AlwaysAI: a Practical Computer Vision Solution to Optimize Operations
We explore AlwaysAI, a solution that detects operational anomalies for a variety of industries including retail, restaurants, and event venues.
Background
AlwaysAI is a Series B startup (1) founded in 2018 (2). AlwaysAI is focused on providing computer vision solutions across 5 core markets: smart retail, restaurants, manufacturing, construction, and venues (2).
Their core product is a computer vision solution that detects objects, people, anomalies, and events to improve operations; AlwaysAI provides significant depth and clarity using clients’ existing cameras and captures data to target inefficiencies and detect anomalies (2). Their value proposition is to provide visual insights to apply process improvements, and increase revenue (2). While safety is one aspect to help with compliance, it is not the sole benefit.
This video (2) provides a product summary:
Where Does AI Fit In?
In terms of how it works, AlwaysAI collects and uploads client data to their site, and then annotates data by highlighting and labeling objects to detect. AlwaysAl trains the model based on the annotated dataset, develops an app using their API library, and deploys the app remotely. Then, AlwaysAI processes incoming data from the app and provides real-time analytics. As a part of their offering, AlwaysAI has pre-trained, pre-optimized models that the client can browse, choose training parameters, and can later adjust to balance speed and cost.
Recently, AlwaysAI announced a release of enhanced ML Operations platform features (3), improving ease of use in creating and deploying enterprise-grade models. Through this, AlwaysAI provides a model evaluation tool called “modelIQ”, which provides performance insights to improve speed and training effectiveness, and allows for better model optimization through fine tuning and increased precision (3).
Below shows their dataset management (2):
How does AlwaysAI Create and Capture Value?
Example use cases to create value, and pricing are highlighted below:
Smart Retail
In retail, AlwaysAI utilizes its solution to count customers in real-time and monitor checkout wait times (4). This enables clients to better allocate staffing and attend to shoppers promptly (4). Historically, 86% of retailer customers have left stores due to long wait times (4); through adequate staffing based on real-time alerts, AlwaysAI can enable additional revenue capture while also improving the client’s service levels. Furthermore, foot traffic tracking enables a better understanding of which products customers are most engaged with to optimize product placement (4).
Smart Restaurants
For restaurants, AlwaysAI assists with service speed by tracking where the bottlenecks occur in a food line during order fulfillment (5). Similar to retail, this solution enables the client to make better staffing decisions and reduce potential customers from turning away. Additionally, occupancy counting at restaurants can assist with better demand planning (5). Finally, in drive-thrus, AlwaysAI provides real-time vehicle ID tracking to optimize order fulfillment and service speeds by alerting the client when the lines are too long (5).
Smart Venues
For event spaces, AlwaysAI can count patrons to understand crowd dynamics for traffic management (6). Concession stand lines can be monitored to notify fans of less crowded stands and reduce wait times. For sponsored ads, AlwaysAI’s product can track how many people interact with the specific display and the time they linger (6), thereby providing data to assist with pricing decisions (e.g. greater display engagement based on a specific location can enable them to charge higher to advertisers).
Smart Construction
In construction, AlwaysAI reduces liability by detecting safety violations, akin to other security and safety competitors. One use case is monitoring inadequate PPE (7) to reduce injury. Additionally, AlwaysAI can detect when site gates are open and shut and send alerts (7). One interesting use case is its ability to track construction progress in real-time, providing updates on tool usage, and material delivery (7); an example app notification is below (7):
Pricing
AlwaysAI captures value through a subscription model (8). For enterprise customers, pricing depends on usage and number of deployments (8). This is a core advantage, as it can be viewed as a “pay as you need” model. Through this offering, they provide unlimited inferencing (processing images, camera streams), unlimited users, and onboarding support (8). One aspect of the pricing is flexibility – clients can adjust the number of computer vision product platforms they want over time (8). Thus, as they scale, clients have the option to invest in additional platforms based on their application. AlwaysAI also offers discounts for long-term contracts, which helps improve user base stickiness.
Opportunities, Challenges, Changes
Challenges include competition, such as those focused on security/safety. Competitors include firms such as Intenseye (workplace safety) and Verkada (AI/ML security systems). With more firms rising in the security space, it may be harder to win long-term contracts, or they may face pricing pressure, especially in construction and manufacturing. However, a core advantage is AlwaysAI’s diversity in applications; strategically, they should continue expanding applications to reduce risk.
AlwaysAI includes numerous documentation around application issues (2); thus, technical challenges and data complexity may exist for clients who aren’t experts. One opportunity is to expand their offering by including personalized 1:1 ongoing support. This may involve an organizational change of hiring a broad technical sales team to better manage customers, but may be an opportunity for additional revenue (e.g. premiums for support). Furthermore, the model’s training accuracy in detecting anomalies may not be 100%, leading to false negatives/positives, or alert biases. Another challenge lies in data privacy; though using clients’ existing cameras suggests AlwaysAI might be aligned with the clients’ privacy rules, individual privacy is a concern given the data volume processed.
While AlwaysAI’s value is rooted in providing alerts for an event, it does not fix the issue. Rather, the client must divert personnel to address it. Thus, if staffing is unavailable or past capacity, the alerts are not useful, running risks of low retention. Since AlwaysAI leverages businesses’ existing cameras to assist in detection (2), they may be limited in detection quality (e.g. poor camera resolution). If a client has poor IT infrastructure, the “real-time” analytics nature can also be impacted.
Sources
1. https://www.crunchbase.com/organization/alwaysai
4. https://alwaysai.co/solutions/alwaysai-smart-retail
5. https://alwaysai.co/solutions/alwaysai-smart-restaurants
6. https://alwaysai.co/solutions/alwaysai-smart-venues
7. https://alwaysai.co/solutions/alwaysai-smart-construction
Cool post !…One of the biggest concerns here would be data privacy as the AI field is striving to be heavily regulated based on regions. This would pose a challenge when they seek to expand their business. Also it would be cool to how they expanded in the manufacturing sector to attain operational efficiency as organizations already have some sort of measuring criteria in place . It would interesting to note how did they approach this.