Pathway to Production: Implementing GenAI in Your Organization in 2025
At Portal26, we’ve been at the forefront of the GenAI revolution, witnessing firsthand the excitement and challenges organizations face as they navigate this transformative technology. Through countless conversations with customers eager to harness the power of GenAI, we’ve heard a consistent refrain: while the potential is undeniable, the path to successful adoption is fraught with obstacles.
Time and again, we’ve listened to leaders grapple with fundamental questions: How do we measure ROI from our GenAI investments? What are the best practices for implementation? How do we balance innovation with security and compliance? These conversations have revealed a critical gap between GenAI’s promise and its practical deployment in enterprise environments.
Drawing on our deep expertise in GenAI governance and security, and in partnership with AWS, we’ve developed this comprehensive whitepaper to address these very challenges. Our insights are further reinforced by cutting-edge research from our colleagues at MIT and McKinsey, whose studies validate the widespread nature of the obstacles organizations are encountering.
This guide represents our commitment to helping you navigate the complex landscape of GenAI adoption, transforming uncertainty into actionable strategy, and pilots into production-ready solutions.
The Importance of GenAI
Key concerns:
- Security & risk: 85% of C-suite executives cite privacy/security concerns, and 88% recognize the risks of shadow AI—unauthorized tools used by employees
- Governance & visibility: 73% report issues with security or data misuse, while 21% regret not having better governance in place
- Adversarial threats: 55% of CTOs believe GenAI benefits attackers more than defenders, citing deepfakes, phishing, and malware concerns (source)
Pathway to Production: Key Challenges
Unlike traditional ML models, GenAI projects introduce unique obstacles:
Black-box complexity:
SaaS and foundational models function as opaque systems with unpredictable outputs.
Model selection uncertainty:
With 25+ credible foundational models and evolving guardrails, organizations struggle to make optimal choices.
Deterministic vs. stochastic expectations:
Traditional software delivers consistent results; GenAI responses can vary significantly, challenging enterprise expectations.
Use case alignment:
Applications where variation is acceptable thrive; deterministic use cases may require fine-tuned models or alternative approaches.
What Is “Production”?
Production AI deployment varies in complexity:
- Plug-and-play AI tools (e.g., GitHub Copilot): Quick to deploy, minimal risk.
- Enterprise-integrated AI solutions: Custom-built GenAI tools require data integration, monitoring, and compliance guardrails.
The latter poses greater challenges—many projects remain in PoC or testing due to unpredictable model behavior, governance gaps, and evolving AI infrastructure.
Pathway to Production: Use Case Selection
A critical first step in projects that successfully go-live is use case (i.e. a problem to solve) identification. Identifying the right use cases to invest your time and effort is the most important step in going live—like laying a solid foundation, it grounds your genAI initiatives on real data.
Many organizations identify use cases through surveys (i.e. asking employees how they’re using AI) or bottoms-up brainstorming. The best way to find good use cases is to understand all existing usage of GenAI within the enterprise. If you allow your employees to access public GenAI responsibly, they will guide you. Classify and categorize the use cases
to understand value and risk.
How Portal26 can help
Shadow AI discovery:
In-depth visibility into who is using GenAI, what tools are being used, and what the tools are being used for
Usage & risk analytics:
Track the impact of GenAI—both its value and the risk it introduces. As many as 85% of C-suite tech executives have concerns about GenAI privacy and security, and 88% feel shadow AI is a high risk (source).
Cost insights:
Portal26 offers great insights into both risk and costs (token counts that are input into cost calculators).
Single control plane:
Monitor GenAI across all platforms—Public LLMs and Internal Models on AWS Bedrock, Google Vertex, and beyond.
Pathway to Production: Going Live
Regardless of how well the project tested out during development phase, as soon as you place a model under production, you have to know two things:
1. If the model is being subject to unusual activity, including:
- Major variation in real-life usage patterns from test patterns —e.g. your project scope didn’t anticipate the actual prompts that are being posed to the model.
• Harmful prompts that may indicate jailbreak attempts,
DAN (Do Anything Now) attacks, data poisoning, etc.
2. If the model response isn’t in line with expectations, it could
be due to the above variation between testing and real life usage, or hallucination or other inherent limitations in the model.
To successfully go live, you will need to systematically collect requests and responses from your production models and subject them to analysis so you can identify and address concerns.
Portal26, a participant in the AWS Global Startup Program, can help you get to both prompts and responses for your internally deployed models—be it on AWS Bedrock or other platforms. It acts as a single plane of glass for all your models, externally, internally, and across all your model deployment platforms
Case Study
Here’s what production looked like for one Portal26 client:
The company deployed the Portal26 GenAI Adoption Platform in a half-day with full
DLP integration. Within days, they began a phased rollout of GenAI permissions across
the company. Within weeks of opening the door, the Portal26 dashboards provided
enough data for the company to understand their risks and define policy. With policy
distribution and education being available out-of-the-box, the company was up and
running with all the required capabilities in record time.
Getting Started with Portal26
Portal26 essentially has a simple four steps to implementation plan:
Adopt GenAI Fearlessly With Portal26
The path to production begins with understanding real-world GenAI use cases, mitigating risks, and establishing governance. With Portal26, organizations gain the visibility, security, and control needed to harness GenAI effectively in 2025 and beyond.
