Factory Intelligence · Offsight AI
MES Adoption Missed the SaaS Revolution, AI Offers a New Opportunity
Why shop floor software has struggled to keep pace with the cloud era, and how vision intelligence can make operational data easier to capture and act on.
An ERP can tell a manufacturer what was ordered, what materials are needed to complete a job, and what a job should roughly cost. Yet a production manager may still need to walk the floor, call a supervisor, or wait for a spreadsheet to determine whether that job will actually meet the deadline and what the real cost may be.
That gap has persisted through generations of technology: on-premise software, cloud platforms, and SaaS (Software as a Service). The question is no longer whether factories need software. It is whether the software can reflect what is happening on the floor while there is still time to act.
That gap was supposed to be filled by a manufacturing execution system (MES). A real time shop floor tool that can relay feedback from the floor to the front office to manage the core competency of any manufacturing business.
Why MES Adoption Fell Short Through the SaaS Era
A 2026 industry survey from Rockwell Automation found that only 28% of respondents had implemented an MES enterprise-wide, and just 23% reported full integration with ERP, product lifecycle, quality, and operational technology systems.
The Manufacturing Leadership Council’s 2024 Data Mastery survey paints a bleaker picture, where 70% of respondents still entered at least some manufacturing data manually, and 68% used spreadsheets to analyze some or all of it. Despite the rapid adoption of ERP systems, CRMs and other software platforms over the last two decades MES software has lagged behind.
Today a manufacturer can own sophisticated front office software and still depend on handwritten notes, delayed updates, and heroic spreadsheet work to understand whats happening each shift.
Resistance can be a rational operational decision
It is tempting to blame slow deployment on a reluctance to change. On a shop floor, hesitation often reflects a legitimate calculation. A new screen, scan, form, or approval step can interrupt an established sequence of work. Every extra interaction competes with the operator’s real job: making a good product safely, consistently, and on time.
If a system asks people to document every micro-activity, it can turn data collection into a second shift. The resulting records may look comprehensive while becoming late, inconsistent, or incomplete under production pressure. At the other extreme, a system may be easy enough to use but capture so little context that the output has holes.
Both failure modes damage trust amongst operators. Managers question the data. Executives see a dashboard but cannot connect its numbers to throughput, schedule risk, or cost. The implementation stalls even when the underlying need for visibility is real.
What AI and Vision Intelligence Changes
Vision intelligence offers a different way to capture events. A supported camera can be connected in minutes and positioned to observe a machine, product, work area, or shared piece of equipment. One manufacturing engineer can then use a simple self guided framework to teach models what matters in their environment: a machine’s On and Off states, a product at a station, a tool cart in a bay, or an area under repair.
Cloud computing handles model training and recognition without requiring each plant to buy and maintain its own AI servers or GPUs.
In Offsight AI’s Vision AI feature set, cameras send timestamped photos through the shift. The console and iPhone app let a team mark examples, name visible states, set confidence criteria, and review results. A state change for example, a machine under repair or a product moving to its next work area, is recorded in seconds after multiple photos agree. Camera installation can be fast. Vision works best when it reduces operator input while keeping people in control of production.
Once common shop floor events, products, tools, equipment and machines are marked in the camera frames. The model is trained and autonomous detection begins. In this world the barriers that hindered MES adoption, including push back due to falling productivity or major data gaps that never allow the system to achieve true ROI, can be easily overcome.
AI changes the narrative by reducing the implementation cost, onboarding and long term value of an MES to the point that it is a no-brainer.
Vision Intelligence Acts as the “Eyes” and “Ears” for LLMs
Once a sophisticated vision model is trained on your shop floor, one camera view can enable AI to be your full time operations assistant and can address:
- Production trackingSee when work enters and leaves a station, monitor progress, estimate cycle times and throughput, and surface bottlenecks before a late job becomes a surprise.
- Machine monitoringRecord visible On, Off, In Use, Under Repair, idle, and other trained states to build a usable history of utilization and downtime.
- Tools and equipmentLocate shared carts, welders, test equipment, and other visible assets by work area, reducing time spent searching or duplicating purchases.
- MaintenanceCapture visible stops and repair activity as timely events that technicians and managers can review alongside work orders and machine records.
- AI analyticsTurn a series of confirmed events into live dashboards, shift summaries, alerts, trends, and questions a manager can ask in their preferred format.
Vision is especially valuable where a machine’s state is visible but expensive to obtain through a custom controller integration. It can complement PLC, sensor, ERP, and other system data where those sources are available.
From floor events to decisions and financial visibility. The return comes from connecting observations to action. If a machine stops, the right person can be alerted while the shift is still in progress. If a product waits too long in one area, a manager can move work or investigate the handoff. If a shared tool is missing, the team can find its last visible location.
The strongest implementations make useful data a by-product of work, rather than another task imposed on the people doing it. Operators keep their attention on safe, high-quality production. Supervisors gain timely context. Managers can act in real time. Executives can see where capacity and margin are being lost and measure whether changes actually improve performance.
The opportunity ahead
AI can make manufacturing software genuinely easier to live with: fewer manual updates, more complete operational context, and faster answers to the questions that matter. It will not remove the need for sound process design, worker input, or data governance. It can remove much of the friction that has kept digital systems from reaching their potential on the shop floor.
That is a shift worth pursuing: technology that observes what matters, doesn’t disrupt the work already happening, and helps everyone from operators to executives focus on what they do best.
Explore Offsight AI Vision AI and see how camera-based recognition can connect visible shop-floor activity to real-time factory management.
Reach out to Offsight AI to trial Vision AI for your factory, and we will send you a free camera to get started.
See what your floor can show
Explore Offsight Vision AI
Start with a camera view and the signals that matter to your operations.