Every technological revolution has a decisive moment when specific use cases drive widespread adoption of the technology. With the rapid spread of the co-pilot, the generated AI (Genai) had already arrived at that time.
Genai has made great progress as a technology in the past few years. Despite all the headlines and hype, the company’s adoption is still in its early stages. The 2024 Gartner CIO and Technology Execution Survey will adopt only 9% of the respondents, with 34% saying they plan to do so next year. A recent survey by the Corporate Strategy Group put Genai adoption at 30%. However, the investigations all came to the same conclusions around 2025.
Prediction 1. Most businesses will use Genai in production by the end of 2025
Genai adoption is considered critical to improving productivity and profitability and has become a top priority for most businesses. But this means that the company must overcome the challenges faced by the Genaii project so far, including:
- Poor data quality: Genai ends up being as good as the data it uses, and many companies still don’t trust their data. Data quality, as well as incomplete or biased data, are all problems that lead to poor results.
- Genai Cost: Genai models like Chatgpt are mainly the best among the best Genai teams, while the cost of computing power is millions. Therefore, a technology called retrieval enhanced power generation (RAG) has been used. But even with a rag, the experts needed to access and prepare data and assemble successfully quickly became expensive.
- Limited Skills: Many early Genai deployments required a large number of coding from Genai’s experts. Although this group is growing, there is still a real shortage.
- Hallucination: Genai is not perfect. It can hallucinate and give the wrong answer when it thinks it is right. You need a strategy to prevent wrong answers from affecting your business.
- Data Security: Genai exposes data to the wrong person because it is used for training, fine-tuning or rags. You need to implement security measures to prevent these leaks.
Fortunately, the software industry has been dealing with these challenges over the past few years. 2025 looks like a year when several challenges begin to be solved, Genai is becoming mainstream.
Prediction 2. Modular rag co-pilot will be the most common use of Genai
The most common use of Genai is to create assistants or co-pilots to help people find information faster. The co-pilot is usually built using rag pipes. A rag is one way. This is the most common way to use Genai. Since large language models (LLM) are common models without all or even the latest data, you need to enhance the query (otherwise called a hint) for a more accurate answer.
The co-pilot helps knowledge workers to increase productivity, solve previously unanswered questions, and provide expert guidance while sometimes performing routine tasks. Perhaps the most successful Copilot use case to date is how to help software developers code or modernize legacy code.
However, when using the external, the co-pilot is expected to have a greater impact. Examples include:
- In customer service, co-pilots can receive support queries and upgrade to human intervention or provide resolution for simple queries such as password reset or account access to obtain higher CSAT scores.
- In manufacturing, co-pilots can help technicians diagnose and recommend specific operations or repairs for complex machinery to reduce downtime.
- In the healthcare field, clinicians can use co-pilots to obtain patient history and related research and help guide diagnosis and clinical care, thereby improving efficiency and clinical outcomes.
The rag pipes mostly work the same way. The first step is to load the knowledge base into the vector database. Whenever a person asks a question, the Genai Rag pipeline is called. It uses the retrieved information as context as context, evaluates and formats the results and displays them to the user, redesigns the problem as prompt, querys the vector database to find the vector database. .
But it turns out that you can’t support all co-pilots well with a single rag pipe. As a result, rags have evolved into a more modular architecture called modular rags where you can use different modules for each step involved:
- Index includes data blocks and organization
- Include query (hint) engineering and optimization before returning
- Retrieval of searcher fine-tuning and other techniques
- Re-evaluation and selection after return
- Use generator for fine-tuning, use and compare multiple LLMs and verify
- Manage orchestration of this process and iteratively help get the best results
You will need to implement a modular rag architecture to support multiple co-pilots.
Prediction 3. No code/low code genai tools will be the way
By now, you may realize that Genai Rag is very complex and rapidly changing. It’s not just new best practices that keep coming. All the technologies involved in the Genai pipeline are changing so quickly that you end up with swapping some or supporting several of them. Plus, Genai is more than just a modular rag. Retrieval enhanced fine tuning (RAFT) and complete model training also become cost-effective. Your architecture needs to support all these changes and hide complexity from engineers.
Thankfully, the best Genai codeless/low code tools offer this architecture. They continuously add support for leading data sources, vector databases, and LLMS and enable fine-tuning or training to be made by modular rags or built into LLMS. The company is successfully using these tools to deploy co-pilots using its internal resources.
Nexla not only uses Genai to make integration easier. It includes a modular rag pipe architecture with advanced data blocks, query engineering, recalculation and selection, multi-LLM support with result ranking and selection, orchestration, and all of which are not encoded.
Prediction 4. The line between co-pilot and agent will blur
Genai Codilots like chatbots are agents that support people. Ultimately, people decide how to deal with the resulting results. But the Genai agent can fully automate the response without involving people. These are often called proxy or proxy AI.
Some people think these are two separate methods. But the reality is more complicated. The co-pilot has begun to automate some basic tasks, optionally allowing the user to confirm the operations and automatically complete the steps required to complete them.
It is expected that the co-pilot will develop into a combination of co-pilot and agent over time. Just as applications help redesign and simplify business processes, assistants can and should start using intermediate steps to automate tasks they support. Genai-based agents should also include personnel handling exceptions or approving plans generated using LLM.
Prediction 5. Genai will drive adoption of data fabrics, data products and open data standards
Genai is expected to be the driver of the biggest change in the coming years as it needs to adapt to enable the company to realize the full benefit of Genai.
As part of the Gartner Hype cycle for data management, Gartner has identified 3 in 2024, and only 3 technologies are transformations in data management, as well as organizations that rely on data: data fabrics, data products and open-table formats. All 3 help make data easier to use with Genai as they make data easier to use with these new Genai tools.
Therefore, NEXLA implements a data product architecture based on data structures. Data structures provide a unified layer to manage all data regardless of format, speed or differences in access protocols. Data products are then created to support specific data needs, such as rags.
For example, a large financial services company is implementing Genai to enhance risk management. They use Nexla to create a unified data structure. Nexla automatically detects the mode and then generates connectors and data products. The company then defines data products for specific risk metrics that aggregate, clean and convert data into the correct format as input to implement rag agents for dynamic regulatory reporting. Nexla provides data governance controls including data lineage and access controls to ensure regulatory compliance. Our integrated platform for analytics, operations, B2B and Genai is implemented on a data fabric architecture that uses Genai to create reusable connectors, data products and workflows. Support for open data standards such as Apache Iceberg, making it easier to access more and more data.
How to plagiarize your own way of proxy AI
So, how should you prepare to mainstream Genai in your company based on these predictions?
First, if you haven’t already, start the first Genai Rag Assistant for clients or employees. Identify an important and relatively simple use case that you already have the right knowledge base to succeed.
Second, make sure you have a small team of Genai experts who can use the right integration tools to help you with the right modular rag architecture to support your first project. Don’t be afraid to use codeless/low code tools to evaluate new vendors.
Third, start identifying data management best practices that you will need successfully. This involves not only data structures and concepts like data products. You also need to manage your AI.
Now is the time. 2025 is a year for most people to succeed. Don’t be left behind.
