SSelector AI

Selector’s annual recurring revenue reached 230% of its 2024 level in 2025, the fourth year in a row it more than doubled, and investors responded with a $32 million round at a $375 million valuation in February 2026. Its product is an agentic layer for network operations: agents that collapse alert storms into a single incident, trace root cause and blast radius, open the ticket and stage the fix. It suits large telecom operators and Fortune 1000 network teams. Pricing is shared by quote, and public user ratings are still few.

Telecom investors know it well. AT&T Ventures, Bell Ventures, Comcast Ventures and Singtel Innov8 all back the company, and Comcast and Bell engineers appear in its testimonials. We rate it 8.0/10 on our scale, a score that mostly reflects pricing transparency. We have not used Selector ourselves; this review is based on Selector’s product pages and blog, its funding announcements and Gartner Peer Insights ratings.

Selector’s Growth, Funding and Operator Investors

Selector publishes growth and funding figures rather than customer case metrics, and its investor list reads like a roster of large network operators.

The investor list is the strongest signal for a telecom buyer. When the venture arms of AT&T, Bell, Comcast and Singtel invest, it usually means their network teams have looked closely at the product. Bell and Comcast engineers also appear by name in Selector’s testimonials.

Outcome data is the area Selector could build out. Selector reports company growth in percentages, and publishing named customer metrics such as hours saved or tickets avoided would help buyers; Gartner Peer Insights holds seven ratings so far. Expect to build your own business case during a proof of concept.

The growth numbers deserve a careful read. ARR at 230% of the prior year is a strong signal of demand, but it is a relative figure from a private company, without a base amount. Treat it as evidence that large buyers are signing, not as a measure of how well the agents work in any one network.

Data to Correlation to Agents to Approved Action

Selector is a network intelligence platform with an agentic layer on top. The flow, as Selector describes it, runs from data to correlation to agents to approved action.

Collecting and Correlating Data

Selector pulls metrics, logs, configurations and flows from more than 300 telemetry sources across network, cloud and edge, without installing agents, and normalizes them into one data model. Selector’s blog names the layers: a Data Hypervisor that cleans and enriches telemetry, a Knowledge Service that adds context through real-time correlation, and LLMs with retrieval that produce answers. Purpose-built models trained on network telemetry look for cause and impact, and cross-domain correlation ties network, cloud and application data into a live knowledge graph.

The Foundry Agents

• Rosetta: the conversational interface. It interprets what an operator wants, calls the right specialist agents and returns one answer with evidence attached. It can start from an event or from an operator’s question.
• Loom: a correlation agent that turns an alert storm into a single incident with root cause, timeline, blast radius and evidence.
• Artist: builds live network dashboards from a plain-language request, with no query syntax.
• Sculptor: lets teams create new agents, including by describing them in plain language, with a Git repository for building, versioning and governing them.

Selector published posts on Rosetta, Artist and Loom in October 2026, presenting them as parts of Foundry.

Alert Filtering, Ticket Routing and Dashboards on Request

According to Selector, the agents filter alert noise, link related incidents, open tickets and route them to owners, build dashboards on request and deliver recommendations in real time. Customer testimonials mention background agents and auto-incidents. When a fix is needed, an agent stages it, Selector’s demo example being a rollback of a configuration push, and waits.

Nothing Runs on the Network Without Approval

Nothing executes until a user approves it. Agents act inside user-defined guardrails and existing role-based access controls, every answer links to its evidence and steps, and actions leave an auditable trail. Selector’s May 2025 post describes its framework as already supporting automated actions and ITSM workflow integration while evolving toward self-healing, and recommends introducing agents with human-in-the-loop guardrails.

Regional Support and Container Deployment Options

Selector’s Japanese and cloud marketplace expansion, announced with its 2026 funding, matters for global operators that want regional support and procurement through an existing cloud account.

Selector supports containerized and hybrid cloud deployments and runs on the LLM of your choice. It connects to Slack, Microsoft Teams and ITSM tools, works beside existing tools such as Splunk, NetBox and ThousandEyes, and links to custom tools and agents through MCP and A2A. Buyers start with a demo, and one Gartner reviewer reports that deployment took longer than expected.

Selector Contracts Follow a Demo, With No List Price

Selector does not publish prices. Contracts are negotiated after a demo, and the company says it is adding cloud marketplace and regional partner channels.

Sizing the Alert Work at an Illustrative NOC

With no list prices, the useful exercise is sizing the work the agents target. The volumes below are our own illustrative assumptions, using Selector’s claim that up to 90% of repair time goes to finding where an issue started.

• A regional operator NOC: 400 incidents a month at three hours each is 1,200 engineer hours. If up to 90% of that is spent locating the origin, as much as 1,080 hours a month is the pool correlation agents aim at.
• An alert storm: a fiber cut that triggers 2,000 alerts across routers, links and applications is exactly what Loom is built to fold into one incident with a blast radius.
• Dashboard requests: if network engineers spend a day a week building views for other teams, Artist targets that time with plain-language requests.

Costs to plan for beyond the license: the infrastructure for containerized or hybrid deployment, LLM usage if you bring your own model, staff time to tune correlation and write agents, and a proof of concept long enough to measure results on your own incidents.

Selector AI Compared With Kentik AI Advisor, Nokia Autonomous Network Fabric and Resolve AI

Selector sits between network observability tools and broader incident agents. These are the closest options listed on our site.

Selector goes further toward action than Kentik, whose agent is read-only, and stays more vendor-neutral than Nokia. Resolve AI covers similar incident work for software production systems rather than carrier networks. For more options see our AI agents for telecom and IT and DevOps guides.

Large Telecom and Cable Operators

• Large telecom and cable operators: the investor list and named testimonials from Comcast and Bell suggest the product is built for networks at that scale.
• NOCs drowning in alerts: Loom is designed to turn alert storms into one incident with root cause and blast radius.
• Multi-vendor, multi-domain estates: 300+ telemetry sources and cross-domain correlation suit networks that mix vendors, clouds and applications.
• Teams ready for supervised action: staged fixes with approval and audit trails give a path to automation without unattended changes.

Small ISPs and Buyers Needing a List Price

• Small ISPs and single-site networks: the platform is sold to Fortune 1000 organizations on negotiated contracts.
• Buyers who need a list price before a call: Selector shares pricing by quote; Kentik publishes a price.
• Teams that want self-healing with no approval step: Selector describes that as a direction, not today’s default.
• Operators whose need is billing or customer care: BSS-focused platforms such as Amdocs aOS or Totogi BSS Magic fit better.

Peer Reviews, Reporting and Deployment Planning

• Public reviews: seven Gartner Peer Insights ratings is a small sample so far for a platform at this price level.
• Built-in reporting: some reviewers say built-in reports are limited and most must be created by users, an area Selector could expand.
• Query phrasing: one reviewer says natural-language questions must be phrased in a specific way.
• Deployment time: one reviewer found implementation took longer than expected, so plan the proof of concept with slack.
• Customer metrics: Selector’s product pages describe lower time to repair; publishing named customer metrics would make that easier to verify.
• Roadmap language: self-healing examples in Selector’s blog are framed as scenarios, so confirm which actions are available in your contract.

AVP’s $32 Million Round and Entry Into Japan

• On February 18, 2026, Selector raised $32 million led by AVP at a $375 million valuation and said it had entered the Japanese market.
• On May 23, 2025, Selector described its path from visibility to autonomous operations, with agents that already take automated actions and integrate with ITSM workflows.

Convincing Triage Agents, Scored at 5.9

For large operators that want agents to do real correlation and triage work with an engineer approving each change, Selector is one of the more convincing options. It scores 8.0/10 on our scale: strong on autonomy and integrations, with the score reflecting quote-only pricing, a demo-led setup and a small set of public ratings. The backing of AT&T, Bell, Comcast and Singtel venture arms, and ARR that more than doubled for a fourth year, suggest operators are buying.

Smaller teams and anyone who needs a price up front may find other tools a better fit. If you shortlist Selector, run a proof of concept on a real alert storm and measure how often Loom’s single incident matches your engineers’ own root cause. Ask which staged actions are available at signing, how approvals map to your change management process and how the platform handles your own LLM if you bring one. Those answers matter more than the funding headlines.

Doubts Network Teams Bring to a Selector Demo

How Much Does Selector AI Cost?

Selector does not publish prices. It sells on negotiated contracts after a demo and is adding cloud marketplace and regional partner channels.

Do Selector’s Agents Make Changes Without Approval?

No. Selector says nothing executes until a user approves it. Agents can stage a fix such as a configuration rollback, open and route tickets and build dashboards, and every action leaves an audit trail.

What Are Rosetta, Loom and Artist?

They are agents in Selector’s Foundry layer. Rosetta is the conversational interface that calls other agents, Loom folds alert storms into one incident with root cause and blast radius, and Artist builds dashboards from plain-language requests.

How Does Selector AI Compare With Kentik AI Advisor?

Both correlate network data to find root causes. Kentik’s agent is read-only and comes with a published platform price from $2,000 a month billed annually. Selector’s agents can stage fixes for approval, and it is sold on custom quotes.

Who Invests in Selector?

Its February 2026 round was led by AVP. Earlier backers include Ansa Capital, Two Bear Capital, Atlantic Bridge, Sinewave Ventures, AT&T Ventures, Bell Ventures, Comcast Ventures, Singtel Innov8 and Hyperlink Ventures.