Novaworks

Action across the customer's HR stack is claimed, but no outside system is named. The homepage makes the claim: "Our agentic platform connects people, workflows, and agents to enterprise systems with one intelligent layer that doesn't just recommend next steps, but takes them."

The product page adds the reach, automating HR transactions and approvals "across onboarding and offboarding, document management, benefits, payroll, and compliance", to "simplify complex workforce processes across your entire HR stack without replacing what already works", so the intent is to act in systems the customer already owns. Novaworks also cites "native access to AI control tower and integration hub" through the ServiceNow foundation, a real inherited route to connectors.

Not one outside system is named on the site, whether a payroll provider, an ITSM tool, an HRIS or an identity provider, and there is no connector count or directory. What is published is an action claim over an unnamed set plus an inherited hub, and the ServiceNow AI platform's integration estate belongs to ServiceNow rather than being a mechanism of Novaworks' own. The partners page or the FAQ may name systems.

Sourcenovaworks.ai homepage and /productread 2026-09-12

Specialized agents are assigned to HR domains and worked flows are published end to end, but no orchestration engine is named and customers cannot author the sequence. Specialized agents handle "onboarding and offboarding, document management, benefits, payroll, and compliance", alongside a Nova Super Agent giving "everyone in the organization their very own always-on concierge agent". One worked sequence is published end to end, "Policy gaps found. Amendments drafted.

Approvals routed", described as handling "the full compliance workflow from identifying what's missing to getting the right people in the room". In a second flow an employee submits a request and "Nova takes it from there" through instant policy guidance and automatic approval routing. The participants are distinct: the employee and their own agent, the specialized domain agents, and the approving manager.

What is not described is how it runs. No orchestration runtime is named, nothing says how the Super Agent hands off to a specialized agent or how a plan is formed, and there is no builder or other authoring surface, so the vendor defines the sequences for standard HR processes and the customer does not author them.

Sourcenovaworks.ai/product and homepageread 2026-09-12

The customer's own policies and configuration are what the agents read to find gaps and answer employees, but nothing says how that knowledge is held.

The published compliance flow opens with "policy gaps found", meaning the agent evaluates the customer's existing policy set to find what is missing, and the trust section describes "AI-assisted policy authoring that maps directly to your system configuration", so the customer's own configuration is an input the agent reads and writes against rather than a generic template.

Employees "get instant policy guidance and automatic approval routing, straight from their own Nova agent", a question answered against the organization's own policies rather than a model's general knowledge. The platform also unifies data across the workforce in one connected view.

There is no mechanism behind it: no retrieval structure, index or knowledge base is named, nor any connector to a document store, and no page says whether the policy corpus is kept between runs or assembled for each request. The site is two marketing pages carrying capability claims, from a vendor five months out of stealth with no documentation site.

Sourcenovaworks.ai/product and homepageread 2026-09-12

The agents prepare decisions and route them to a person to approve, but nothing documents a configurable hold over the agents' own actions. The homepage states the principle: "Every AI driven action is transparent, governed with a human-in-the-loop", with a "self-service, mobile-first model allowing people to stay in control".

The product page's trust section states the control: "Every decision our operating system enables you to make is transparent, auditable, and subject to human override, so your team stays in control at every step."

The worked flows put a named person in the path: "approvals routed" as a step in the compliance workflow, "automatic approval routing" on an employee request, and a manager step captioned "Approve in seconds", since "Nova handles the eligibility checks upfront so managers can act in one step."

So the agent prepares a decision and routes it to the person who makes it, which is good plumbing for keeping the decision with a person rather than a configurable hold over what the agent itself does. Nothing documents an approval queue the customer configures or a risk threshold, a checkpoint that can be placed or removed, or what happens when nobody responds. Human override is asserted as a property rather than shown as a mechanism.

Sourcenovaworks.ai/product and homepageread 2026-09-12

ServiceNow's access controls carry over, but Novaworks holds no security certification of its own. The vendor names the inherited model itself: "Novaworks is built natively on ServiceNow's enterprise security model", "zero-trust by default". A customer's ServiceNow access controls therefore govern the product. Novaworks also publishes four legal documents of its own, linked from the footer: a Privacy Policy, an Acceptable Use Policy, Terms of Service and a Data Processing Agreement.

Attestation cannot be inherited, and none is claimed: no SOC 2, ISO or other certificate is claimed as held by Novaworks, and there is no trust center or security page. The one sentence that gestures at it, "designed to meet the security and compliance standards your CISO demands", states intent rather than a held attestation. With no trust surface on the site, the DPA or the FAQ may say more.

Sourcenovaworks.ai/product, homepage and site footerread 2026-09-12

Every decision the agents enable is called transparent and auditable, but no log or audit trail is shown behind the claim. The product page's trust section states that "every decision our operating system enables you to make is transparent, auditable, and subject to human override", and the homepage that "every AI driven action is transparent". The vendor applies auditable to the agent's own decisions rather than the customer's HR records. The ServiceNow foundation is cited as giving "native access to AI control tower", a governance and monitoring surface a customer would already have.

No page names a run log, action history or audit trail product, nor a retention period or export. Nor is there a view an operator opens to see what the specialized agents did, such as which policy gap the compliance agent found or what the payroll agent changed, when and on whose authority. Transparent and auditable are asserted as properties of the design, and nothing shows the artifact. The site is four marketing pages and the vendor is five months out of stealth.

Sourcenovaworks.ai/product and homepageread 2026-09-12

An always on concierge agent is about availability, not retention, and no agent memory is described. Neither the homepage nor the product page mentions retained state, a preference store or decision traces carried between runs, or any lifetime or scope for what an agent holds. The Nova Super Agent is an "always-on concierge agent" for everyone in the organization, but an agent that is always reachable is not thereby one that remembers, and nothing says it carries context from one request to the next.

What the platform holds is the customer's records: a Total Workforce Management system keeping employees and contractors, interns and agents in "one connected view" is an application data model, the corpus the agents read rather than memory. The site says agents are managed alongside people; managing an agent's task assignments would be workforce administration, not the agent's own retained state. A thin marketing site is all there is to go on, so this records what the vendor publishes rather than settling what the product does.

Sourcenovaworks.ai/product and homepageread 2026-09-12

Novaworks publishes no region, residency or hosting information of its own; the product runs wherever the customer's ServiceNow instance runs. Neither the homepage, the product page nor the site navigation carries a region, data location or environment statement, and there is no trust center, security page or documentation surface where one would sit. Being built natively on ServiceNow is a real and arguably favorable position for a buyer who already has an instance, but the deployment estate is ServiceNow's, not Novaworks'.

Novaworks names no region list, no environment the customer controls and no way to choose, and does not make even a one sentence residency claim. A Data Processing Agreement is published at novaworks.ai/legal/dpa, and it is the most likely place for a residency commitment or a subprocessor region.

Sourcenovaworks.ai/product, homepage and site footerread 2026-09-12

None of the named HR agents is something a buyer can select, switch on or remove. The product page lists HR domains handled "by specialized AI agents", from onboarding and offboarding to payroll and compliance, plus a Nova Super Agent for each person. Those are named agent types describing how the platform carries out its own transactions, not units a buyer picks among. There is no catalog, library or marketplace, no template gallery or install path, and no page presents the agents as separable.

The two things a buyer can name, Nova Core HR and Nova Super Agent, are product modules rather than a set of complete agents to choose among, so nothing adoptable as a unit is documented. A catalog of the specialized agents may appear in later product documentation.

Sourcenovaworks.ai/product and site navigationread 2026-09-12

Nothing names a channel or a trigger for the agents. Neither the homepage nor the product page mentions Teams, Slack, email or SMS, or any portal, named mobile app or other messaging surface, and no event framework, schedule or webhook is described.

The Nova Super Agent is said to deliver a conversational experience "wherever and however your team works", and the homepage cites a "self-service, mobile-first model", but wherever and however names nothing, and mobile first is a design posture rather than a channel a buyer can count. The one thing that works as a trigger is internal: "automatic approval routing" moves a request to the right approver, which is workflow routing inside the product. Channels a customer's ServiceNow instance already carries belong to ServiceNow. The site has four pages, and its FAQ could add detail.

Sourcenovaworks.ai/product and homepageread 2026-09-12

No model or provider is named, and the customer has no choice of model. There is no routing description, model picker or admin setting, and no way to bring your own key. No LLM is named for the Super Agent, the specialized HR agents or the AI-assisted policy authoring, which is unusual even among thin sites.

Running on the ServiceNow AI platform means whatever model flexibility that platform offers is ServiceNow's rather than Novaworks'. The customer chooses nothing about the models running Novaworks' agents and is not told which provider is in use. For a vendor whose pitch is being AI native, naming no model anywhere is a visible choice not to disclose.

Sourcenovaworks.ai/product and homepageread 2026-09-12

Outside systems can reach Novaworks through the ServiceNow surfaces it is built on, but it publishes no API, SDK, webhook or MCP server of its own. The vendor names the inherited route on its own page, "native access to AI control tower and integration hub" through the ServiceNow foundation, describing itself as built natively on the Now platform and as extending what the customer has already invested in rather than replacing it. A customer already running ServiceNow therefore reaches Novaworks through surfaces ServiceNow publishes.

No API reference, endpoint list or SDK specific to Novaworks appears, nor any webhook or MCP server. The site navigation runs from Product and Partners to News and FAQs with no developer, docs or API entry, the footer carries only legal pages and social links, and no docs or developer subdomain is linked.

Nothing published lets an outside system call Novaworks itself as distinct from ServiceNow. The partners page names an implementation partner, Ateko, a services relationship rather than a programmatic one. A vendor this early may not have published a reference yet, and developer documentation after general availability would show more.

Sourcenovaworks.ai/product, /partners and site navigationread 2026-09-12

There is no published way to test or measure the agents. The homepage and product page show no test environment, sandbox or simulation, no versioning or regression check, and no evaluator or quality metric, and there is no documentation surface where one would be described. Every decision being "transparent, auditable, and subject to human override" is oversight and visibility, and neither measures the agent's behavior or puts a change under test.

Claims about saving time and boosting productivity describe outcomes for the customer's HR operation. There is not even a live metric on the agents' own output, such as a resolution rate or an accuracy figure. The site is four marketing pages from a vendor five months out of stealth, so public documentation could change the picture.

Sourcenovaworks.ai/product and homepageread 2026-09-12

HR transactions run inside the platform, with no browser or computer use. No hosted or local browser, desktop session, computer control or RPA appears on the site. The agents automate HR transactions and approvals, route approvals to managers, draft policy amendments and answer employee questions, all inside the platform and, where they reach outward, through the ServiceNow integration hub. An agent working a system through an interface that is not visual is not using a computer the way a person does.

How often Novaworks breaks when a target application changes its on screen layout is a meaningless question here, because nothing it does depends on what is drawn on a screen. The mobile first, self service model describes where employees reach their own agent, not an interface the agent drives, and the conversational Super Agent answers rather than navigates. This does not depend on the site being thin: the product executes workflows on a platform, and computer use would be a different product.

Sourcenovaworks.ai/product and homepageread 2026-09-12