Walk any security trade show floor and the pitch sounds identical from booth to booth: AI-powered detection, real-time alerts, enterprise-grade reliability. FSIE 2026, running September 3 to 5 at the Jio World Convention Centre in Mumbai, will put more than 180 exhibitors in front of exactly that pitch. For a security head with a real budget and a real deployment to plan, the challenge isn't finding a vendor who claims AI. It's finding the handful whose claims survive a direct question.
This is the checklist to bring with you, organized the way an actual evaluation should run: what you need before you get to the booth, then the questions that separate a real AI video surveillance vendor from one selling a dashboard with the word "AI" on the login screen.
Before You Get to the Booth: Know What You're Solving For
The single biggest mistake at a trade show is evaluating vendors against a generic idea of "good AI CCTV" instead of the specific problem your organization actually has. A warehouse worried about perimeter breaches and forklift safety needs a different analytics mix than a retail chain tracking footfall and shrinkage, and both need something different from a hospital managing visitor flow and restricted-zone access. Write down the two or three problems you're actually trying to solve before you walk the floor, and score every vendor against those, not against whichever feature list sounds most impressive on a booth screen.
Detection and Accuracy: Where Most Claims Fall Apart First
This is the category where a specific question separates a real capability from a marketing line, faster than any other on the floor.
Where detection actually happens is the single most revealing question you can ask, because it determines everything downstream: latency, bandwidth cost, and whether the system still works if connectivity drops. On-camera detection and server-dependent detection get sold with the same word, "AI-powered," but they behave completely differently on-site.
Real detection speed and false-alarm rate, measured under normal conditions rather than a controlled booth demo, tell you more than any spec sheet. A vendor who can only quote lab numbers, or who can't quote a false-alarm rate at all, hasn't tested the system the way you're about to use it.
Resilience during an outage matters more than it sounds like it should. Ask directly whether detection and alerting continue if the internet connection drops, because plenty of "AI-powered" systems quietly stop working the moment they can't reach a server.
What's actually bundled versus gated behind a separate license is worth pinning down before, not after, a quote arrives. Vendors often lead with one headline analytic and treat the rest as add-ons priced separately.
Red flag: an answer that leads with "our AI is very advanced" instead of a number.
Certification and Compliance: The Filter That Runs Before Features Do
For government, PSU, BFSI, and an increasing share of private-sector buyers, this category isn't a nice-to-have. It's checked before a single feature gets compared.
Independent certification, not a self-declared claim, is the distinction that matters. BIS-ER certification on the hardware and STQC certification on the video management software are both things a vendor should be able to document, not just state verbally at the booth.
Compliance posture, including NDAA status and whether the system contains any high-risk foreign components, is worth confirming directly rather than assuming from the brand's country of origin.
Where it's actually manufactured, and whether that can be documented for a tender or audit, closes the loop on the certification conversation. A verbal "yes, made in India" means little without paperwork behind it.
Red flag: certification claims that can't be backed by a document on request.
Integration and Scalability: What Breaks Between the Demo and Site 10
A system that performs well on one camera at a booth is a different animal from one running across a hundred cameras spread over ten sites, and this is the gap most demos are built to hide.
Compatibility with what you already have, through ONVIF support or a dedicated bridge device, determines whether an existing camera or NVR investment survives the upgrade or gets thrown out entirely.
Behavior across multiple locations is worth asking about directly: does the platform give one unified view across sites, or does it mean a separate login and a separate management burden per location as the deployment grows.
Performance at scale, not just at the booth's single demo camera, is the honest question. Ask what actually happens to alert reliability and system responsiveness once camera count moves from dozens to hundreds.
Red flag: a vendor who can describe the single-site demo in detail but goes vague the moment you ask about site eight, nine, and ten.
Cost and Support: The Number That Isn't on the Booth Slide
The price quoted on the show floor is rarely the number that shows up on an invoice two years into the deployment.
Total cost of ownership, across hardware, licensing, cloud subscription, and support, is the number to press for, not the per-camera price alone. Subscription-based pricing in particular can look inexpensive upfront and compound significantly as storage and analytics tiers scale up.
How support actually works, including what the warranty covers and the real response time on a hardware or software issue, matters more once the system is live than it ever does during the sales conversation.
Update cadence for firmware and security patches is easy to skip past on the floor and expensive to discover later. Ask whether it's included in the subscription or billed separately.
Red flag: a TCO answer that only ever comes back as "it depends," with nothing more specific offered.
Data, Privacy, and Ownership: The Category Buyers Ask About Last
This is consistently the category that gets the least floor time and matters the most once a contract is signed.
Who owns the footage and the metadata analytics generate, the customer or the vendor, is a question worth asking in plain terms before signing anything, not inferring from a terms-of-service document afterward.
Where data physically lives, and whether that location satisfies your organization's data residency requirements, is non-negotiable for government and many BFSI buyers, and worth confirming even if your organization isn't in either category yet.
Who can access feeds, and how that access is logged, closes the loop. A system with no audit trail on who viewed what footage, and when, is a gap that tends to surface during an incident review, not before one.
Red flag: hesitation or a non-answer on data ownership specifically, more than on any other question here.
Quick-Reference: Bring This to the Booth
Where does detection run: camera or server?
What's the real false-alarm rate, on a real site?
Does it keep alerting during an internet outage?
Is certification documented, not just claimed?
Does it support ONVIF or a bridge device for existing hardware?
One unified view across sites, or a login per location?
What's the total cost of ownership, not the per-camera price?
What does support actually cover, and how fast?
Who owns the footage and analytics data?
Is there an audit trail on who accessed what, and when?
Any vendor worth a serious relationship should answer all ten directly, on the floor, without deferring every one to "let's set up a follow-up call."
How ArcisAI Answers This Same Checklist
ArcisAI is an AI-native CCTV brand by Adiance Technologies — BIS-ER and STQC certified, designed and manufactured in India. Running ArcisAI against its own checklist: detection for face recognition, ANPR, crowd counting, fire and smoke detection, PPE compliance, and intrusion runs directly on the camera, at sub-100ms speed and with roughly 80 percent less bandwidth used than sending raw footage to a server, and it keeps running through an internet outage rather than stopping with it. On certification, the hardware is BIS-ER certified, the Cloud VMS is separately STQC certified, and the line is NDAA compliant and manufactured in India with no high-risk foreign components, documented rather than asserted. On integration, ONVIF support and the ArcisAI Bridge Device bring existing cameras and NVRs into the system without a rip-and-replace, and the Cloud VMS gives one unified view across a single site or a multi-location deployment rather than a login per location. On cost, there's no separate per-camera annual license fee for edge AI analytics — that capability is built into the hardware itself — and Cloud AI and VMS licensing scale in tiers by camera count rather than a single large upfront commitment, so ask for the tier breakdown against your specific site count rather than a single blended quote. On data, ArcisAI Cloud is hosted on secure Indian servers, which lines up directly with the residency requirements most government and BFSI buyers are already working against.
Meeting Vendors at FSIE 2026
Bring the quick-reference list above, printed or saved on your phone, and put every vendor through the same ten questions in the same order, including ArcisAI. Score answers, not enthusiasm. Ask for a live demo specifically on the analytics that matter to your actual use case rather than whatever the booth is set up to show by default, and treat any hesitation on the certification or data-ownership questions as a flag worth resolving before a contract is signed, not after.
FSIE 2026 runs September 3 to 5 at the Jio World Convention Centre, Mumbai. Ready to run this checklist against ArcisAI before the show? Book a free demo at arcisai.io.
Frequently Asked Questions
What's the most important question to ask an AI video surveillance vendor? Whether detection happens on the camera or requires footage to reach a server first. That single answer determines latency, bandwidth cost, and whether the system keeps working during an internet outage, and it filters out most vague "AI-powered" claims immediately.
Why does certification matter more than feature lists for many buyers? Government, PSU, and BFSI procurement processes check certification, such as BIS-ER and STQC, before features are compared at all. A system with strong analytics but no independent certification often can't clear the first filter of a tender, regardless of performance.
How should a security head evaluate cost across different vendors at a trade show? Ask for total cost of ownership across hardware, licensing, cloud subscription, and support, not the per-camera sticker price alone. Subscription models can look cheaper upfront and cost more over several years, depending on how storage and analytics tiers are priced.
Can existing CCTV cameras be upgraded with AI instead of replaced entirely? Often yes, if the existing cameras or NVR support ONVIF, or if the vendor offers a bridge device that adds AI and cloud capability to already-installed hardware. This is worth confirming before assuming a full system replacement is necessary.
What data questions should be asked before signing a video surveillance contract? Who owns the footage and analytics metadata, where the data is physically hosted, whether that location meets your organization's compliance requirements, and who has access to live and historical footage, with an audit trail of that access.