Kana

The agent acts in systems the vendor does not own. Kana's FAQ states its agents analyze data, adjust ad bids, optimize audiences and protect brand visibility autonomously, and that the platform plans multi-step workflows and executes them across the customer's tech stack.

Reach is supplied by a proprietary Just In Time Data Integration that connects to existing marketing and media systems without infrastructure changes, with agents calling out to ad platforms, analytics, CDPs and data providers and reasoning over the responses.

Adjusting a live bid is authenticated action rather than a recommendation handed back, and the classes span advertising, analytics, customer data and third-party data.

Sourcekana.ai/faq and kana.airead 2026-09-01

Kana intelligently orchestrates highly aligned, loosely coupled AI agents assigned to distinct jobs, from refining a brief to identifying audiences, planning channel mix, and running autonomous tracking and optimization, working simultaneously per TechCrunch and kana.ai.

Sourcekana.airead 2026-10-01

A maintained structure over the customer's own data, and the vendor's positioning is that it stays where it is.

Kana structures campaign performance, customer insight and business context into a layer the agents reason over, grounding them in the customer's first-party data and CDP and combining external market signals with proprietary data in one workspace, with synthetic data generation filling gaps in third-party coverage.

The Just In Time Data Integration reaches that material in place rather than requiring a migration, and the layer persists and is queried as campaigns run rather than assembled per prompt.

Sourcekana.ai/solutions and kana.ai/faqread 2026-09-01

A shipped approval gate plus configured boundaries, and the vendor states the limit plainly. Kana's augmented intelligence design has marketers review, approve, edit, coach and reroute agent work in the flow, and its own framing is that agents move fast without touching budgets or going live unapproved, so spend changes and live launches stop for a person.

Role-based controls set what agents may do for brand safety and policy compliance, which bounds behavior rather than only bounding access, and the FAQ describes agents acting autonomously while the marketing team supervises. That is a mechanism the vendor ships rather than a division of labor it describes.

Sourcekana.ai/faq and kana.airead 2026-09-01

The FAQ states SOC 2 compliance and the home page displays SOC 2 Type II and HIPAA badges; a privacy consultant's CIPP/E and CIPM credentials are an individual's, not the company's. No SSO, roles or permissions the buyer operates are documented.

Sourcekana.ai/faqread 2026-10-01

Campaign reporting rather than an agent record. What Kana supplies is tracking and reporting on campaign performance plus outputs a marketer reviews before approving, both of which describe the work product rather than the agent's execution. No run log, action trail, decision trace or record of what an agent did and why appears on the homepage, the solutions pages or the FAQ.

Sourcekana.ai/solutions and kana.ai/faqread 2026-09-01

The vendor's own thesis is accumulation: it describes an operating layer built to learn so a marketer's judgment and decisions compound instead of being rebuilt from scratch each time, and agents adapt as a campaign evolves, which means state carries between runs. No memory feature is named, with no scope or lifetime, and the campaign, customer and business context that persists is the knowledge layer.

Sourcekana.ai and kana.ai/blogread 2026-09-01

A disclosed single environment with no choice: the FAQ says services are hosted in Kana's own environment on Google Cloud Platform, with a just in time integration approach that reads the customer's data where it sits, and no region choice, private tenancy or self hosting is offered.

Sourcekana.ai/faqread 2026-10-01

Named solution areas, delivered bespoke rather than adopted from a library. Kana's solutions page covers data generation, campaign strategy validation and optimization, an insight layer combining external and proprietary data, answer engine presence, and churn prevention, and the platform composes reusable configurable domain-specific agents quickly. Applications are custom-built to the client's goals and architecture, co-developed with Kana's team, and the solutions page presents the list as a sample of what has been delivered for customers rather than a catalog to switch on.

Sourcekana.ai/solutions and kana.ai/partnersread 2026-09-01

Work reaches the agents without a person starting it: Kana's agents track campaigns continuously and adjust bids, audiences and budgets as performance data arrives, and the platform identifies patterns, risks and behavioral signals in real time.

Sourcekana.ai/faqread 2026-10-01

Documented vendor-internal routing. The vendor's FAQ states the platform uses a model-agnostic architecture with logical model roles such as planning, evaluation and summarization, and names the underlying providers as Anthropic, OpenAI and Google, accessed over their official APIs with no customer data persisted by them under contract. It also describes a process for assessing new models as they are released. That is routing across three providers by role, decided by Kana. Nothing documents a customer or admin selecting a model, bringing a key or pinning a provider.

Sourcekana.ai/faqread 2026-09-01

Nothing callable from outside is documented. No developer, documentation, API or status entry appears across the site, whose surfaces run to the homepage, solutions, partners, blog and FAQ, and there is no docs subdomain. The partners page describes Kana co-developing bespoke agentic applications alongside the client, which is the vendor's people rather than a surface the customer extends.

Sourcekana.ai/partners and the kana.ai site mapread 2026-09-01

A real experimentation engine, pointed at the media rather than at the agent. Kana's optimization agents run experiments and shift budget as performance data arrives, which is a mechanism with a criterion and a controlled comparison, and synthetic data generation fills gaps in third-party data so a marketer can test scenarios before committing.

Nothing compares one agent configuration against another, no holdout runs on agent behavior and no scored verdict on the agent is produced; shadow mode rollouts are described only as expected, the evaluation named in the FAQ is an internal model role rather than a customer surface, no failure log, run inspection or error surface is offered, and coaching lets a marketer correct an agent but returns no readable verdict.

Sourcekana.ai/solutions and kana.ai/faqread 2026-09-01

No browser driving or computer use capability; Kana operates as an orchestration layer over marketing systems rather than a browser or GUI.

Sourcekana.airead 2026-10-01