AI at the edge, people at the controls600-store connectivity case$30M growth financing in 2022Twenty20 acquired in 2024Bad camera angles still beat good AI

Company profile / AIoT + connectivity

EPIC iO Bet the Camera Was Never the Point

The Fort Mill company is stitching cameras, sensors and wireless links into one operating layer for remote sites - a useful idea whose economics depend less on magical AI than on good placement, clean signals and humans who still make the call.

By YesPress EditorialSeptember 1, 20269 min read

The least glamorous part of artificial intelligence is the part that has to work in the rain. A camera on a remote oil pad can identify a person, a vehicle or a leaking pipe only if it has power, a useful angle, enough light and a connection home. If any link breaks, the clever model becomes an expensive spectator. EPIC iO Technologies has built its business around this stubborn fact.

The Fort Mill, South Carolina company sells an operating layer for physical places: cameras, environmental sensors, cellular links, satellite service, edge computers, dashboards and, when required, humans watching around the clock. Its customers are the organizations with awkward maps - electric utilities, oil producers, construction sites, retailers, hospitals, schools and municipalities. Their assets are scattered. Their wiring is unreliable. Their existing systems rarely speak the same language.

EPIC iO did not begin as a clean-sheet software startup. It was assembled in December 2020 when IntelliSite, an AI and IoT operator, bought Broad Sky Networks, a managed wireless business founded in 2003. The new holding company later merged both operations under the EPIC iO name. That origin explains the product better than any acronym does: connect the site, collect the data, decide what matters, and keep someone accountable for the handoffs.

600locations in a published retail failover case
20-30%claimed data-cost savings in a SIM-management case
70K+square miles in a utility deployment

A roll-up with an operating thesis

The company’s catalog can sound like a trade-show aisle. DeepVision handles video management and computer vision. RemoteVision covers low-activity places where wired infrastructure is impractical. DeepInsights combines camera, sensor and third-party data into alerts and visualizations. WirelessWindow tracks SIMs, devices, carriers and usage. Mobile Surveillance Units put cameras, edge processing and wireless links on equipment that can move. A security operations center adds trained remote operators. EPIC Cortex is the umbrella over the collection.

The point is not that every customer buys every box. The point is that EPIC iO can own more of the failure chain. A retailer can buy automatic 4G or 5G backup for its checkout lanes. A utility can move a surveillance trailer as work shifts between substations. An oil producer can collect tank, pressure and camera data at distant pads, filter events at the edge and see the useful ones in a central room. A mobile medical provider can keep a secure imaging workflow alive with two active carriers.

That breadth is also the competitive argument. A buyer could assemble Genetec or Verkada for video, a cellular provider for backup links, a separate SIM-management platform, a local integrator and a guarding company. EPIC iO’s offer is to collapse several contracts and finger-pointing sessions into one managed system. It sells directly and through systems integrators, resellers and technology advisers, with partners including Dell, Intel, SAS, carriers and channel distributors.

A surveillance camera overlooking a busy roadway
The camera has one job. The network, edge computer, dashboard and sleepy human behind it have several more.

What failed first was the boring stuff

EPIC iO’s customer stories are vendor-written, so their unqualified triumphs deserve a raised eyebrow. The operational failures underneath them are more instructive. One multi-state utility said moving a competitor’s surveillance units took two to three weeks. Its officers struggled with field support and could not change coverage quickly. Another engineering company managed thousands of SIM cards manually and depended on a single carrier. An oil producer had cameras that frequently went offline and delivered poor images. A municipality found that fixed cameras stayed put while risk moved elsewhere.

These were not failures of frontier AI. They were failures of mobility, inventory, signal, maintenance and service. EPIC iO changed the architecture accordingly: mobile units instead of fixed-only coverage, multi-carrier connections instead of one network, edge processing instead of shipping every frame, and one remote console instead of spreadsheets. The utility said redeployment fell from weeks to days. The engineering customer reported average data-cost savings of 20 to 30 percent. An oil and gas case said remote automation recovered dozens of operator hours each week.

Where the measurable value showed up

SIM cost saving
20-30%
Retail sites
600
Oil cameras
200+

Bars normalize unlike units for visual comparison; labels show the disclosed figures. Customer contract values were not published.

What did it cost? EPIC iO does not post a price list and its customer contract values are not public. The business is quote-based, mixing subscriptions, managed connectivity, monitoring, deployment, hardware and support. The one large disclosed capital number belongs to EPIC iO itself: $30 million in growth financing from Runway Growth Capital in 2022. That money helped integrate and expand the combined teams. DFW Capital Partners is the private-equity backer, and Capital Southwest has also held preferred equity.

“The model isn’t the hard part. Design and leverage are usually the larger challenge points.”Michael Knight, EPIC iO chief technology officer, 2026

The company changed its mind about where AI belongs

The early EPIC iO pitch leaned heavily on an “open AI” platform and the promise of turning abundant IoT data into insight. The 2026 version is more restrained and more credible. In a detailed note on generative AI in security operations, chief technology officer Michael Knight wrote that AI should summarize events, prioritize risk, gather context and draft reports. It should not decide whom to confront, when to escalate or how to respond. Humans keep the consequential authority.

Knight also named the conditions no model can repair: blind spots, poor camera angles, bad lighting, blurry images and incomplete coverage. That is a meaningful correction to the usual “add AI” story. The software can reduce review time and alert fatigue, but system design and response procedure determine whether the alert changes an outcome. The company’s newer language emphasizes confidence levels, audit trails, human override and integration with equipment already in place.

Its corporate choices followed the same direction. EPIC iO folded IntelliSite and Broad Sky into one brand in 2022, bought Twenty20 Solutions in 2024, and installed Twenty20 leader Duncan Riley as chief executive that June. By 2026 the portfolio was framed around Cortex, a platform joining physical security, monitoring and connectivity rather than treating AI as a standalone attraction. The company also added veteran telecom executive Dan McDuffie to its board and demonstrated ZeroEyes threat detection on a mobile surveillance unit.

The playbook worth stealing

Most readers do not need EPIC iO’s entire catalog. They can still borrow its sequence. Start with the physical job and work backward. A camera is not a strategy; “detect a person inside this fence after 10 p.m. and notify this operator within 30 seconds” is. Then find the mundane point most likely to break.

Five moves to copy before buying anything

  1. Write the response outcome in plain language, including who acts and how quickly.
  2. Audit angles, lighting, power, bandwidth and blind spots at the actual site.
  3. Process obvious noise near the edge, then send compact events and context upstream.
  4. Design for carrier or circuit failure where downtime has real operational cost.
  5. Keep critical decisions with people and log both machine output and human action.

This is where EPIC iO fits in the market. It is not trying to train the world’s biggest foundation model or manufacture every camera. It occupies the messy middle between operational technology, physical security, telecom and managed service. Its expertise is orchestration: make old and new equipment interoperate, get useful data off the site, reduce the pile to a manageable queue and support the people who respond.

When the full stack is the wrong answer

The integrated approach is strongest when locations are numerous, remote, temporary or exposed to costly outages. It weakens when a customer has one well-connected building, simple recording needs and staff who already manage the system. A bundled managed platform can add cost and vendor dependence where an ordinary camera recorder and a local internet circuit are enough.

Good fit

Distributed sites, unreliable terrestrial links, moving risk zones, multiple carriers, mixed sensors, expensive truck rolls and a defined security or operations team.

Poor fit

One simple location, stable wiring, low consequence of downtime, no owner for alerts, weak privacy governance or a budget that cannot support ongoing managed service.

There are harder constraints too. Public-sector surveillance can stall on procurement, privacy and retention rules. Computer vision performance changes with weather and scene design. Cellular redundancy fails if every carrier is weak at the same location. Satellite brings geographic reach but not immunity from obstruction or service limits. And no centralized dashboard helps when an organization has not agreed on who receives an alert, what evidence is sufficient and when to call for help.

EPIC iO’s lasting idea is almost unfashionably practical: intelligence is a property of the whole system, not a feature pasted onto a camera. The company still sells plenty of AI, and its case studies still read like case studies. But its best advice is the caveat hidden inside the pitch. Fix the angle. Fix the signal. Fix the handoff. Then let the software earn its place.