‍

Assessment

Our assessment revealed the following:

While the data velocity flowing across the shop floor was high, the value of the first-party data that the vendor was able to mine and process for its SMB customers had not been determined.
The leadership team was new to AI/ML and required help to drive and implement strategy, as well as create AI/ML algorithms and models.

The solution value proposition was strong, however there were incumbent vendors who held strong market positions.

There was no clear vision to set the direction. Defining a new market category or niche was critical to stand out.

The results are summarized below.

‍

AI Business Problem and Value

Manufacturing component and finished goods quality inspection is not failsafe using human-only visual inspection.

The business value is to improve product quality on the manufacturing shop floor and use AI-assisted quality control powered by the client SaaS solution.

‍

AI Use Cases

We looked across the manufacturing value chain to identify potential AI use cases that could be powered by our client’s shop floor process control SaaS solution.

We identified the following potential AI uses cases for a potential representative customer

Source

• Component supplier selection
• Digital marketing operations
• Front office order management
• Component ordering
• Component forecasting
• Inventory receiving

Prep

• Inventory management
• Inventory picking
• Factory scheduling
• Labor scheduling
• Labor training

Make

• Work instructions
• Chat bot enhanced work
• AR-based work help
• SKU Kitting
• Assembly
• IoT Sensor control
• Quality control
• Reporting
• Business intelligence
• Predictive maintenance

Route

• Inventory stacking
• Delivery tracking to distributors
• Widget EOL prediction status tracking

‍

We started with Quality Control for the AI MVP concept.

‍

Prediction Variables

After identifying the key use cases, we identified the key prediction variables that would be generated by a data-driven ML system.

Pass/Fail of manufactured components and finished goods on the shop floor.

‍

Modeling Approach

In collaboration with the client’s technical team, we proposed a comprehensive set of experiments on data sets collected from representive customer shop floor data.

Logistic regression (categorical) with computer vision anomaly detection models.

​Linear regression to determine pass/fail based on real world data.

‍

Data Strategy

Our next steps were to quantify the available data sets in terms of their daily volume and diversity of sources.

Product images, audio, operator notes and process data from the shop floor manufacturing system will be captured in real time during the manufacturing process.

Data could be ingested in the 3rd party managed private or hybrid data lake, where it will be stored, processed and correlated in real time at the edge or in the cloud.

‍

Results

The client accepted the findings and rolled out the ML server in a product that was released in a private beta and is now available to its select clients for a full release testing.

‍

Leadership team and AI readiness

The client management team was relatively new to AI/ML so additional training and consulting resources would be needed to enhance and train them to become proficient in order to enable the identified potential AI use cases.

Data pipeline, algorithms and models

With high velocity of operational data from the manufacturing, there is a significant potential to apply AI techniques; however, the signal in the data was unclear so an experimentation and discovery process was needed to better understand the value.

AI algorithms, models, architecture and deployment

The client needed help to implement algorithms and build new models to train and deploy on the shop floor data.

Category Strategy

Via our category process we defined a new market category, including a market blueprint and ecosystem strategy summary.

Product, Roadmap, Value Prop and Competitive

Further work was needed to help the client map the product roadmap to the new category.

‍