Deploying artificial intelligence (AI) tools should be seen as a business initiative rather than a technology initiative, says Martin Soukup, CTO of Solink, a cloud-based video security and data intelligence platform. The approach makes a difference that plays out in practical terms.
A technology initiative is tactical and solution-oriented, often solving an isolated problem within a single department's silo. In other words, it might solve just a security problem.
In contrast, a business initiative is cross-disciplinary, driven by executive leadership to impact the entire organisation in a measurable way. “Viewing AI tools as a business initiative ensures they adapt to multiple departmental needs and drive high-level strategic growth rather than just fixing a localised technical issue,” says Soukup.
AI platform combines video and business systems
Demonstrating the value of embracing AI as a business initiative is Solink’s new video intelligence platform
Demonstrating the value of embracing AI as a business initiative is Solink’s new video intelligence platform that combines video with business systems to detect threats and send real-time alerts. The capabilities are designed to help security teams operate smarter and more efficiently, supporting the next generation of lean, intelligence-driven security operations centers (SOCs). And benefits of the platform extend beyond security to include other parts of the business.
“At the end of the day, video cameras only tell part of the story, much like looking at a 2D shape,” says Soukup. “They capture dimensions and colors, but only from a single perspective, even with powerful AI analysis providing traffic metrics, wait times, product interaction, and compliance logs. The addition of operational data, such as inventory, shipping, labor, and sales data, fills out that shape into 3D, adding critical context.”
For example, while a camera simply shows three people standing in a line, and AI data flags that those individuals have been waiting for over five minutes, operational data lets you know also that a cashier has not checked into their shift yet. Combining these data streams gives businesses full situational awareness, turning passive observation into immediate, contextual intervention.
AI agents handle repetitive tasks
The Solink platform uses AI Agents, which are software that operates autonomously to handle manual, repetitive, or high-volume tasks based on specific instructions and tools. For Solink users, AI Agents unify their existing camera and data source investments into a single automated workflow.
“AI Agents transform Solink from a security tool into an operational investment that benefits all departments,” says Soukup. “Modern operational leaders are turning to AI to scale productivity without increasing overhead.”
Integrating business intelligence
Operations checks compliance, loss prevention investigates theft, and security assesses threats
Customers already use Solink to centralise video, integrate business intelligence, and trigger real-world responses like real-world alerts, two-way communication, or smart locks. AI Agents tie these pieces together, analyse the data and identify what matters most, and then take action either without human intervention or with humans in the loop.
Businesses currently watch video manually across separate workflows. Operations checks compliance, loss prevention investigates theft, and security assesses threats. However, a human must always watch the footage to get results. Solink AI Agents automate this entire workflow. If an operational event can be seen on camera and matters to the business, a Solink Agent can be integrated to automatically review, assess, and trigger the next step in the workflow across hundreds or thousands of locations simultaneously.
Calling all stakeholders to the discussion
When discussing applications, stakeholders should include revenue and marketing officers, compliance officers, and line-of-business executives. Agents leverage existing infrastructure by adding an intelligent automation layer that drives efficiency and grows revenue across the entire organisation without large capital investment.
Solink AI Agents will not replace employees; they empower lean teams to maximise their existing resources. Job descriptions will shift toward human-in-the-loop workflows where humans and machines collaborate. The AI Agent handles the high volume, filtering out noise and processing data at scale. Human employees set up and guide the Agents, give feedback to improve performance over time, and step in to make final decisions, managing the nuanced situations that require critical thinking, empathy, and deep context.
Passive data visualisation
In contrast, a Solink AI Agent is deployed with a clear mandate for a specific task, deliverable, or outcome
Given the convergence of physical and digital security, AI Agents, not dashboards, are the right unit of enterprise AI deployment. Dashboards typically aggregate data without evaluating the context or importance of the source. Data looks identical whether it comes from 10 sources or 100. In contrast, a Solink AI Agent is deployed with a clear mandate for a specific task, deliverable, or outcome.
This approach shifts the focus from passive data visualisation to active intent, creating a direct path of accountability and clear attribution for business results. With dashboards, everyone can see the problem, but nobody owns it. An agent with a mandate either did the job or it did not, and you can audit exactly what it saw and why it acted or did not.
Overcoming obstacles to deploy AI agents
The primary obstacles to integrating AI Agents into the physical security workflow are change management, building trust, and system interoperability. Physical security requires absolute reliability, which demands rigorous initial configuration to establish trust. Furthermore, Agents are only as effective as the data they can access.
Overcoming these hurdles requires comprehensive team training, shifting organisational mindsets, and ensuring seamless integration with existing tools so the Agents have the full context needed to execute tasks accurately and integrate with human workflows seamlessly.
Addressing common security challenges
Several use cases are driving the deployment of AI Agents among Solink's customers
Several use cases are driving the deployment of AI Agents among Solink's customers. Overnight guarding, loss prevention, and store readiness will be the first wave of use cases. These are proven, high-frequency scenarios in which Solink provides pre-built template Agents to address common security challenges out of the box.
The second wave of use cases will include highly customisable occupancy tracking, health and safety, and food quality. Pre-built template Agents are coming for these next.
In addition, custom, niche operational use cases can be tailored to a specific business environment. Just like hiring a specialised contractor, companies can deploy custom Agents to tackle unique operational challenges rapidly across all locations.
Measuring the effectiveness of AI agents
The effectiveness of AI Agents can be measured by multiple factors.
- Core Accuracy, and consistency and accuracy of detections and automated outputs.
- Business Outcomes delivered, such as reduced inventory loss, reduced fraud, and faster incident resolution.
- Financial Impact, measured through recovered budget, reduced operational overhead, and increased per-store conversion.
There is a perception that AI Agents possess predictive or flawless capabilities, like stopping crime before it happens or operating perfectly in dark, chaotic environments. Not true, says Soukup. AI cannot understand ambiguous requests like "look for suspicious activity." Success requires explicit parameters, tools, and skills, defining what is acceptable, what is unauthorised, and how to handle the gray areas.
Another misconception is that Agents need a fundamentally different security model than people. “I run security for this company as well as engineering, and I hold our own Agents to the same standard I would hold a third-party vendor,” says Soukup. “That means least privilege, full audit trails, anomaly detection, random audits, and no action they can't explain after the fact.”
