AI applied to real results

Challenge

Interactions were lengthy and required specialized knowledge. A significant part of each conversation was spent identifying the cause of the customer's problem.

Diagnosis

From a consulting perspective, the opportunity was to make the knowledge of the best agents available during every conversation, making it easier to identify solutions.

How AI was used

A generative AI agent, trained on support history, suggested responses and internal documents in real time. The support agent decided what to use and remained responsible for the conversation.

Economic benefit

As an illustration of capacity, a volume that required 1,000 hours could require approximately 820 hours with 21.9% higher productivity: about 180 hours freed up, to improve response speed or reduce the team.

Challenge

Purchasing and replenishment depend on spreadsheets and the experience of those in charge. The chain overbuys in some stores and faces stockouts in others, leading to expired products, disposal and emergency transfers. Each store spends two hours a day analyzing and preparing orders.

Diagnosis

The proposed analysis cross-references sales, inventory, shelf life and losses by product and store. The goal is to identify where orders poorly matched to demand cause waste and repeated work, prioritizing the perishables with the greatest economic impact.

How AI was used

Forecasting models estimated demand by product and store, considering history, day of the week, promotions and seasonality. Using inventory and shelf-life rules, the system suggested purchase and replenishment quantities and flagged expiration risk. A pilot in two stores validated the gains without increasing stockouts.

Economic benefit

The reduction in losses and transfers totals R$26k per month. After deducting R$8k per month for the solution and its maintenance, the recurring gain would be R$18k per month, or R$216k over 12 months of stabilized operation.

With an initial investment of approximately R$80k for implementation and integrations, simple payback occurs five months after the expected gains are reached. The implementation period is added to this timeframe.

The freed-up hours could be allocated to inventory checks, replenishment and customer service. They were not counted as financial savings, since they do not automatically represent a reduction in payroll or overtime.

Challenge

Each proposal requires interpreting needs, combining products, checking technical rules and preparing the quote. This work takes ten hours per customer. Knowledge is concentrated in experienced professionals, creating queues and limiting personalization.

Diagnosis

The proposed analysis maps the time spent at each step and organizes catalogs, prices, constraints and approved examples. The goal is to make this knowledge available to the whole team and expand the capacity to develop tailor-made proposals.

How AI was used

A series of agents was created to interpret the customer's request, suggest combinations compatible with their preferences and budget, and prepare proposals using the catalog and commercial rules. Consultants validate measurements, feasibility, prices and the final solution. A pilot project was implemented to verify time, quality and the need for revisions before expansion.

Economic benefit

The capacity to send proposals nearly doubled, with a 20% increase in the monthly cost of the team and the solution. The cost per proposal would fall from approximately R$571 to R$357, a 37.5% reduction.

Since the company had the capacity to expand manufacturing, this initiative is expected to generate a 38% increase in the company's gross revenue next year, considering the current rate at which proposals convert into orders.

Challenge

Revenue depends on one-off service calls. Customers usually notice a failure only when the equipment is already malfunctioning, while the service provider has little visibility into its condition between visits.

Diagnosis

The proposed analysis cross-references failure history, installed equipment, service costs and customer interest. A pilot would assess which signals make it possible to anticipate problems and how much establishments would pay for continuous monitoring.

How AI was used

Sensors would record equipment temperature and operation. AI would cross-reference these readings with failure history to estimate risks and prioritize alerts for technical review. The team would advise the customer and schedule interventions. The subscription would include monitoring, reports and a limited number of preventive visits; out-of-scope repairs would be billed separately.

Economic benefit

The 100 contracts would generate R$45k per month. Direct costs of R$180 per contract, including connectivity, visits, consumables and sales taxes, would total R$18k. Technology, operations and support would add R$9k, leaving R$18k per month in contribution margin.

After deducting R$5k in contribution from the replaced on-demand services, the incremental gain would be R$13k per month, or R$156k over 12 stabilized months, before income taxes. The investment in sensors and implementation would be R$130k, equivalent to ten months of this gain. Implementation and building the customer base would extend the payback period.

The company would start selling continuous monitoring and prevention. Customers would gain visibility into their equipment, and the business would combine recurring subscriptions with additional services.