91仓库

By adopting a Data-First approach, you can build connected intelligence while providing AI analysis to automate decision-making, creating highly reactive connected intelligent systems.
By Evan J Schwartz

I have spent 35 years watching companies across Natural Gas, Forestry, Pulp and Paper, Scrap Metal, and 91仓库 and Recycling make the same expensive mistake. They buy software and think that they have bought transformation. The biggest lie companies tell themselves about their technology strategy is: 鈥淲e鈥檙e transforming because we bought the software.鈥�

They assume modern software will fix broken processes. That integrating one new platform means the whole ecosystem is now 鈥渄igital.鈥� That data will magically become cleaner, more complete, or more connected. That buying AI means they are ready for AI. But the costliest lie, the one that quietly drains millions across decades, is the belief that tools transform companies. They do not.

Connected operations transform companies. Information that flows and mimics your processes provides transformation. Data that enables decisions transforms businesses.

The Hidden Cost of Averages Land
Here is what I mean by that. I once worked with a pulp mill that had optimized its entire sales operation around inbound ticket data. Large loads of raw material arrived and were sorted into inventory locations based on grade, quality, and product. Sales thought they could hedge raw materials and build sales slots ahead of processing.

Smart, right? Except they were living in what I call 鈥淎verages Land.鈥� They used conversion factors that had worked for decades. Tons and pounds of whole tree translated to finished goods using historical averages. Then came two years of heavy rainfall, one of the wettest seasons on record. Their production numbers went sideways.

It took over a year of data analysts digging through every step of the process to find the problem: moisture content had increased by nearly 20 percent. Water weighs a lot. You cannot turn water into paper. Their 20-year-old conversion factors were now off by almost 20 percent.

The problem was so deeply hidden in code that no one could figure out why the mill鈥檚 production costs were way off the mark. They had expensive tools. They had data. But their tools were not talking to each other. They looked only at inbound data, not the graded material or production output. They worked off an average instead of reading data from their own machines. This is what happens when you do not have connected intelligence.

The Contamination Problem Nobody Wants to See
In waste and recycling, the same pattern shows up around contamination. Most recyclers do not work from actuals. They sample some loads, spread those numbers around, create supplier averages, and call it good. When things go south as they did during COVID, it is almost always retrospective.

You end up in 鈥渨ould鈥檃 should鈥檃鈥� territory. How much pain will you put up with before you gather enough retrospective data points to act? It gets costly fast. The right answer is knowing in near real-time and doing something about it when costs go outside your limits. But few recyclers chase this. Many do not think the technology exists today.

Here is the uncomfortable truth: The technology exists. I work with it every day. AI-powered vision systems on collection vehicles can scan every container lift and tie event for contamination and trace it to its source. These systems achieve 96.57 percent accuracy in waste classification.

According to the Ellen MacArthur Foundation, AI-enabled recycling could reduce global landfill waste by 20 percent and save $10 billion annually by 2030. Toronto uses AI sensors in public bins to detect fullness and contamination, reducing collection costs by 30 percent annually. So why aren鈥檛 more companies chasing it?

Every business has its own set of fear zones. Maybe that kind of exposure demystifies the business to the point that the current overseer is not needed. Perhaps it shows that there is more contamination than there should be and having edit access to a spreadsheet is a power too good to give up.

Tech is different, and different means change, and change is painful. But as competition leapfrogs your business, you will eventually be forced to adapt. How long you can wait comes down to your pain threshold of how much business you are willing to lose.

What the Winners Are Doing Differently
Large, national waste and recycling companies have adopted a Data-First approach. They looked at the data they have. How good is it? Where does it come from? What do they need to make connected, intelligent decisions? Do they currently have the ability to feed their decision systems that data directly, or is it coming in through 鈥渟wivel chair鈥� or 鈥渉and jam鈥�? Then, that becomes the area of improvement.

Once the data is right, they can build connected intelligence on top of that. They can provide AI analysis to automate decision-making, creating highly reactive connected intelligent systems. This matters because the global waste recycling services market was valued at $65.09 billion in 2024 and is projected to reach $109.8 billion by 2033 according to Grand View Research. The companies that architect connected ecosystems will capture that growth. The ones building isolated tool stacks will not.

Where Each Technology Fits
Here is what most companies get wrong about AI in waste operations. They confuse deterministic optimization, machine learning, and Agentic AI. Each has a specific role. Using the wrong one costs millions.

Deterministic Optimization
Deterministic optimization handles route planning. Optimizing a route is not a 鈥渇uzzy鈥� logic system. You might get real-world problems such as car accidents, road closures, but that does not make the logic fuzzy. It just means the deterministic outcome lives in a chaotic system, and you cope the best you can.

Route optimization software uses algorithms to minimize fuel consumption while maximizing stops. The City of Spokane, WA saves up to $25,000 annually in printing costs alone after implementing telematics solutions, with hundreds of thousands saved over the contract鈥檚 lifetime.

Machine Learning
Machine learning handles the fuzzy logic, business rules that cannot be declared with precision. For example: 鈥淚 want to prioritize my premier customers and make sure they get serviced first, at the highest quality, then my discounted customers come next.鈥� That feels like a declarative statement until you ask: 鈥淲hat defines a premier customer?鈥�

Premier for rural Oklahoma is vastly different from premier in upstate New York. Now we are deep in fuzzy logic territory. This is something learned culturally within your organization.

Agentic AI
Agentic AI is where things get interesting. Yeah, you stop overweight trucks at the scale, but if your highest-margin customer comes in, you take that truck and make a phone call. You do not turn that truck away at the scales.
This is where agents shine. They come with some sense, just like a new employee, but quickly learn: 鈥淭hat customer gets away with murder!鈥� But any other time, we hold firm to the rule. Agents provide the connective tissue between sensing, deciding, acting, and learning in closed-loop operations.

The 24-Month Window
Here is my prediction: The next 24 months will separate industry leaders from followers based on their connectivity architecture decisions today. 91仓库 companies currently invest less than 1 percent of their revenue in ICT, while most funds go to trucks, containers, and installations. Forrester Research found that 47 percent of company revenues would be influenced by digitalization by 2020. Companies like Kodak and Nokia did 鈥渢oo little too late鈥� and did not survive the transformation.

The defensive nature of the waste sector makes it attractive for private equity firms looking for stable returns. But stability does not mean immunity from disruption. Research shows that digitalization in waste management can promote waste avoidance up to 65 percent. IoT sensors now provide real-time data on pollutant concentration levels with 99 percent accuracy. For particulate matter, Random Forest models achieve 84 percent accuracy. The technology exists. The ROI is measurable. The question is whether your organization has the courage to look at what your data actually shows.

What This Means for You
If you are leading a digital transformation project, here is what I want you to do tomorrow morning. Pull up your decision dashboard. Pick any metric. Ask: 鈥淲here does this metric come from?鈥� If you do not see a direct line from a live data source feeding that metric, you are in trouble. You are in Averages Land. Some smart analyst did the work at some point and produced a factor. That factor sits at the heart of your business and has for decades.

Now things are starting to shift, and no one really knows why. Track back to a live data source for every decision. If you cannot build that plumbing, if you are not sure how data gets in there, you are in trouble.

A new ERP will not fix a bad master data model. Route optimization will not succeed if operations and dispatch continue to schedule work the way they did 20 years ago. Vision AI will not scale if fleets lack consistent onboard connectivity. An IoT program will not produce ROI if the organization has not agreed on what decisions the sensors are meant to trigger.

An Agentic AI solution absolutely will not work if foundational business rules, functional APIs, and data governance are weak. In every failed transformation I have seen, whether a sawmill, a pipeline operator, or a recycling MRF, the root cause is the same: Leadership thought they could 鈥渂uy鈥� a future-state instead of architecting one.

Your competitors are not winning because they have better tools. They are winning because their tools actually talk to each other. That is connected intelligence. And over the next decade, it will decide who survives. | WA

Evan J. Schwartz is Chief Innovation Officer at AMCS Group, driving trustworthy AI vision across resource-intensive industries. He teaches technical project management, data and model building, and architecture design at Jacksonville University, and he created The Customer Journey Framework, which codifies 35+ years of enterprise-level ERP and digital solution deployment experience. For more information, visit .

Resources

• Klover.ai. (n.d.). 91仓库 Management AI strategy analysis of dominance in environmental services. Retrieved from www.klover.ai/waste-management-ai-strategy-analysis-of-dominance-in-environmental-services-waste-management-ai/
• Klover.ai. (n.d.). Republic Services AI strategy analysis of dominance in environmental services recycling waste management AI. Retrieved from www.klover.ai/republic-services-ai-strategy-analysis-of-dominance-in-environmental-services-recycling-waste-management-ai/
• MDPI. (2025). AI vision systems achieving 96.57% accuracy in waste classification. Retrieved from www.mdpi.com/2313-4321/7/1/9?type=check_update&version=1
• arXiv. (2020). AI-based waste classification research. Retrieved from https://arxiv.org/abs/2006.05873
• IJIRMPS. (2025). AI waste classification study. Retrieved from www.ijirmps.org/papers/2025/2/232392.pdf
• McKinsey & Company. (n.d.). Artificial intelligence and the circular economy. Retrieved from www.mckinsey.com/~/media/mckinsey/business%20functions/sustainability/our%20insights/artificial%20intelligence%20and%20the%20circular%20economy%20ai%20as%20a%20tool%20to%20accelerate%20the%20transition/artificial-intelligence-and-the-circular-economy.pdf
• Ellen MacArthur Foundation. (n.d.). Plastics and the circular economy deep dive. Retrieved from www.ellenmacarthurfoundation.org/plastics-and-the-circular-economy-deep-dive
• Zipdo. (n.d.). AI in the waste industry statistics. Retrieved from https://zipdo.co/ai-in-the-waste-industry-statistics/
• Grand View Research. (n.d.). 91仓库 recycling services market analysis. Retrieved from www.grandviewresearch.com/industry-analysis/waste-recycling-services-market
• GlobeNewswire. (2024). 91仓库 recycling services market to hit USD 93.24 billion by 2031. Retrieved from www.globenewswire.com/news-release/2024/11/12/2978857/0/en/91仓库-Recycling-Services-Market-to-Hit-USD-93-24-billion-by-2031-Coherent-Market-Insights.html
• Geotab. (n.d.). City of Spokane case study. Retrieved from www.geotab.com/case-study/city-of-spokane/
• DigitalOcean. (2022). DigitalOcean Currents report. Retrieved from https://anchor.digitalocean.com/rs/113-DTN-266/images/DigitalOcean-Currents_Dec-2022.pdf
• ResearchGate. (2022). Unlocking digital technologies for waste recycling in Industry 4.0 era. Retrieved from www.researchgate.net/publication/360215946_Unlocking_digital_technologies_for_waste_recycling_in_Industry_40_era_A_transformation_towards_a_digitalization-based_circular_economy_in_Indonesia
• Emerald. (n.d.). The impact of digital traceability on supply chain management. Retrieved from www.emerald.com/scm/article-abstract/29/3/497/1230888/The-impact-of-digital-traceability-on?redirectedFrom=fulltext
• ScienceDirect. (2022). Unlocking digital technologies for waste recycling in Industry 4.0 era. Retrieved from www.sciencedirect.com/science/article/abs/pii/S0959652622003845?via=ihub
• ResearchGate. (2023). Evaluation of machine learning algorithms for air quality prediction. Retrieved from www.researchgate.net/publication/376837255_Evaluation_of_Machine_Learning_Algorithms_for_Air_Quality_Index_AQI_Prediction
• ScienceDirect. (2025). Air quality monitoring using machine learning algorithms. Retrieved from www.sciencedirect.com/science/article/pii/S2773049224000278
• Frontiers in Environmental Science. (2024). Advances in AI and IoT for environmental pollution monitoring. Retrieved from www.frontiersin.org/journals/environmental-science/articles/10.3389/fenvs.2024.1336088/full
• ResearchGate. (2022). Regression trees modeling and forecasting of PM10 air pollution in urban areas. Retrieved from www.researchgate.net/publication/320375758_Regression_trees_modeling_and_forecasting_of_PM10_air_pollution_in_urban_areas
• AIP Publishing. (2020). Hybrid CART鈥揂RIMA approach for PM10 pollutant forecasting. Retrieved from https://pubs.aip.org/aip/acp/article-abstract/2302/1/060015/755025/Hybrid-CART-ARIMA-approach-for-PM10-pollutant?redirectedFrom=fulltext