hope for the best," he said. "Errors got worse throughout the pipeline. Hallucinations were common because the generator was not trained to stay on the ground."
When Kiela founded Contextual AI in June 2023, it set out to systematically address these issues. The company has developed what it calls a "unified context layer" — a set of tools placed between a company’s data and its AI models, ensuring that the right information reaches the model in the right format and at the right time.
This approach is recognized. According to a Google Cloud case study, contextual AI has reached the highest performance on Google’s FACTS benchmark for grounded and hallucination-resistant results. The company refined Meta’s open source Llama models on Google Cloud’s Vertex AI platform, focusing specifically on reducing the tendency of AI systems to invent information.
Inside Agent Composer, the platform that promises to turn complex engineering workflows into minutes of work
Agent Composer extends Contextual AI’s existing platform with orchestration capabilities – the ability to coordinate multiple AI tools in multiple steps to achieve complex workflows.
The platform offers three ways to create AI agents. Users can get started with pre-built agents designed for common technical workflows such as root cause analysis or compliance verification. They can describe a workflow in natural language and let the system automatically generate a working agent architecture. They can also create from scratch using a visual drag-and-drop interface that requires no coding.
According to the company, what sets Agent Composer apart from competing approaches is its hybrid architecture. Teams can combine strict, deterministic rules for high-stakes steps (compliance checks, data validation, approval gates) with dynamic reasoning for exploratory analysis.
"For highly critical workflows, users can choose fully deterministic steps to control agent behavior and avoid uncertainty." » said Kiela.
The platform also includes what the company calls "One-Click Agent Optimization," which takes user feedback into account and automatically adjusts agent performance. Every step of an agent’s reasoning process can be audited, and responses are accompanied by sentence-level citations indicating exactly where the information originated in the source documents.
From eight hours to 20 minutes: what the first customers are saying about the real performance of the platform
Contextual AI claims that early customers have reported significant efficiency gains, although the company acknowledges that these figures come from customer self-reporting rather than independent verification.
"These come directly from customer reviews, which are approximations of actual workflows," » said Kiela. "The numbers are self-reported by our customers as they depict the scenario before and after the adoption of contextual AI."
The announced results are nevertheless striking. An advanced manufacturer reduced root cause analysis from eight hours to 20 minutes by automating sensor data analysis and log correlation. A specialty chemicals company reduced product research from hours to minutes with agents searching patents and regulatory databases. A test equipment manufacturer now generates test code in minutes instead of days.
Keith Schaub, vice president of technology and strategy at Avantesta semiconductor test equipment company, offered its endorsement. "Contextual AI has played an important role in our AI transformation efforts," » Schaub said. "The technology has been deployed to several Advantest teams and selected end customers, saving significant time on tasks ranging from test code generation to customer engineering workflows."
Other clients of the company include Qualcommthe semiconductor giant; ShipBoba tech-enabled logistics provider that claims to have solved problems 60 times faster; And Nvidiathe chipmaker whose graphics processors power most AI systems.
The eternal business dilemma: should companies build their own AI systems or buy off-the-shelf?
Perhaps the biggest challenge Contextual AI It is not competing products that compete, but the instinct of engineering organizations to create their own solutions.
"The biggest objection is “we’ll build it ourselves”." Kiela admitted. "Some teams try. This sounds exciting to achieve, but it is exceptionally difficult to do well at scale. Many of our customers started out DIY and found themselves debugging recovery pipelines instead of fixing the actual issues 12-18 months later."
The alternative – off-the-shelf solutions – presents its own problems, the company says. Such tools deploy quickly but often prove inflexible and difficult to customize for specific use cases.
Agent Composer attempts to occupy a middle ground, offering a platform approach combining pre-built components with extensive customization options. The system supports models from OpenAI, Anthropic, and Google, as well as Contextual AI’s own Grounded Language Model, which has been specifically trained to stay true to retrieved content.
Pricing starts at $50 per month for self-service use, with custom enterprise pricing for larger deployments.
"The rationale for CFOs is actually to increase productivity and get them to produce faster through their AI initiatives." » said Kiela. "Every tech team struggles to recruit the best engineers, so making existing teams more productive is a top priority in these industries."
The Road Ahead: Multi-Agent Coordination, Write Actions, and the Race to Create Compound AI Systems
Looking ahead, Kiela outlined three priorities for the coming year: automating workflows with actual write actions to company systems rather than just reading and parsing; better coordination between several specialized agents working together; and faster specialization through machine learning from production feedback.
"The compound effect matters here," he said. "Every document you ingest, every feedback loop you close, those improvements add up. Companies building this infrastructure will now be difficult to catch."
The enterprise AI market remains fiercely competitive, with offerings from major cloud providers, established software companies and many startups seeking the same customers. Whether AI’s bet on context rather than models succeeds will depend on whether companies come to share Kiela’s view that basic model wars matter less than the infrastructure around them.
But there is a certain irony in the company’s positioning. For years, the AI industry has focused on building ever-bigger, ever-more powerful models, pumping billions into the race for artificial general intelligence. Contextual AI makes a quieter argument: For most real-world work, the magic is not in the model. It’s about knowing where to look.
