Cameron Tabor, senior vice president of software engineering for Protos Security, recently talked exclusively to Security Sales & Integration about the growth of artificial intelligence operational assistants and how they can help security teams.
Security Sales & Integration: Nearly every security technology company now claims to use artificial intelligence. What specific operational problem did Protos set out to solve with its new AI-powered operational assistant?
Cameron Tabor: Security teams don’t have a data shortage. They have a time shortage. The information is already in the system: incidents, shift coverage, site history, vendor performance and billing. But it’s spread across so many reports that answering one question can mean opening five of them and reconciling the information by hand.
We set out to remove that friction. The Protos Assistant lets a user ask a straightforward question, such as which locations had the most incidents this month or where fulfillment is trending down and receive a useful answer quickly. The goal wasn’t to add AI for its own sake. It was to shorten the distance between an operational question and an informed decision.
SSI: Security programs generate enormous amounts of information across incidents, shifts, locations, and service providers. How can AI help users find what matters without simply adding another layer of data?
Tabor: AI becomes valuable when it reduces complexity instead of decorating it. A dashboard requires the user to already know which report to open, which filters to apply and what pattern to look for. A conversational assistant starts where the user starts: with the question.
It surfaces the exceptions: the sites with repeat incidents, the shifts that went unfilled and the provider whose performance slipped against the rest of the portfolio. The user still brings the context and judgment. They just don’t have to hunt through every row to get there.
SSI: Protos’ technology serves clients, internal operations teams and service providers. How do the information needs of those groups differ, and how did that influence the platform’s design?
Tabor: Each group looks at the same security program from a different position. Clients want portfolio-level visibility. Are posts being filled? Where are incidents increasing? Which locations or providers need attention? Our operations teams need to identify exceptions early so they can coordinate coverage, investigate issues and keep service moving. Providers need clear, timely details about the work they’re responsible for delivering.
A pre-built report has to guess at all of that in advance. Somebody decides which columns matter and which cut of the data will be useful and every user after that works within somebody else’s guess. An assistant takes its direction from the person asking: the role they hold, the account they’re examining and the question they’ve asked. It then shapes the answer around those needs.
It’s the same information underneath, with the same permissions governing what the user can see. But the view is built for the question in front of them rather than the one we anticipated months ago.
SSI: What safeguards are needed to ensure that AI-generated information is accurate, secure and worthy of customer trust?
Tabor: Trust starts with where the data comes from. The assistant reads the operational record we already maintain: incidents, shifts, post-coverage, provider performance and billing. That record is cultivated, reconciled and checked at intake. It’s the same data our teams and clients already rely on, under the same access controls.
Our approach to trust emphasizes transparency rather than relying solely on restrictive guardrails. We could handcuff the assistant until it answers only a short list of pre-approved questions but it would be safe and close to useless. We’d be back to pre-built reports.
Instead, it shows the basis for its answer: what data it pulled, how it interpreted the question and which records produced the number. The user also has a role in getting useful results. Asking a good operational question means understanding the operation, what’s being measured and over what period.
Transparency makes that possible. If the assistant interprets the question differently than the user intended, the user can see that and ask again.
When a number is going to a client or becoming part of a billing dispute, traditional reporting retains its value. A report is fixed, repeatable and auditable. It’s the layer against which the assistant’s answer can be checked.
SSI: Where should AI support human judgment in physical security and where should people remain firmly in control?
Tabor: AI is well suited to speed, scale and pattern recognition. It can scan large volumes of operational information, highlight anomalies and help people decide where to look first. Across a portfolio of locations, shifts and providers, that’s real work nobody has time to do by hand. The initial scanning doesn’t require human judgment.
Judgment is everything that comes after that. Safety, escalation, deployment, investigation and the interpretation of real-world context stay with people. The assistant can tell you that incident activity changed at a location. An experienced security leader knows the site, the client, the officer, the neighborhood and what changed there last month. That leader determines what drove the change and how to respond.
The consequences are real and physical, and accountability for those decisions belongs to a person. AI shortens the time it takes to see what’s happening. It doesn’t decide what to do about it.
SSI: What measurable results should a customer expect before an AI security tool can be considered operationally valuable?
Tabor: Start with time to insight. How long does it take someone to get from a question to an answer they trust? That’s easy to measure because everybody knows what the old path cost: the reports pulled, the people asked and the days spent waiting.
Then look at what changed operationally. Are service gaps caught earlier? Are exceptions resolved faster? Are recurring issues surfacing that used to stay buried because nobody thought to run that particular report? Those are the numbers that show the tool changed the work, not just the interface.
Then look at adoption and be honest about who’s using it. If the only person who can get value from it is a technical analyst, the accessibility problem hasn’t been solved. It’s been relocated. Usage by operations managers and account managers is the real signal.
SSI: For security leaders attending GSX 2026 who want to move from AI promises to practical results, what can they expect to learn by connecting with you and other Protos executives at booth 2405?
Tabor: They can see the Protos Assistant working inside the Client Portal and experience the difference between searching for data and asking an operational question. We can walk through practical use cases involving incident trends, fulfillment, provider performance, and multi-location oversight.
Just as importantly, our executives can discuss how the technology fits within a fully managed security program spanning guarding, off-duty law enforcement, remote and specialized services. We want visitors to bring us the operational questions they struggle to answer today.
The best AI conversation does not begin with the model. It begins with the decision you need to make.
