Paul Lewis, Chief Technology Officer

Enterprises are investing heavily in artificial intelligence, launching pilots and proofs of concept across functions, but outcomes remain limited. Nearly 90 percent of AI initiatives never make it to production, leaving organizations with isolated experiments instead of real business impact.

Pythian addresses this disconnect by helping organizations transition from early-stage exploration to production-ready systems that deliver measurable outcomes in real-world environments.

“The real challenge isn’t just getting AI into production, but knowing how to support and evolve it once it’s in production,” says Paul Lewis, chief technology officer.

As a data, analytics and AI services company, Pythian brings over three decades of experience managing 30,000 databases and 2,000 environments for 1000+ customers. This background in running mission-critical systems informs how it approaches AI in production environments.

How do operational support models improve long-term performance of enterprise artificial intelligence systems?

Where AI Initiatives Break Down

Insights from Pythian’s engagements point to two recurring barriers. The first is poor use case selection. Many organizations either focus on low-impact improvements that do not justify investment or highly complex transformations that are difficult to execute with current tools and skills. The most effective initiatives lie in a middle ground where impact is visible and execution is feasible.

The second issue emerges after deployment. Once systems go live, accuracy must improve over time, models must evolve with new data and performance must remain stable under changing conditions. Organizations often lack the specialized skills to continuously monitor and adapt these systems in production.

The real challenge isn’t just getting AI into production, but knowing how to support and evolve it once it’s in production.

Execution is organized through a dual center of excellence structure. One stream focuses on end-user enablement, where Pythian develops no-code AI agents and trains employees to use them. The second stream, an agentic center of excellence, acts as the AI build team responsible for developing complex, multi-data-source agents and workflow-based agentic systems, and implementing them in production.

Supporting both streams is a dedicated AI operations layer that manages these systems once they go live. This includes monitoring model performance, improving accuracy, handling updates, and managing the overall lifecycle of AI applications.

Alongside this, Pythian’s managed services provide the operational backbone required to run these environments at scale. It delivers 24/7 monitoring, incident management and performance optimization. Given the volume of activity, much of this support is AI-assisted, where systems analyze past incidents, identify likely resolutions and in some cases automate fixes. The same approach extends to data pipelines and AI workloads, ensuring consistent security and performance across conventional IT systems and newer AI-driven environments.

Why is measurable business impact important when deploying enterprise artificial intelligence solutions at scale?

Delivering Measurable Impact

The impact of Pythian’s capability is evident in practice. In one engagement with a logistics provider, it addressed a manual document processing workflow that was slowing down operations. Applying computer vision to extract and validate information, it reduced processing time from two hours to ten minutes. The solution delivered immediate returns and scaled into millions of dollars in long-term savings.

An unintended value was accuracy. The accuracy of manual double-key entry was in the high 80 percent range, while the accuracy achieved through computer vision with a tuned model and out-of-the-box Gemini reached the mid 90 percent range, improving the overall outcome quality.

To what extent are enterprises integrating artificial intelligence into everyday operational system management today?

These outcomes are driven by a deliberately high-caliber team built for complex problem solving. With industry-recognized experts, including Oracle ACEs and Microsoft MVPs, Pythian focuses on challenges that require depth rather than scale.

As AI capabilities become embedded across enterprise systems, from databases to analytics platforms, organizations will need to manage these capabilities as part of everyday operations. Pythian’s approach reflects this shift. By turning AI into an operational discipline rather than a one-time initiative, it supports enterprises in sustaining accuracy and delivering measurable business outcomes at scale.