Network automation provider BackBox is adding AI-powered analysis and recommendations to its network resilience platform with Kilter AI, an intelligence layer that lets enterprises apply AI across their network infrastructure while keeping humans in the loop, according to BackBox.
BackBox has renamed its network resilience platform Kilter AI and introduced AI capabilities that work across the platform’s lifecycle, configuration, and vulnerability management functions. The updates are designed to help network teams identify problems and use automation to remediate them to some degree.
BackBox says it is keeping humans in the loop. AI-generated automations cannot execute until an administrator reviews and approves them, according to the company. The approach is intended to address a problem BackBox sees among large enterprises: Network teams need more automation to manage increasingly complex, multi-vendor environments, but they aren’t necessarily ready to let AI make changes independently.
“We’re seeing customers that are getting more comfortable with automation that are afraid of AI,” says Rekha Shenoy, CEO of BackBox. One large global service provider, she notes, realized it had effectively given an AI system enough access to start making network changes on its own. “There’s no accountability, and that’s what scares them.”
Kilter AI is designed to operate more as what Shenoy calls a “smart intern,” analyzing large amounts of network and vulnerability data, identifying what requires attention, and recommending actions. Kilter AI then leaves execution decisions to the network engineers.
For instance, an enterprise might have hundreds of CVEs across network devices from multiple vendors running different software versions. Kilter AI can analyze vendor vulnerability information, determine which vulnerabilities apply to an enterprise’s devices, and identify whether a configuration workaround or patch is required, according to BackBox.
If a configuration change is needed, Kilter AI can offer to create an automation to perform the remediation. The automation is presented as a human-readable visual chain showing each step before it runs, Shenoy explains. A workflow might first back up a device, test its configuration, make a change, test again, and create another backup if the change succeeds. If the change fails, the automation could generate a ServiceNow ticket for further investigation, she says.
BackBox Kilter AI
Shenoy explains that administrators can review AI-generated automations and test them in non-production environments before approving them for production. Enterprises can also roll out changes in a “canary-style deployment” rather than across the entire environment at once. The recommendations draw on vendor information, BackBox’s library of more than 5,000 tested automations across 180 vendors, and activity in the enterprise environment, according to BackBox
According to Shenoy, that combination is important as network teams struggle to keep pace with patching and vulnerability management across diverse network infrastructure. Unlike server environments that might have large numbers of similar Windows or Linux systems, enterprise networks can include switches, routers, firewalls, and VPN infrastructure from many vendors, which are difficult to track and patch. And the growing infrastructure demands associated with AI add to that complexity.
“In the AI world, what we’re seeing is a ton more infrastructure buy,” Shenoy says. As enterprises add network devices to support AI-related bandwidth and connectivity requirements, they also increase the number of devices they must maintain and patch.
Shenoy says the company deliberately stops short of a “self-healing network,” which lets software make infrastructure changes. The goal is to automate time-consuming work while keeping network pros accountable for what happens in production. “It’s giving you the control; it’s not taking it away from you,” Shenoy says.
Kilter AI is available now as an upgrade for existing BackBox customers and as part of the Kilter platform for new customers.