SolarWinds Launches Avior as AI Agent Startups Draw Fresh Funding

The launch addresses a reliability problem SolarWinds says comes with letting AI agents act on production systems: an agent that produces different outcomes from one run to the next is a liability where uptime and security are at stake. Chief Technology Officer Krishna Sai said Avior grounds "probabilistic models in deterministic execution," with predefined goals and approval controls governing how work gets done and every workflow auditable.

Avior runs as its own application on the SolarWinds Platform and reaches third-party systems through the Model Context Protocol, according to SiliconANGLE. No data migration is required for it to work across self-hosted and cloud environments. At launch it can merge related alerts into one incident, detect configuration changes made outside normal channels and restore the last known good configuration, and retire stale inventory records for devices no longer on the network.

Autonomy is tiered by risk. Actions SolarWinds classifies as potentially dangerous wait for explicit human approval, each action runs with the permissions of the person who started it, and rollback support is built in. Only data relevant to a request is sent, customer data is masked, and it is not used to train underlying models, the company said. Avior is available through a guided onboarding program; pricing was not disclosed. Cullen Childress, SolarWinds interim CEO and chief product officer, said "reaction time alone is no longer enough" as IT environments grow more complex.

Harmonic Security Inc. said it raised $50 million in an oversubscribed Series B round led by Crosspoint Capital Partners, with existing investors participating. The round brings total funding to $76 million, after a $17.5 million Series A led by Next47 in October 2024. Harmonic's platform inspects what employees send to AI tools before it arrives, using custom small language models to read each interaction in context and return a verdict within milliseconds. It was first built for browsers and has been extended over the past year to desktop AI applications and agentic workflows.

Harmonic also offers a Usage Intelligence layer that shows which AI use cases are delivering value, down to individual employees, according to SiliconANGLE. It has integrations with Cisco Systems Inc.'s Agent Builder through the Model Context Protocol and with Anthropic PBC's Claude Compliance API. More than 100 companies are customers, including some of the world's largest enterprises, and millions of their employees are covered. Staff numbers have passed 90 across the U.S. and U.K. Co-founder and CEO Alastair Paterson said the company has "real momentum."

Uvi, officially Naboo Research Ltd., raised $7.5 million in seed funding led by CT Innovations, with Cardumen, Mana VC and 91 Ventures participating. The startup is selling a context reasoning layer that gives AI agents a shared institutional memory instead of forcing each new agent to relearn how a business operates. It links systems such as Slack, Jira, Confluence and Salesforce with code repositories and meeting minutes, then builds a knowledge graph with entity mapping and a semantic layer for company jargon. Uvi says access controls and data tagging carry permissions across platforms.

Uvi co-founder and CEO Gilad Salinger said the problem is structural: large companies run dozens of tools with different data shapes, and an agent that completes a task may fail the same task next time because it never retained how the work gets done. Uvi's benchmark tests show internally built agents cost the average company about eight times more than they should, with agents that run without its layer costing between $0.75 and $1.41 per response and sometimes more than $2, according to SiliconANGLE. Uvi says routing responses through its layer lowers that cost.

Evos, officially Evolutionary Operating System Inc., raised $9 million in pre-seed funding led by Uncork Capital and Field Ventures, with Crosslink Capital, Utopia Capital Management, Sandwith Ventures and angel investors participating. Its Workhorse platform creates autonomous operators for physical-economy businesses in manufacturing, freight and logistics. The agents begin by watching and learning from experienced workers behind an approval gate; humans review, approve, modify or reject their decisions. Founder and CEO Urav Shah said the average agent reaches roughly 82% decision accuracy after 24 hours of learning and can then handle about 70% of a human's daily workload autonomously.

MyFreightWorld is using Evos agents to audit less-than-truckload carrier invoices against its transportation management system. In the first two months, it made more than 3,500 auditing decisions with 97% accuracy and helped process more than $600,000 in invoice value, according to SiliconANGLE. Evos went live with select customers a month ago and has reached more than $1 million in annual recurring revenue. Shah said the company plans to make Workhorse generally available later this year.

CodeWall Ltd., a London-based offensive security startup, launched with $4 million in pre-seed funding led by Osney Capital, with InReach Ventures, Antler Global and angel investors. Its autonomous AI agents attack enterprises the way an outside hacker would, starting with no credentials or inside knowledge. Several agents work at once, logging findings to a common investigation graph so a weak spot in one application can become another agent's route into connected systems. Founder and CEO Paul Price calls the ongoing testing "a permanent capability" and pitches it as a replacement for annual penetration tests.

Earlier this year, CodeWall turned its agents on the Big Three management consulting firms. At McKinsey & Co. Inc., an agent with no credentials gained read and write access to the production database behind the firm's Lilli AI platform within two hours, according to SiliconANGLE. Research published in the weeks that followed found exposed databases at Boston Consulting Group Inc. and Bain & Co. Inc. Each firm fixed its flaws within days of being told. A McKinsey spokesperson told The Register that a third-party forensics investigation found no evidence that client data had been accessed by CodeWall or any other unauthorized party. Over the summer, Sysdig Inc. documented what it said was the first ransomware operation run from start to finish by an autonomous AI agent.

The product and funding announcements came as industry observers debated how enterprises should manage agents. A TechRadar Pro article argued that AI agents require an Agent Development LifeCycle, borrowing from the traditional software development life cycle and from human employee management, because probabilistic actors must be onboarded, given scoped access, trained and governed differently. It recommended treating agent setup like drafting a job description, configuring guardrails and organizing specialized agents in an agent mesh with role-based access controls.

Sifted reported that AI is beginning to shift finance from reactive work to proactive systems, pointing to Intuit Intelligence rolling out across QuickBooks and Intuit Accountant Suite. Intuit's AI on Autopilot turns conversational descriptions of recurring tasks into automations, while Autonomous Invoicing creates invoices and handles reminders and updates. Intuit QuickBooks' impact report found regular AI use among UK businesses rose to 70%, but trust rather than cost is now the biggest barrier to deeper adoption. Intuit research surveying 1,003 UK accountancy students and early-career finance professionals found 90% believe AI will help them advance faster, while more than a third expect to spend more time on higher-value tasks earlier in their careers.