How to Use AI in Operations: From Demand Forecasting to Operator Assistance
By Pedro Sonda, dooop
In operations, AI can be applied in two main areas: the analytical front, which forecasts demand, equipment failures, quality defects, and bottlenecks based on historical data; and the generative front, which supports people with quick access to procedures, incident logging, shift reports, and problem investigation. The first depends on reliable data; the second, on organized documentation. Choosing the right approach for each problem is the first step.
Updated in October 2026.
Why Operations Is Where AI Reduces Costs
In McKinsey’s global research, cost reductions attributed to AI appear most frequently in supply chain management, service operations, and manufacturing (McKinsey, 2026). BCG estimates that about 70% of AI’s potential value lies in core business functions such as sales, marketing, manufacturing, supply chain, and pricing (BCG, 2025).
In Brazil, among industries using AI, 52% apply it in production, a lower percentage than in areas like administration and sales (IBGE, 2025).
Interpretation: The potential in operations is high, but usage is still lower than in administrative areas, partly because it requires factory floor data and integration with operational systems.
dooop Framework: The 2 AI Fronts in Operations
| Analytical Front | Generative Front | |
|---|---|---|
| What it does | Forecasts, detects, optimizes | Explains, summarizes, drafts, guides |
| Examples | Demand forecasting; predictive maintenance; visual defect inspection; route and inventory optimization | Assistant answering questions about procedures; shift reports from notes; support for root cause analysis; training new operators |
| Depends on | Historical data, sensors, integrated systems | Updated procedures, manuals, and records |
| Time to results | Medium to long term | Short to medium term |
| Main risk | Model trained on poor data | Responses based on incorrect or outdated documents |
Typical Opportunities by Type
| Type | Examples in Operations |
|---|---|
| Automation | Incident logging by voice or text; report filling; issuing work orders |
| Decision Support | Production sequencing suggestions; stockout risk alerts |
| Analysis | Failure patterns by equipment, shift, or batch; recurring causes of scrap |
| New Services | Remote equipment monitoring sold as a service to clients |
Data Readiness Test
Before starting an analytical project, answer:
- Do the data exist? Is the event we want to predict recorded (failures, defects, sales)?
- Are they digital? Or are they in loose spreadsheets, paper, or people’s memory?
- Are they reliable? Are the records complete and consistent over time?
- Is there enough history? Has the event occurred enough times for a model to learn?
- Are they accessible? Can systems be integrated without a months-long project?
If most answers are “no,” the first project is likely organizing data collection, not building the model. Gartner identifies data quality and availability as the main barrier to AI adoption in finance (Gartner, 2024), and the pattern repeats in operations.
Why the Generative Front Is Often a Good Starting Point
Many operations have documented procedures that are difficult to consult when needed. An assistant based on official documents can reduce search time and support less experienced operators. In a study with 5,179 agents, generative AI support increased productivity by an average of 14% and by 34% among less experienced workers (Brynjolfsson, Li, and Raymond, NBER). The context was customer service, but the mechanism—transferring knowledge from experts to beginners—also applies to operations.
Caution: The assistant is only as good as the documents it relies on. Outdated procedures generate incorrect answers that appear official.
What to Do Now
- List the three biggest costs or losses in your operation and ask whether they are forecasting problems (analytical front) or information problems for people (generative front).
- Apply the data readiness test before any analytical project.
- Review and organize procedures before implementing a generative assistant.
To map opportunities in operations, explore dooop’s discovery tool at dooop.com.br/discovery-de-oportunidades-de-ia.
Frequently Asked Questions
Does predictive maintenance require sensors on all equipment? Not necessarily. Start with critical equipment whose downtime costs are higher, and evaluate if current data already allow some type of prediction.
Is AI in operations only for large industries? No. Medium-sized companies often start with the generative front and simple analyses of existing data, which require less investment.
Also Read
- How to Use AI to Improve Administrative Processes
- How to Evaluate if AI Automation Is Really Worth It
- How to Use AI to Improve Business Productivity
About dooop
dooop is a consultancy based in Caxias do Sul (RS) that prepares leaders and organizations to transform artificial intelligence into business capability through strategy, governance, and training.