How Agent-Based AI Makes Companies More Successful<\/h2>\n\n\n\n
After two years of experimenting with artificial intelligence (AI), companies have come to realize which form of AI actually delivers results. It is not the common AI assistants that answer questions, summarize documents, and write drafts\u2014useful for individuals, but they do not change the flow of business processes. It is innovative AI systems that take on a task rather than simply answering a question. They understand the initial situation, plan the steps, execute them in the company\u2019s development environment, verify the result, and present it ready for approval\u2014in a transparent, repeatable manner, and in hours rather than weeks. The key point here is that such a system handles the task across all roles\u2014business analysis, development, infrastructure, and testing\u2014which previously had to be passed back and forth among several specialists. Because it acts like an agent who sees a task through to completion, rather than like an assistant who answers a question, these are referred to as \u201eagent-based AI systems.\u201c The difference from a standard assistant, such as a chatbot, is the same as the difference between a consultant who writes recommendations and a well-coordinated, experienced team that fully completes the assigned tasks\u2014though one thing is clear: humans make the final decisions regarding production deployment.<\/p>\n\n\n\n
From the Boardroom<\/h2>\n\n\n\n
A scene like the one currently playing out in many companies:<\/strong> Thoughts from a board member the night after the steering committee meeting on the SAP migration: \u201eIt\u2019s been the same thing for three years. It\u2019s taking longer, it\u2019s going to cost another fifty million, and I don\u2019t know if I\u2019ll even go live with the new system. It\u2019s been postponed for the third time, and the second project manager has been replaced. My IT team and the consultants are at each other\u2019s throats, and none of them is ultimately responsible\u2014except me. I\u2019m sick of hearing about it. So is my supervisory board.\u201c<\/p>\n<\/div><\/div>\n\n\n\n There are three reasons why this makes a company more successful\u2014the same three factors that set competitors apart from one another. First, the speed with which a company responds to the market: a new product, a new distribution channel, a new pricing model\u2014everything must be reflected in the business systems before the company can act on it. Whoever does this first achieves the best margin; whoever comes last competes on price. Second, the costs and risks of day-to-day operations: The more capacity operations tie up, the less is left for change\u2014and the less a company has documented and understands its own code within its systems, the greater the risk of cyberattacks. Third, the knowledge of what sets a company\u2019s business apart from others\u2014the processes, rules, and exceptions that have distinguished a company from the competition over decades. Agent-based AI addresses all three of these points because it is deployed where it makes a difference: in the systems the company uses to run its business. This applies to large-scale modernization as well as to companies that want to leave their systems as they are and simply make day-to-day operations faster, more secure, and more cost-effective.<\/p>\n\n\n\n Two leading global industrial companies\u2014which are among the first to integrate the Nova Intelligence platform into their core systems and are willing to share their results with the public\u2014have demonstrated what this means in terms of numbers: Festo, the automation specialist based in Esslingen, and the KION Group in Frankfurt, one of the world\u2019s largest suppliers of forklifts and warehouse technology. The time from a business requirement to a productive change in the system drops from months to weeks, ongoing operations require only a fraction of the previous capacity, and the costs of the transformation itself are estimated by experienced consulting firms to be reduced by about half. The figures for time and capacity are measured, not estimated; all numbers are derived step by step in this article. The result achieved at the end of such a modernization now has a name: Cleanfield.<\/p>\n\n\n\n In most German industrial companies, the system used to run the business is called SAP. That\u2019s where the processes, master data, orders, and financial statements are stored\u2014along with the business logic that defines the company. That\u2019s also where the largest budgets for the coming years will be allocated, because the old systems are being phased out and modernization is on the agenda for every executive board member, whether they like it or not. That\u2019s why SAP is the place where agent-based AI has the greatest impact.<\/p>\n\n\n\n One prerequisite is often underestimated in many companies: agent-based AI is only as good as the foundation on which it operates. It needs the company\u2019s data and business logic in digital form\u2014complete and of reliable quality\u2014not stored in people\u2019s heads, not in Excel spreadsheets, and not in code that no one reads anymore. Where this foundation is lacking, even the best AI can only provide plausible guesses. That\u2019s why every serious implementation begins with the same task: fully understanding and documenting the existing system so that the agents can work based on verified facts. For humans, this is a laborious process. It is the first step that an agent-based platform handles on its own\u2014and the reason why the results are reliable afterward.<\/p>\n\n\n\n That is exactly why Nova Intelligence was founded\u2014the company behind the platform used by Festo and KION: an agent-based AI platform for enterprise systems, including SAP and others, founded in Silicon Valley in 2024 by researchers who develop AI systems themselves, and by the co-inventor of SAP HANA, who, together with SAP founder Prof. Hasso Plattner, initiated the HANA project and led it as his deputy\u2014the \u201eNew Architecture\u201c that became the foundation of SAP S\/4HANA. The platform\u2019s capabilities are most clearly demonstrated by a specific case\u2014a system that was already optimized but still managed to improve significantly once again.<\/p>\n\n\n\n At KION, a business-critical logistics process was running on an application that had recently been developed for S\/4HANA and had already been optimized\u2014yet it was still too slow for operational use. The code was familiar, and the team had even tried to speed it up on their own. It wasn\u2019t until an AI platform\u2014one that understands not only the code but all the underlying processes in the system\u2014was brought in that the key areas for improvement were identified. The result: The program now runs 400 percent faster.<\/p>\n\n\n\n This wasn\u2019t an isolated incident. Across all tasks that KION\u2019s SAP experts have measured since then, they\u2019ve saved an average of 86 percent of their working time. Troubleshooting a system failure\u2014which used to take about four hours\u2014now takes less than 20 minutes. Users rate the platform 9.7 out of 10; the recommendation rate is 100 percent.<\/p>\n\n\n\n \u201eNova Intelligence addresses the two requirements that are most important to us: operating systems that are truly Clean Core-compliant and achieving the development performance that our SAP landscape demands,\u201c says Carsten Engelbrecht, CIO of the KION Group. \u201eThe results were outstanding.\u201c<\/p>\n\n\n\n For a CIO, figures like these are a success. For a CEO or CFO, they are more than that: they are an indication that the cost-benefit analysis that executive boards have been using for twenty years to make decisions about their core systems is no longer valid.<\/p>\n\n\n\n The cost-benefit analysis to date has been as follows: The SAP system is the backbone of the business, and its in-house developments\u2014which have evolved over decades\u2014contain the process logic that truly sets the company apart from the competition. But modernizing this system\u2014the transition from SAP ERP, commonly referred to as ECC, to the new S\/4HANA product line\u2014has, in most cases, failed to provide companies with a competitive advantage: the same processes, the same logic, just on a new technical platform, at a cost running into the hundreds of millions. So companies invested as little as possible, postponed the major modernization for as long as they could, and treated it as a cost center. According to a Gartner forecast, by the end of 2027, just under half of ECC customers worldwide will still not have migrated to S\/4HANA. Even those who have begun the migration often still run part of their systems on ECC\u2014not out of negligence, but because the traditional route through consulting firms is too expensive, too complex, and too risky. Anyone expected to spend hundreds of millions on a program that ultimately delivers the same business processes as before is right to hesitate.<\/p>\n\n\n\n At the same time, there\u2019s a second question on every executive board\u2019s agenda: Where should we invest in AI, and where will it pay off? In most companies, the honest answer is: We don\u2019t know yet. There are many pilot projects, but few viable business models. The problem is rarely the technology. It\u2019s the knowledge that AI lacks\u2014knowledge about how the company actually operates: which pricing rules apply to which customers, how custom orders are scheduled, and why a process runs differently at Plant B than at Plant A.<\/p>\n\n\n\n And it is precisely this knowledge that is concentrated in one place: in the SAP system\u2019s proprietary code, which has evolved over time. For decades, companies have used it to capture what sets them apart from their competitors; the core business logic is almost always embedded in the SAP system itself.<\/p>\n\n\n\n However, this knowledge is mostly undocumented; it leaves the company with every employee who retires, and it remains inaccessible to any AI as long as it exists only as code.<\/p>\n\n\n\n This means both questions are closely related. An agent-based AI platform that modernizes an SAP system must first fully capture this knowledge\u2014every process, every rule, every dependency\u2014and store it in a documented, machine-readable format. The result is the core of what the company needs for its own AI: a company-specific knowledge model of its business processes that belongs solely to it, grows with every use, and serves as the foundation for the AI applications of the coming years.<\/p>\n\n\n\n That\u2019s the new trade-off. Modernizing core systems is no longer a cost center that companies keep putting off. It\u2019s the strategic path through which a company builds the foundation for its own AI platform\u2014and its benefits are evident from day one in measurable results.<\/p>\n\n\n\n What does that mean in terms of numbers? Nova Intelligence is now being used by a growing number of customers. Two of them, Festo and KION, are making extensive use of the platform within their SAP teams, have disclosed their metrics, and are publicly endorsing it. At Festo, the team first compiled these metrics for its own executive board to demonstrate the return on investment from the Nova investment.<\/p>\n\n\n\n The first lever:<\/strong> Productivity. Festo has been working with SAP since the 1970s and was among the first thirty companies worldwide to implement the system. Since Nova has been in use within the SAP teams, their productivity across all tasks has increased more than fivefold. For individual tasks, the increase is even greater: documentation, for example, which used to take three months to complete, is now produced in about one day. Festo\u2019s CIO, Dr. J\u00f6rn Weigle, puts it this way: \u201eThe way we answer the three big questions about our SAP program\u2014how long does it take, what does it cost, and how big is the risk\u2014has improved significantly since Nova became part of our teams. Our team estimates that, on average across all tasks, it is more than five times as productive. A capacity gain of this magnitude fundamentally changes the rules of the game for our organization.\u201c The value of such a capacity gain depends on how a company uses it.<\/p>\n\n\n\n The second lever:<\/strong> Costs. The impact on costs stems from productivity\u2014in a specific order. The lever that has the fastest effect is external consultants: Any task that your own team can complete five times faster no longer needs to be outsourced. Next comes internal capacity, then the software itself, and finally risk\u2014disruptions, outages, and security incidents, which become less frequent once a company fully understands and has documented its own code. For a CFO, one point is important here: A blanket percentage for cost reduction is deliberately not mentioned here because it depends on the starting point\u2014on the scope of the in-house code, the proportion of external consultants, and the number of disruptions. What is available, however, are the measured productivity figures from Festo and KION, which can be used to calculate the cost impact for your own company.<\/p>\n\n\n\n The third lever:<\/strong> Capacity. This is where the calculation becomes interesting for the executive board. In a typical SAP organization with 100 specialists, 80 are responsible for day-to-day operations\u2014maintenance, troubleshooting, adjustments, and support tickets. Twenty work on new processes and business requirements. If productivity increases fivefold, as measured at Festo, sixteen specialists can handle the same operational workload that previously required eighty (16 \u00d7 5 = 80). Sixty-four are freed up and join the twenty: Eighty-four instead of twenty work on new initiatives\u2014more than four times as many, without a single new hire and without a single additional consultant. That is why Weigle speaks of changed rules of the game: work that used to be completely outside the scope of planning becomes possible.<\/p>\n\n\n\nA program that had already been optimized<\/h2>\n\n\n\n
The Investment Question, Reexamined<\/h2>\n\n\n\n
Cost-Effectiveness in Detail<\/h2>\n\n\n\n
![\"\"](\"https:\/\/e3mag.com\/wp-content\/uploads\/2026\/10\/Grafik1_Kapazitaet_Innovation_DE_web.jpg\")