Data democratization is broadly described as capturing more data and making it more widely available to more people. It’s a straightforward definition; however, it’s odd that we aren’t questioning it more now that the sheer volume, variety, and velocity of data the world creates is almost incomprehensible.
-
- ❖ In 2020, the amount of data the world created exceeded the number of detectable stars in the universe: 64 zettabytes.
- ❖ By 2023, it had doubled to 120 zettabytes.
- ❖ It’s forecast to be at a mind-boggling 180 zettabytes by 2025.
While there are a number of platforms, dashboards, and data visualization tools that make data democratization partially possible, they still require managers to set up, maintain, and interpret the data, which often negates the reasons that data democratization was pursued in the first place.
In reality, a correctly-configured AI system needs to be in place before attempting to democratize data and is, in fact, the very best way to achieve it. AI technology can replace the goal of data democratization with an “information on demand” approach. Let’s explore how.
Defining the New Data Democracy – Information on Demand
The goal of capturing more data and making it more widely available to more people needs to be exchanged for a different goal — information on demand: providing the right information to the right people at the right time, making further exploration easy, facilitating quicker, smarter decisions, and sparking innovation.
We use the term ‘information’ rather than data and define the difference as information is data which is formatted, made relevant and consumable to a specific user, in other words information is aware of its audience, whereas data is simply an output from a system.
AI technology holds the promise of sorting through all the available data, turning it into relevant, presentable, consumable information, and providing it to the people who need it at that time.
As a discipline, data democratization isn’t inherently good or inherently bad. It’s a tactic that can yield either positive or negative results depending on how it’s implemented. If it’s implemented well, friction goes down and productivity goes up. If it’s implemented poorly, it becomes just one more thing to do and may replace systems that were working better before the initiative was launched.
An well-configured AI approach to information on demand provides many advantages and removes many of the disadvantages to traditional data democratization.
| AI-Enabled Information on Demand | Traditional Data Democracritization |
|---|---|
| ✓Right information | X Data you don’t need |
| ✓ Right people | X Data someone else needs |
| ✓ Right time | X Data before or after it’s needed |
| ✓ Trustworthy options for action | X Uncertainty |
| ✓ Better, faster decisions | X Status quo |
| ✓ Innovative action | X Missed opportunity |
| ✓ Consumable and Relevant | X Raw Data requiring interpretation |
If you’re like most people, you think first about the human resource side of the equation: skills, training, and temperament. But the most experienced, well-trained, and motivated people can only go so far with applications that are unable to achieve the objectives due to the application limitations or the time required to operate the application properly.
Barriers to Achieving Information on Demand with Agentic AI
While this situation appears to be perfectly suited for artificial intelligence to solve, none of the agentic AI platforms currently available seem to be up to the task. They rely on programming the AI to solve problems one at a time and “do not retain feedback, adapt to context, or improve over time”, according to a July 2025 report from the MIT Media Lab’s Project NANDA which found that 95% of AI business initiatives fail.
Agentic AI’s Achilles Heel
AI agents use LLMs to solve specific types of problems. AI Agents typically are AI wrappers calling one to two instances of a single LLM and rely on the LLM for both logic and comprehension of the task. They rely on RAG from conversation history summaries and unstructured inputs combining those with static prompts to compile a final AI input.
Agents sometimes send inputs and outputs to each other in a relatively unstructured manner forming a ‘‘swarm’. LLMs determine which functions or APIs to call when integrated with other systems, with little or no control over the inputs which get passed to the functions or APIs. Improvement of AI agents is typically done through static prompt engineering, LLM fine tuning, and adjusting LLM parameters. These improvements are typically made manually by AI engineers or, in the case of fine tuning, require a significant amount of time to generate new questions and answers for LLM training.
To reiterate, AI can solve the data democratization problem, but not in its present form. Current AI tools are like having an extremely motivated and hard-working research assistant that is also dumber than a bag of hammers. Until AI’s deficiencies have been overcome, its use will make information on demand worse instead of better. Two things are holding it back: public trust in its ability to provide the right information, and its own inability to understand and manipulate data across silos.
The Trust Issue
2025 AI has a trust issue. Any SMEs reading this can immediately think of multiple examples of misinformation they’ve encountered while trying to be more productive with LLMs. Hallucinations aside, would anyone today ever really trust AI to update their enterprise data? Even more to the point, would your stakeholders trust it to update their data and follow all of your company’s security and business rules?
The obvious answer is no. While AI can be trusted to give you different options to choose from, all of which will need to be fixed; it can’t be trusted to be 100% thorough or 100% accurate with those options — nevermind selecting the best one from among them. And it certainly isn’t ready to be responsible for data governance.
The Integration Issue
The foundational requirement for information on demand is a highly integrated artificial intelligence system where there are no inaccessible data silos. In addition, the AI must be able to accurately merge data across them and deliver it in a way that can be acted upon.
In the current state of play, we need a system that can understand the user’s role at that moment in the context of the business, sync what the user is doing in the system, and deliver the right information in the right format for informed decision making.
Therefore, in order for AI to make information on demand possible, two things must happen.
- The AI must be fully integrated within and across the system (data integration)
- It must be aware of and obey the data security and business rules by which the current system operates ( governance)
It’s an outsized problem that needs an equally outsized solution. An AI-augmented ERP could make the goals of data democratization possible without the use of multiple platforms and extensive staff training through the implementation of an AI-powered universal user interface which provides an information on demand service.
AI as the User Interface
An AI system that has been programmed and trained properly, and that is fully integrated within and across an ERP system, can quickly get information from multiple silos and create a temporary view of how the data should be presented across them.
In this way, AI becomes the ‘‘universal’ user interface in the moment and effectively removes barriers to data democratization caused by too much data and users not sure how to access or manipulate it.
The four “C’s” of good UI design are control, comfort, clarity, and consistency. The best user interfaces make it easy for people to accomplish tasks without confusion or frustration. This is where AI is uniquely positioned to be a game changer. The fluid, just-in-time interface it creates allows users to control when and where they access data and gives them ways to act on the information that are clear, consistent, and can be carried out with minimal additional user input.
For this to happen, the AI system needs to actually be part and parcel of the interface, which brings us full circle back to integration and trustworthiness.
- The integration solves the silo problem and gives you a system you can trust to give you the right information and options each and every time.
- The integration allows the system to update your data and the UI after each interaction, while ensuring flawless data governance.
Most data democratization goals are made possible by true integration across silos because the ERP can provide all of the relevant data to what the user is doing at that time in the system.
A Note About Security & Business Rules
For businesses to run well, security rules must be set in stone. But business rules are highly conditional and their fields are allowed to be changed under certain circumstances which are captured in minute detail. This adds another level of complexity for the AI to address.
To enable unencumbered data democratization, the AI network needs to know all the rules that govern data, the user interface needs to allow for only the correct changes, and then the system needs to be able to update the data while adhering to all the security and business rules. It’s a big lift that no agentic AI systems have mastered yet due to their inherent limitations.
Data Democratization, Corner Cases, and ERPS
Business corner cases or workflow process gaps are situations that fall outside of a company’s normal business processes and are not easily accounted for in their enterprise software solutions. They occur when a company’s enterprise software can’t accommodate the business processes employees use to get work done.
Corner cases can arise in any software platform, but they arise most commonly in ERP systems due to their ubiquity and the large number of corner cases they intentionally do not address. The data democratization that AI-ERP transformation enables is the solution to the productivity dilemma of corner cases.
Let’s look at a widespread ERP example. Many heavy equipment companies have reservation systems that are separate from their delivery systems. Consequently, if a piece of equipment is reserved, the reservation and the delivery tickets are in different locations.
If for any reason the equipment is not accepted at the job site, the reservation system needs to know that the equipment that was supposed to be there is not actually there. The job still needs that equipment, but now it needs a replacement, and so the reservation ticket needs to be reopened. The vast majority of ERPs can’t handle this very basic scenario.
An AI system that is tightly integrated within and across the system would reopen the reservation ticket, notify the responsible employee, and give them the following information:
- Why the first piece of equipment was rejected
- What equipment needs to replace it
- Where that equipment is currently located
- When and where it needs to be transported
- Who is available to complete the delivery
- The action or actions that need to be taken next
This is information on demand — true data democratization — in action. With the AI-generated universal user interface, the employee has all the information she needs to quickly make the right business decisions and, seconds later, the process of replacing the equipment is initiated. The implications for operations, customer service, and data management are far-reaching. Employees are more productive, customers are more satisfied, and data quality remains high. Multiply this across hundreds of transactions a day, five days a week and the productivity gains compound quickly.
AI-Enabled Information on Demand
Data democratization is about one thing: pursuing productivity for your business. There are a lot of ancillary benefits, such as data accessibility, simplification, cooperation, and empowerment; but ultimately better data democratization must result in increased productivity because that’s how it will be measured.
In 2025, all businesses are data driven, whether they realize it or not. If information isn’t available in real time, bottlenecks occur and the organization’s ability to leverage data to its advantage evaporates. Data democratization can free the business to leverage its proprietary information to its full potential, but only if the right information gets to the right people at the right time with trustworthy options for further action.
AI-powered information on demand, correctly implemented, improves data quality, enables productive data analysis, and frees the company from having to worry about extensive data literacy programs. Once AI becomes the user interface, the only thing employees will need to be concerned with is the next most appropriate action. They won’t need to worry about becoming data analysts or achieving data literacy. They will be free to make the best decision in the moment, knowing they have accurate, actionable information — and that will be a force multiplier for productivity.
About Atigro
Atigro is a proven ERP transformation firm that pairs its modular augmentation capabilities with AI-native frameworks. Atigro’s experience and capabilities generate the rapid development and provisioning of new ERP functionality that meets dynamically changing business processes.