The real problem: investing millions with nothing to show for it
There is a massive disconnect between what companies expect from automation and what they actually manage to deliver on a daily basis. Many organizations dive headfirst into modernization projects, purchase expensive platforms, hire well-known consulting firms, and at the end of the day, remain stuck in pilots that never move beyond the drawing board. Data scattered across different systems, teams that do not know how to operate the new tools, and a chronic lack of strategy mean the return on investment simply never materializes. This scenario is far more common than it seems, and it is not limited to small companies — major industry players also fall into this kind of trap. The problem starts when technology is treated as a magic bullet without considering the operational context, organizational culture, and most importantly, the people who will be dealing with all of it on the factory floor.
Recent research reinforces this perception by showing that Artificial Intelligence tends to intensify the volume of work before generating any measurable efficiency gains. In other words, during the first months of implementation, teams spend more time feeding models, fixing integration issues, and adapting processes than actually reaping benefits. On top of that, the energy costs associated with these technologies keep climbing, which raises an important question about financial and environmental sustainability. Companies face resource shortages, unexpected disruptions, and delays that can be difficult to reverse. The question that remains is straightforward: what separates the initiatives that actually work from those stuck in endless testing cycles?
In a conversation with Rahul Negi, Head of Industrial AI at Honeywell, the answer became much clearer. According to him, success comes down to a few fundamental pillars — smart data integration, physical AI with deep sector knowledge, and above all, valuing the people who operate the machines and processes every day. It is not about replacing people with algorithms, but about empowering those who already understand the operation with tools that genuinely make sense. Honeywell has worked with a range of customers and partners and observed that physical AI improves business outcomes by enhancing safety, connecting data, and improving decision-making — all while supporting workers who may not have decades of experience. This perspective completely changes the game and explains why some companies are achieving concrete results while others continue to spin their wheels 🏭
The most common mistakes in the race toward automation
In the rush to implement new technologies, many companies make missteps that compromise the entire project before it even gets off the ground. One of the most frequent mistakes is underestimating how much technology deployment consumes in terms of time and resources. Global supply chains, energy systems, and critical infrastructure are already under unprecedented pressure, and adding layers of technological complexity without proper planning only makes things worse.
Another recurring error involves a lack of understanding about how different data points interconnect within an operation and impact results. When information is disconnected, adaptability becomes limited and the optimization process slows down considerably. To make matters worse, the global shortage of skilled labor in industry makes it harder to interpret that data and act on the results it generates. All of these challenges, when combined, create significant cybersecurity risks involving IT and OT environments that can take months, or even years, to reverse.
The good news is that a proven path exists. Companies that invest first in understanding their own data ecosystem, that map the interdependencies between systems, and that plan deployment gradually and realistically are able to avoid these traps. The secret is not in how fast you adopt, but in the intelligence behind each implementation decision.
Data integration as the starting point
One of the biggest bottlenecks in modern industry is not a lack of data — it is an overload of disconnected data. Sensors, legacy systems, ERPs, maintenance platforms, manual records in spreadsheets, and quality control software generate enormous volumes of information every single day. The problem is that these sources rarely talk to each other. When a company tries to apply Artificial Intelligence on top of a fragmented foundation, the result is predictable: inaccurate models, wrong predictions, and decisions based on incomplete assumptions. Smart data integration is not just a technical step — it is the foundation on which any serious automation initiative needs to be built. Without it, every investment in advanced algorithms and cutting-edge infrastructure turns into waste.
Receive the best innovation content in your email.
All the news, tips, trends, and resources you're looking for, delivered to your inbox.
By subscribing to the newsletter, you agree to receive communications from Método Viral. We are committed to always protecting and respecting your privacy.
Negi explained that Honeywell works with the concept of a unified data layer, where information from different sources is normalized, contextualized, and made available in real time to Artificial Intelligence models. A concrete example is the Experion Process Knowledge System (PKS), an alarm assistant for operators that analyzes all historical data and provides real-time recommendations so operators can identify and resolve risks and issues. In practice, this means an operator on the factory floor can receive predictive alerts about equipment failures, adjust production parameters based on actual trends, and make faster decisions without having to check five different systems.
This approach eliminates informational noise and allows automation to work where it truly matters — at the critical points of operation that directly impact efficiency and safety. When data flows in an organized way, technology stops being a burden and becomes a genuine ally in the production process.
Another relevant point is that well-executed integration drastically reduces the time needed to move Artificial Intelligence projects into production. Companies that invest first in data governance and architecture are able to move past the pilot phase much faster, because their models are born fed by consistent information that accurately represents operational reality. This contrasts sharply with organizations that skip this step and end up spending months — or even years — trying to clean up and organize databases that should have been structured from the start. The lesson here is simple, even though it is often ignored: before thinking about sophisticated algorithms, you need to make sure the raw material for those algorithms is in order.
The driving force: physical AI as the new pillar of automation
If the question about what should be the driving force behind every automation strategy had been asked a year ago, the answer would probably have been the convergence of three technologies: Artificial Intelligence, 5G, and cloud. But the landscape has evolved. According to Negi, these technologies are now enabling a new lever — physical AI.
Physical AI is the foundation that supports successful automation implementations. In practice, it leads to greater efficiency gains, enables preventive maintenance that reduces downtime, and helps close the skills gap in the industrial sector. Workers with limited experience can operate with the competence of someone who has been on the job for years, because the technology provides the context and recommendations that previously only existed in the minds of veteran operators. All of this translates into real productivity gains that show up in plant performance indicators.
A case that illustrates this potential well is Honeywell’s partnership with ADNOC Borouge to develop AI-powered autonomous operations. According to ADNOC Borouge, once deployed across their facilities, these solutions have the potential to improve efficiency by 20 percent, reduce downtime by 20 percent, and lower operational costs by 15 percent. This is a practical example of the ideal role of AI — extending decision-making capability beyond individual plants to reach the entire enterprise.
Historically, achieving full autonomy has been difficult because data remained trapped in silos and traditional computing power could not handle the volume of information that facilities produced. Agentic AI and physical AI are changing this reality by processing massive data sets in the cloud and at the network edge simultaneously.
Industrial AI needs context, not just scale
There is a huge difference between a generic Artificial Intelligence and an AI built with deep knowledge of a specific sector. The large language models and generalist tools that dominate the headlines are impressive in terms of capability, but when it comes to industry, they fail to capture nuances that only someone who lives the operation truly understands. A pulp mill faces completely different challenges than an oil refinery, and the Artificial Intelligence models that generate real efficiency are those trained on domain-specific data, calibrated by industry experts, and validated in real-world production scenarios.
Negi pointed out that the market is flooded with AI tools and vendors, but what really matters is the depth of domain knowledge. For those navigating this vast sea of options, the recommendation is to seek partners who understand how automation can be applied to operational technology in sectors like energy, manufacturing, utilities, and industrial and commercial buildings. Generic tools may look attractive because of pricing or promises, but without that vertical expertise, results tend to fall short of expectations.
This domain knowledge is what allows, for example, an automation system to differentiate a real equipment anomaly from a normal variation caused by a shift change or weather condition. Without that context, false alerts multiply, teams lose trust in the technology, and the system ends up being ignored — which is, ironically, the opposite of what was expected when the project was approved. Industrial Artificial Intelligence needs to be precise, reliable, and above all, useful to those on the front lines. When an operator receives a recommendation that makes sense within their practical experience, they adopt the tool naturally. When they receive out-of-context alerts, they turn the system off and go back to doing everything the way they did before. This dynamic is decisive in the success or failure of any digital transformation project in industry.
Industrial facilities like oil refineries and petrochemical plants generate enormous volumes of operational data every day from thousands of endpoints and assets. These are precisely the environments where AI-powered autonomy thrives, because there is enough volume of contextually rich data to train and refine models that truly make a difference in operations.
Another aspect that deserves attention is the computational and energy cost of Artificial Intelligence solutions. Very large and generic models consume enormous resources to run, which is not always justified in industrial environments where latency needs to be minimal and processing often happens at the network edge. Verticalized solutions optimized for the specific operational context tend to be lighter, faster, and more sustainable from an energy standpoint. This matters both for the budget and for the environmental goals that many industry companies are pursuing. Efficiency cannot be only operational — it also needs to account for the environmental impact of the very technology used to achieve it.
The human factor as a competitive advantage
Perhaps the most important point in this entire discussion is the role of people. The dominant narrative around automation and Artificial Intelligence tends to revolve around workforce replacement, but the reality on the factory floor tells a very different story. Industrial automation always requires human oversight — people are the catalyst for better outcomes because of the specialized knowledge, contextual understanding, and judgment that ensure safe and beneficial deployment. While concerns about job displacement are understandable, they do not reflect how industrial automation actually operates.
Industry faces a growing shortage of skilled professionals, especially in technical and operational roles. In the United States, approximately 26 percent of the national industrial workforce is already eligible for retirement. Experienced workers are leaving and taking with them decades of tacit knowledge that no manual documents. New operators face a steep learning curve and struggle to manage peak performance from their plants and assets. This creates significant side effects, including unplanned downtime, rising maintenance costs, and aging infrastructure.
In this context, Artificial Intelligence functions less as a replacement and more as a tool for preserving and transferring knowledge. Intelligent systems that capture patterns from veteran operators and translate them into recommendations for new employees are a concrete example of how technology can amplify human capability rather than eliminate it. Manufacturers are increasingly leaning on AI to help new operators sharpen their skills and maximize their productivity.
Negi emphasized that Honeywell believes companies have a collective responsibility to equip today’s workers and leaders with the tools and knowledge needed to navigate the path toward autonomy. This means intuitive interfaces, alerts that respect the actual workflow, and recommendations that arrive at the right moment — not as an avalanche of notifications that do more harm than good. Real efficiency happens when technology adapts to the human, not the other way around. Companies that force their operators to radically change the way they work to accommodate a new system are creating unnecessary resistance and sabotaging their own automation projects. The approach that works is incremental, respectful, and collaborative — involving teams from the solution design phase all the way through implementation and continuous refinement.
Trends that will shape industry in 2026
Beyond Artificial Intelligence, there are other important trends that should materially impact the manufacturing sector this year. Negi highlighted three movements that deserve special attention:
- Exodus of experienced operators: the challenge of mass retirement has intensified in recent months. The loss of this institutional knowledge puts direct pressure on plant productivity and demands technology solutions that fill this gap in a practical and efficient way.
- API-first approaches and interoperability: customers are increasingly asking for consistent user experiences and interfaces across different systems and platforms. The ability to integrate solutions from different vendors seamlessly is becoming a decisive criterion when choosing technology partners.
- A real path toward full autonomy: manufacturers are beginning to achieve full autonomy through more open solutions, connected ecosystems, embedded knowledge, and value-driven deployments. Many customers were stuck in the so-called infinite pilot phase last year, but they are now moving past that stage and seeing time-to-ROI accelerate. Solutions are being delivered more quickly, and end-to-end lifecycle management is paving the way for autonomy.
These trends reinforce that industrial transformation is not just a technology issue. It involves people, processes, data governance, and a strategic vision that connects every investment to tangible results. Companies that understand this dynamic will lead the next phase of industry, while the rest risk falling behind in an increasingly competitive and demanding landscape.
What actually works in practice
Investing in Artificial Intelligence and automation without simultaneously investing in developing people is a recipe for failure. Hands-on training, upskilling programs, and transparent communication about the role of technology in each team member’s daily routine make all the difference between a successful implementation and a project that dies in a drawer. The industry companies that are seeing real results are the ones that understood that data, algorithms, and machines are only part of the equation — the rest depends on motivated, prepared, and heard people.
The conversation with Rahul Negi made it clear that there are no shortcuts. Automation strategies that truly work combine three non-negotiable elements: an integrated and reliable data foundation, Artificial Intelligence with deep domain knowledge, and a genuine commitment to developing the people who will operate and benefit from these technologies. When these three pillars are aligned, technology stops being a promise and becomes a real competitive advantage — with measurable gains in efficiency, productivity, and safety that show up in the metrics that truly matter 💡
Rahul Negi is an executive at Honeywell Process Automation, leading Digitalization, Autonomy, and AI initiatives, with over 20 years of experience in Strategy, Consulting, Business Development, and AI/ML.
