Aomni

A short fixed list of connectors with no way to add your own, which is the published Partial marker for this axis. The vendor's integrations catalogue enumerates thirteen available third-party connections and closes there: CRM (HubSpot, Salesforce, Pipedrive), mail (Gmail, Outlook, SMTP), calendar (Google, Outlook), social (LinkedIn), research (Ocean.io, Exa) and enrichment (CrustData, Fullenrich), plus built-in web scraping, deep research and call recording.

Nothing on the page offers a custom connector, a generic HTTP action, or any user-defined tool a customer could wire up themselves, so a buyer whose tooling is not among the thirteen has no documented route. The threshold clause is met on the connectors that do exist, since CRM sync and sequencing write into outside systems, but the fixed catalogue holds this at Partial.

Sourceaomni.com/landing/integrationsread 2026-09-01

A linear chain with no documented error path, which is the published Partial case. The product runs a fixed three-stage sequence of research, strategise and engage: the agent assembles account and contact intelligence, generates a plan from it, then drafts the outreach. Waterfall enrichment queries multiple contact-data sources in one pass and the Enterprise tier adds bulk research across many accounts in parallel, but parallelism is throughput rather than control flow.

What the Full bar asks for is not documented anywhere on the product pages or pricing tiers: no branching or conditional logic, no retry or error handling, and no routing of work between steps or agents. Customers are said to build research workflows matching their own playbook, but no workflow builder surface, step reference or condition syntax was reached on either pass. Corrected from an initial grade of Full, which credited the pipeline without testing it against the control-flow bar.

Sourceaomni.com and aomni.com/pricingread 2026-09-01

Grounding on the customer's own material is one of four headline capabilities: the vendor states the AI is trained on your company data for tailored insights and that the agent is built to understand your ICP, differentiators and market landscape, with a configurable playbook carrying value propositions and selling methodology so output is framed the way the customer sells.

On the external side, research spans more than twenty sources and over a thousand data points per account, with Exa and Ocean.io as named retrieval connectors and built-in deep research. Confidence medium rather than high: this rests on product-page description, and no documented knowledge-base surface, ingestion interface or retrieval configuration was reached.

Sourceaomni.com and aomni.com/landing/integrationsread 2026-09-01

No shipped approval or guardrail mechanism is documented.

The agent researches and drafts, and the rep works from the output, but that is the default operating mode rather than a gate the vendor ships: no per-action approval step, no send-approval queue on the email and LinkedIn sequences, no permission scoping on agent actions and no configurable runtime guardrail was reached across the marketing site, integrations catalogue, pricing tiers or legal pages.

The abuse guardrails referenced in the commercial terms govern fair usage of the service, not agent behaviour. Graded as not documented as first-class rather than left unspecified, since both passes reached first-party material.

Sourceaomni.com, aomni.com/pricing and aomni.com/termsread 2026-09-01

This is the published Partial case almost exactly: a SOC 2 badge with nothing documented about who inside the buyer's organization can make an agent do what. The vendor states it is SOC 2 Type II compliant, with a footer badge linking to a live compliance portal at app.delve.co, and the privacy policy adds that reasonable security measures are implemented.

None of the four controls the Full bar asks for is published: no named access model, no SSO or SAML, no audit surface, and no compliance page beyond the badge and the portal link. The only customer-operated control reached is self-service deletion of account data on request. Resolves an Unspecified cell that the April build left ungraded.

Sourceaomni.com footer, app.delve.co/aomni and aomni.com/privacy-policyread 2026-09-01

No observability or audit surface is documented. Two passes reaching first-party material covered the marketing site and its four capability sections, the integrations catalogue, the pricing tiers and both legal pages; none describes run traces, execution logs, an activity or audit trail, or any record of what an agent retrieved and why it reached a conclusion.

The nearest is a notification feed of new deals and events, which is product output rather than a record of agent behaviour, and account research briefs, which are the artefact rather than the trace. Graded as not documented as first-class rather than left unspecified, since both passes reached the vendor and found nothing.

Sourceaomni.com, aomni.com/landing/integrations and aomni.com/pricingread 2026-09-01

Account intelligence persists between sessions and is addressable by account: research briefs, stakeholder profiles and account plans are stored per company and returned to rather than regenerated each time, and shared team account intelligence on the Enterprise tier makes that state readable across sales, marketing and customer success. What holds this below Full is that the persistent layer is authored or produced artefact rather than agent-written memory: the playbook carrying ICP, value propositions and differentiators is customer configuration, which is excluded from this axis, and no accumulation across runs, correction retained between sessions, or memory the agent writes to and reads back is documented.

Sourceaomni.com and aomni.com/pricingread 2026-09-01

Cloud-only with no customer control over data location. The privacy policy documents no region selection, no residency commitment, no sub-processor list and no disclosure, offering only that personal information is stored for a period consistent with business purposes, and the sole customer-side control is deletion on request by email.

No self-hosted, on-premises, VPC or private-tenancy option appears on the marketing site, the integrations catalogue or any pricing tier including Enterprise, whose differentiators are bulk research, CRM sync, annual pricing and onboarding. Confidence raised from the April grade because the absence is now visible in the vendor's own legal text rather than inferred.

Sourceaomni.com/privacy-policy and aomni.com/pricingread 2026-09-01

Ready-to-use capabilities ship on the free tier and are named as such: AI account research, AI sales strategy, AI contact dossier, AI chat, personalized AI sequences and custom deliverables, so a customer gets working agent output without configuring anything.

What was not reached is a catalogue: no browsable library of prebuilt agents, no template gallery and no importable starting configurations appear on the marketing site, the integrations page or the pricing tiers, and the customisation the vendor emphasises runs the other way, with reps building research workflows to match their own playbook. Partial: packaged capability without adoptable packaged assets. Downgraded from an April Full that carried no vendor evidence.

Sourceaomni.com/pricing and aomni.comread 2026-09-01

The agent is started by a person prompting it. The documented entry point is signing in and supplying a target company name or URL, and no schedule, webhook, inbound message, CRM update, inbox rule or monitoring alert is documented as starting a run. Calendar connections to Google and Outlook put meeting context in front of a rep and a notification feed surfaces new deals and events, but neither is documented as invoking the agent.

On the channel side the user meets the agent in the vendor's own console only: email and LinkedIn carry outbound sequences to prospects, which is the agent acting on an outside system rather than a channel where it meets the buyer, and the outbound webhook pushes data to the customer's applications rather than accepting work in. That is the published Partial marker, one surface with no event triggers.

Sourceaomni.com and aomni.com/landing/integrationsread 2026-09-01

The privacy policy states that the platform incorporates large language models provided by third parties, naming Google, Microsoft and Anthropic, and that data is shared with these models through their APIs for the tasks the customer instructs. Three providers named explicitly as model providers rather than as infrastructure suppliers, which clears the Partial floor.

What is absent is the other half: no model selector, no per-agent or per-workflow model choice, no bring-your-own-key and no routing policy the customer configures is documented anywhere on the marketing site, integrations catalogue or pricing tiers. Provider plurality without customer selection is Partial, not Full. Upgraded from an April grade that carried no vendor evidence.

Sourceaomni.com/privacy-policyread 2026-09-01

One outbound extensibility surface is shipped: the integrations catalogue lists Webhook as Available, described as sending real-time data to the customer's applications when specific events occur. The inbound half is explicitly unshipped on the same page, where HTTP API and MCP are both marked Coming Soon, so the platform is not callable from outside today.

This corrects a third-party claim: several review sites report API access on the Enterprise tier, but the vendor's own integrations page is the better source and states the API is not yet available. Partial on the shipped webhook, held below Full under the standing rule that Full requires the platform to be callable from outside through a documented API or SDK, and under roadmap-is-not-shipped.

Sourceaomni.com/landing/integrationsread 2026-09-01

Nothing the customer can test, debug or optimise is documented. No evaluation surface, scored run, A/B or sequence performance comparison, prompt or playbook testing harness, or quality measurement of research output was reached on either pass.

The vendor publishes outcome claims of roughly three hours saved per prospect and a 40 percent close-rate lift, but these are marketing assertions about the product rather than a measurement the customer runs, and are excluded under the convention that Eval grades what the customer can test rather than what the vendor reports. Graded as not documented as first-class rather than left unspecified.

Sourceaomni.com and aomni.com/pricingread 2026-09-01

The vendor's integrations catalogue lists Aomni Web Scraping as a built-in capability gathering intelligence from public sources, alongside Aomni Deep Research and the Exa search API, so the agent reaches pages it has no interface to. Web scraping is the published Partial case for this axis. The Full case is not documented: no hosted or local browser the platform drives itself, no desktop session and no remote computer control appears on any product page, and the vendor describes the mechanism as scraping rather than browser operation.

Recorded because the surrounding sources disagree with the vendor: an internal pricing dossier and several directory listings describe research running in real browser sessions on Browserbase and Playwright, which would be the Full case, but no first-party page makes that claim and directory listings are not evidence of capability. Corrected from an initial grade of none, which read scraping as zero.

Sourceaomni.com/landing/integrationsread 2026-09-01