The 2025 GenAI SitRep

 

As 2026 announces itself as the year where “the hombres will be separated from the niños” , I’d like to look back at the evolution in GenAI and present a few predictions on it for the next two to three years.

When ChatGPT was launched in November 2022; it immediately raised high hopes for a new wave in machine learning, making the interaction with neural networks more intuitive using natural language. And as time progressed, the models were trained on ever increasing data volumes as the incumbents saw this as the only way to win the platform war and telling their investors about first mover advantage to get funds at surreal valuations. Now they are finding out two things. One: size is not the distinguishing factor in the platform competition. It’s about purpose built and well curated models where the competition is heading and two: there is no platform competition because the platforms were already there: Microsoft integrating ChatGPT in Office365 and Google doing the same with Gemini in its product suite.

Other platforms like Meta, shopping platforms like Amazon, Ali Baba or banking applications, either promote home grown LLMs or use one or more models among the hundreds of thousands open access models they can find on Hugging Face.

Nevertheless, there is one aspect of platform economics emerging, the creation of exit barriers. Once you’ve chosen for a platform to integrate in your business applications the SDKs and the models used and improved by RAG and other means (which we are working on) will increase switching costs.

The first three years in GenAI were a lot about overpromising and under delivering. I am not checking if Gartner already talks about the trough of disillusionment but I have met a few clients where disillusion has set in. Cases of organisations reducing their after sales service staff and replacing them with chatbots had to reverse their decision. We all know of lawyers producing phony legal precedents are being fined.  Not langer than a couple of weeks ago, a judge in California has fined plaintiffs’ law firm Hagens Berman, one of its partners, and another lawyer a combined $13,000 for the “misuse” of artificial intelligence in several court filings in a lawsuit against the parent company of adult content social media site OnlyFans.

Much of the GenAI offers are “inside out” like when a top official of ChatGPT bluntly stated that the users will need to learn what they can do with their product. To use Lee James and Warren Schirtzinger’s concept of the marketing chasm, later popularised by Geoffrey Moore’s terminology, we are very much pleasing the techies but the pragmatists are left in the cold and are waiting on the other side of the chasm for a useful application to cater to their needs.  That doesn’t absolve the pragmatists from the obligation to examine the impact of GenAI on their organisation, their business processes and their talent management. But today, little efforts are made to support them in that endeavour.

Forget about network effects, it’s all about trusted data

The real battle will be fought on the data terrain. And this is where our knowledge modelling methodology comes in play. Its output combines the strengths of “older” machine learning techniques with state of the art LLM technologies and prepares domain knowledge for reliable use in a well-defined, professional context.

It is also about scope management. There is a simple, over and over empirically conformed law that illustrates the correlation between consistency and scope of information: the smaller the scope, the higher the consistency.

Let me give you two examples to illustrate this thesis.

In the accounting domain the concept of cash flow is unequivocally defined as the money that flows in and out of a business and is measured from different perspectives: operations, investing and financing. Cash flow is an important metric to determine the value of a business.  But a business valuation in itself is a more fuzzy concept composed of various, sometimes contradicting metrics as not only quantitative metrics like free cash flow, debt to equity ratio, price/earnings ratio etc… are under scrutiny. There are a lot of qualitative or more fuzzy metrics and evaluations like risk, customer loyalty, market position,  growth potential and brand preference that are equally important to this concept.

In the supply chain domain, Solventure’s Bram De Smet pointed out in his book “SupplyChain Strategy and Financial Metrics” that crisply defined metrics like cash, cost and service level become a dynamic and fluid mix as a function of your business strategy. Cash, cost and service level may be opposing forces in realising an optimum outcome to support the strategic direction. The author leverages Treacy & Wiersema’s model of value disciplines, showing how product leadership, customer intimacy, or operational excellence each imply different preferred positions on the service–cost–cash triangle.

Crawford & Matthews’ five value drivers (price, access, service, product, experience) are used to further link market strategy choices to supply chain and financial design.

What we need in GenAI is a method, models and software that support both the low level, crisp and well defined knowledge building blocks while producing meaningful concepts that combine these building blocks in a reliable way. Imagine on top of that, we can implement de Bono’s lateral thinking as a proxy for creativity…

The S-curve, where are we in GenAI?

GenAI technology started out as an expensive platform war with ever increasing investments in GPUs, data volumes and training efforts. But contrary to the classic S-curve evolution, it was rapidly widely adopted as the  major players were seeding the market via free, universally available chatbots in a browser or an app, Microsoft making a move with CoPilot and Google introducing Gemini in its Google Search engine. True to the S-curve’s pattern, improvement is slow as the fundamental concepts are being figured out. This is where we are today.

When the period of rapid innovation and massive adoption will follow is hard to predict but there are signs it won’t take too long. I keep seeing more and more colleagues investigating real life use cases instead of considering it as a glorified search engine or an evolved autocomplete. I think before 2028 we will see some real killer apps like streaming music and video, E-Commerce and social media were for the Internet era. Instead of a top down movement, fuelled by massive investments and even “incestuous” investments compared to Lernout and Hauspie’s bloating its revenue figures during the Internet bubble, we will see bottom up initiatives on these platforms delivering measurable value.

We hope the application we are working on, will one day be part of GenAI’s killer apps.

 

Yogi Berra on GenAI

The inimitable Yogi Berra already knew it: “Predictions are very hard, especially about the future.” So please accept my apologies in advance, as my prediction will likely be inaccurate in terms of both timing and scope when it comes to the transition we are about to witness in the field of Generative AI and its associated market.

The launch of ChatGPT-5 has demonstrated that “parameter escalation” is subject to the law of diminishing returns. Greenfield startups such as OpenAI are burning cash at an astonishing rate, and at some point a shake-out will inevitably occur. The giants—OpenAI (backed and integrated by Microsoft) and Gemini (from Google)—are well positioned to survive. Their model performance is increasingly shaped not merely by parameter counts, but also by architecture, training data quality, deployment efficiency, and—crucially—the fact that their AI functions are embedded into widely used software ecosystems, creating high entry barriers for competitors. What will become of the many other models hosted on Hugging Face remains an open question.

LLM users are already aware that parameter escalation alone cannot eliminate the remaining 1.5–2% hallucination rate. They are also keenly aware that their interactions contribute intellectual property to the models, often without compensation. Furthermore, they know that open models are vulnerable to prompt injections, nonsensical outputs, and coordinated reputation attacks by adversaries.

There is, therefore, a market for closed models built on validated data curated by experts. Beyond pattern recognition, such systems will deliver genuine problem-solving capabilities. Subject matter experts will play a central role in improving data quality—by uploading validated content, stress-testing outputs with thousands of questions, and leveraging Retrieval-Augmented Generation (RAG) to enhance reliability. The logical next step will be to integrate rule-based algorithms for decision support—bringing us full circle to the earliest AI systems of the 1960s, such as Mycin, a pioneering pharmaceutical knowledge base.

Ultimately, these systems will evolve toward mimicking human reasoning and judgment through first-principles thinking.

Beneath the surface of press releases aimed at inflating incumbents’ P/E ratios or maximizing the IPO valuations of new entrants, something more substantial is brewing. 

 My bold prediction is this: in the long run, closed models will generate more value than today’s “gorillas,” who are largely providing the infrastructure for them. 

And so, one day, another of Yogi Berra’s paradoxical dicta may well come true: 

“Nobody goes there anymore. It’s too crowded.”

Data Governance for GenAI

Introduction

In this article I will define what data governance (DG) for Generative Artificial Intelligence (GenAI) is and how it differs from DG as we have known it for decades in the world of transaction systems (OLTP) and analytical systems (OLAP and Data Mining).

In a second post, I will make the case for DG based on the use case at hand and illustrate a few GenAI DG use cases that are feasible and fitting the patterns and the framework.

Die “Umwertung aller Werten”

The German philosopher Friedrich Nietzsche postulated that all existing values should be discarded and replaced by values that -up to now- were considered unwanted. 

This is what comes to mind when I examine some GenAI use cases and look at the widely accepted data governance policies, rules and practices.

Here are the old values that will be replaced:

• Establish data standards;

• The data model as a contract;

• Data glossary and data lineage, the universal truths;

• Data quality, data consistency and data security enforcement;

• Data stewardship based on a subject area.

Establish data standards

As the DAMA DM BOK states: Data standards and guidelines include naming standards, requirement specification standards, data modelling standards, database design standards, architecture standards, and procedural standards for each data management function. 

This approach couldn’t be further away from DG for GenAI. Data standards are mostly about “spelling” which has very low impact on semantics. The syntactical aspects of data standards are more in the realm of tagging where subject matter experts provide standardised meaning to various syntactical expressions. So we can have tagging standards for supervised learning, but even those can depend on “the eye of the beholder”, i.e. the use case. 

OK, we can have discussions about which language model and vector database is the best fit for the use case at hand but it will be a continued trial and error process before we have optimised the infrastructure and it certainly won’t be a general recommendation for all use cases.

And as for the requirement specification standards, as long as they don’t kill the creativity needed to deal with GenAI, I’ll give them a pass, since this is not always a linear process to identify business needs for information and data. The greatest value in GenAI lies in discovering answers to questions you didn’t dream of asking. 

The data model as a contract

fig. 1: Requirements constitute a contract for the data model. Governing this contract is relatively easy.

This principle works fine for transaction systems and classic Business Intelligence data architectures where a star schema or a data vault models the world view of the stakeholders. The only contract is the aforementioned tagging and metadata specifications to make sure the data are exploitable.

Data glossary and data lineage, universal truths?

No longer. The use case context will determine the glossary and the lineage if there are intermediate steps involved before the data are accessible. Definitions may change as a function of context as well as data transformations to prepare them for the task at hand. 

Data quality, data consistency and data security enforcement

In old school data governance policies, data quality (DQ) is first about complying with specs and only then does “fit for purpose” comes in as the deciding criterion as I described in Business Analysis for Business Intelligence(1):

Data quality for BI purpose is defined and gauged with reference to fitness for purpose as defined by the analytical use of the data and complying with three levels of data quality as defined by:

[Level 1] database administrators

[Level 2] data warehouse architects

[Level 3] business intelligence analysts

On level 1, data quality is narrowed down to data integrity or the degree to which the attributes of an instance describe the instance accurately and whether the attributes are valid, i.e. comply with defined ranges or definitions managed by the business users. This definition remains very close to the transaction view.

On level 2, data quality is expressed as the percentage completeness and correctness of the analytical perspectives. In other words, to what degree is each dimension, each fact table complete enough to produce significant information for analytical purpose? Issues like sparsity and spreads in the data values are harder to tackle. Timeliness and consistency need to be controlled and managed on the data warehouse level. 

On level 3, data quality is the measure in which the available data are capable of adequately answering the business questions. Some use the criterion of accessibility with regards to the usability and clarity of the data.  Although this seems a somewhat vague definition, it is most relevant to anyone with some analytical mileage on his odometer. I remember a vast data mining project in a mail order company producing the following astonishing result: 99.9% of all dresses sold were bought by women!

In GenAI, we can pay few attention to the aforementioned level 1 while emphasizing the higher level aspects of data quality. And there, the true challenge lies in testing the validity of three interacting aspects of GenAI data: quality, quantity and density. As mentioned above: quality in the sense of “fit-for-use-case” reducing bias and detecting trustworthy sources, quantity by guaranteeing sufficient data to include all -expected and non-expected- patterns and finally density: to make sure the language model can deliver meaningful proximity measures between the concepts in the data set.

Data stewardship based on a subject area

fig. 2:Like a football steward, a data steward must also control the crowd to prevent chaos

A business data steward, according to DAMA  is  a knowledge worker and business leader recognized as a subject matter expert who is assigned accountability for the data specifications and data quality of specifically assigned business entities, subject areas or databases, who will: (…)

4. Ensure the validity and relevance of assigned data model subject areas

5. Define and maintain data quality requirements and business rules for assigned data attributes.

It is clear that this definition needs adjustments. Here is my concept of a data steward for GenAI data:

It is, of course, a knowledge worker who is familiar with the use case that the GenAI data set needs to satisfy. This may be a single subject matter expert (SME) but in the majority of the cases he or she will be the coach and integrator of several SMEs to grasp the complexity of the data set under scrutiny. He or she will be more of a data quality gauge than a DQ prescriber and, together with the analysts and language model builders will take measures to enhance the quality of the output  rather than investing too much effort in the input. Let me explain this before any misconceptions occur. The SME asks the system questions to which he knows the answer, checks the output and uses RAG to improve the answer. If he detects a certain substandard conciseness in all the answers he may work on the chunking size of the input, but that is without changing the input itself. Meanwhile some developers are working on automated feedback learning loops that will improve the performance of the SME, as you can imagine coming up with all sorts of questions and evaluating the answers is a time consuming task.

In conclusion

Today, GenAI is more about enablement than control. It prioritises the creative use of data while ensuring ethical and transparent use of it. Especially in Explainable Artificial Intelligence (XAI) this approach is enabled in full. I refer to a previous blog post on this subject.

Since unstructured data like documents and images are in scope, a more flexible and adaptive metadata management is key. AI is now itself being used to monitor and implement data policies. Tools like Alation, Ataccama and Alex Solutions have done groundbreaking work in this area and Microsoft’s Purview is -as always- catching up. New challenges emerge: ensuring quality and accuracy is not always feasible with images and integrating data from diverse sources and in diverse unstructured formats is also a challenge. 

The more we are developing new use cases for GenAI, the more we experience a universal law: data governance for GenAI is use case based as we prove in the next blogpost. This begs the question for a flexible and adaptive governance framework that monitors, governs and enforces rules (but not too strict unless it’s about privacy) of its use. In other words, the same data set may be subject to various, clearly distinguishable governance frameworks, dictated by the use case. 

________________________________________________________________________________

(1) Business Analysis for Business Intelligence, CRC Press 2012, Bert Brijs p. 272