Albert

Deep coverage inside one ecosystem class. The FAQ states Albert is integrated with Google's search and programmatic channels plus Facebook, Instagram, YouTube and Bing, reaching what the vendor calls 90 percent of biddable inventory, that its programmatic campaigns are managed through an integration with Google Marketing Platform, and that Gmail is its only non-paid component; Zoomd's product page adds TikTok. Albert operates inside the advertiser's own existing ad accounts rather than reselling inventory.

Every named connection is an advertising platform: no CRM, no analytics or data warehouse, no marketing automation, no commerce platform and no offline conversion source appears anywhere on the property, and the vendor states plainly that Albert holds no company data beyond what the ad platforms expose, which forecloses the whole class. The FAQ also confirms first party the limitation on programmatic, which runs through Google Marketing Platform. Zoomd's 600 plus media sources belong to the parent's own user acquisition platform, not to Albert.

Sourcealbert.ai/faq and zoomd.com/albert-airead 2026-09-02

The whole optimization cycle, executed rather than recommended, as one fixed product pipeline.

The FAQ says Albert takes action rather than leaving it to the marketer, making adjustments every moment of every day and managing budgets in response to market conditions against the goal set for it: across search it adjusts bids on audiences, schedule, devices and ads; in paid social it optimizes audiences, ad sets and lookalikes; in programmatic it works real time engagement and lookalike modeling, with campaign creation, creative rotation, keyword management and reporting around that.

The buyer sets goals, guardrails and creative, and no flow the buyer designs or changes, no branching or conditions the buyer configures and no set of coordinated agents is documented. However many steps it runs, the pipeline is fixed. The ROAS lifts on the site are vendor reported.

Sourcealbert.ai/faqread 2026-10-01

Grounding that is real, bounded on purpose, and holds only what the ad platforms expose. The FAQ states that Albert ingests data from brand sources, creative sources and channel or platform sources, combines it against adjustable KPIs, and reasons over accumulated campaign performance, audience and interest behavior, creative variant results from continuous multivariate testing, placements and keywords.

The ceiling is a stated design choice rather than an oversight: asked whether it holds and analyzes the company's data, the vendor answers no, that Albert is exposed only to data available from the platforms and not to personal or sensitive information from internal company sources. That is a defensible architecture for an agent with write access to ad spend. What is absent follows from it: no document corpus, no retrieval layer, no customer knowledge base, no brand guideline object and no ingestion of first-party data beyond the ad platforms.

Sourcealbert.ai/faqread 2026-09-02

A configured boundary the customer sets, and no gate on anything the agent then does. The FAQ is explicit about both the boundary and the division of labor. On the boundary, Albert optimizes against the business goal you set, all within the guardrails you specify, and after being configured and preset with guardrails and goals it works autonomously.

The division of labor: clients set strategy, tactics, goals and the overall scope of campaigns, and supply the creative, with Albert restricted to mixing and shifting budget across approved ad materials and forbidden from writing its own copy. Those are real constraints on what the agent may do and they are the vendor's own mechanism.

What is never documented is the rest: nothing states what the guardrails can actually constrain, how they are configured, whether spend caps or escalation thresholds exist, or what happens when one is hit, and there is no approval step, review queue or hold before Albert executes, which the vendor presents as the point of the product. For an agent autonomously spending six and seven figure budgets that documentation gap is the thing a buyer most needs and cannot get.

Sourcealbert.ai/faq and zoomd.com/albert-airead 2026-09-02

No security, identity or compliance information is published. Beyond its marketing pages, FAQs, privacy policy and a client login at console.albert.ai, the site has no security page, no trust center, no compliance page and no documentation site.

The privacy policy is the unedited WordPress template, last modified March 2021 and written about the marketing website rather than the product, and its headings for how we protect your data, what data breach procedures we have in place, who we share your data with and what automated decision making and profiling we do are present and empty.

No attestation, certification, named auditor, penetration test, sub-processor list, DPA, single sign-on, role-based access control, audit log, encryption statement or retention control appears anywhere.

Three nearby facts are not security controls either: data minimization, that Albert sees only advertising-platform data and no personal or sensitive information from internal company sources, which is a scope property confirmed first-party in the FAQ but neither a control surface nor a certification; Zoomd's uptime commitment, which is availability; and the parent's TSXV listing, which is corporate governance.

Sourcealbert.ai/privacy-policy, albert.ai/faq and the albert.ai site navigationread 2026-09-02

Campaign reporting, and no trail of what the agent did or why. What Albert documents is performance reporting and insight: the FAQ describes machine-level reporting on how interests perform and actionable insights from placements, keywords, creative items and audiences, and lists looking at campaign insights among the human interactions that matter. That measures how the campaigns performed, which is a different thing from auditing what the agent did.

Nothing documents a run history, an action log, a per-change record of which bid moved when, or any account of why a budget shifted or a creative was rotated out. Albert's own reporting is reporting rather than auditing. Because Albert operates inside the advertiser's own Google, Meta and Bing accounts, those platforms' native change history records what it did, but that record comes from the customer's own systems rather than from Albert. For an agent with write access to live budgets the what is recoverable from the ad platforms and the why is not recoverable at all.

Sourcealbert.ai/faq and zoomd.com/albert-airead 2026-09-02

The accumulation is real, and it is model improvement rather than memory. The FAQ states that Albert is designed to learn, evolve and improve performance over time as learnings continue to accumulate, and that it runs continuous multivariate testing and calculations, and Zoomd describes it as self-learning rather than rule-based.

Learning absorbed into the optimization model is not a memory layer, and a store of data is not a memory layer whoever writes it, so none of that is agent-written state the agent reads back. What is documented is statistical learning over campaign data rather than agent memory, with no cross-session recall, no memory object exposed to the buyer, no way to inspect or correct what the system believes it has learned, and no retention or forgetting control. The learned state directly governs spend and cannot be examined.

Sourcealbert.ai/faq and zoomd.com/albert-airead 2026-09-02

No deployment or residency control of any kind is published. The FAQ describes Albert only as a cloud based platform that plugs into the marketer's existing stack, and the site contains no hosting, region, residency, tenancy or infrastructure statement at all.

No region selection, on-premises, air-gapped, private VPC or single-tenant option is offered, and the privacy policy that would normally carry a processing-location clause is the unedited WordPress template with those sections blank.

Jurisdiction is a separate matter from residency: Albert originates as Adgorithms of Herzliya, is described in the acquisition release as a US-based platform, and the parent Zoomd Technologies is listed on the TSXV.

Sourcealbert.ai/faq, albert.ai/privacy-policy and the albert.ai site navigationread 2026-09-02

A single agent with no catalog behind it. Nothing on the property offers named prebuilt agents with distinct roles, a template library, a playbook or blueprint system, a marketplace, or anything a customer browses, selects, instantiates or clones.

The key point is that campaign structures are generated rather than adopted: Albert builds ad groups, audience sets and budget allocations from the KPIs it is given and the creative the client supplies, each time, which the FAQ frames as the client setting strategy, tactics, goals and scope while Albert handles the computational work. That is a legitimate architecture, just not a catalog of prebuilt agents.

Sourcealbert.ai/faq and zoomd.com/albert-airead 2026-09-02

Four distinct buyer-facing channels, acted on simultaneously, with continuous rather than scheduled intake. The FAQ names search across Google and Bing, social across Facebook and Instagram, video on YouTube and programmatic through Google Marketing Platform, with Zoomd's product page adding TikTok, and states the combination reaches what the vendor calls 90 percent of biddable inventory.

What distinguishes this from platform-native agents confined to their own inventory is that Albert shifts budget between those channels rather than optimizing within one. Triggering is continuous: the vendor describes adjustments every moment of every day without delay, driven by real-time performance, engagement and conversion data rather than a batch cycle or a schedule someone sets. Third party 2026 commentary claims Albert ingests weather, competitor spend, inventory levels and social sentiment, but no first party page confirms it.

Sourcealbert.ai/faq and zoomd.com/albert-airead 2026-09-02

No model provider named, no selection, no routing, no bring-your-own-model and no bring-your-own-key anywhere on the property. One distinction matters to a buyer.

Albert originates as Adgorithms, founded 2010 in Herzliya, and its core is proprietary optimization and machine learning built in house over fifteen years rather than a wrapper around a frontier language model; the vendor describes sophisticated algorithms of its own and self-learning behavior contrasted with preset rules, and the value is in the optimization engine itself.

That is deliberate vertical integration which is the product, rather than an undisclosed third-party dependency left unanswered. On the surface it looks like a vendor hiding its model provider, but for a buyer it means the opposite.

Sourcealbert.ai/faq and albert.ai/newsread 2026-09-02

The platform cannot be called from outside. Albert's whole published property is seven navigation pages, a FAQ, a privacy policy and a client login at console.albert.ai. There is no developer portal, no API or docs subdomain, no SDK, no webhook documentation, no MCP server and no partner or certification program.

That is coherent with how it is sold rather than in tension with it: the FAQ describes a dedicated Customer Success team providing implementation, account management and support as included services, and the interface is English only, which is a managed enterprise product consumed through its own console rather than a platform others build on.

Searching for an Albert API surfaces a different product: the French government's sovereign AI gateway, also called Albert, which publishes an OpenAI compatible API with EU data sovereignty guarantees and has nothing to do with this vendor's API, model or deployment position. Nor is help.albert.com this vendor; it is the help center of an unrelated consumer finance company.

Sourcealbert.ai/faq and the albert.ai site navigationread 2026-09-02

A strong experimentation engine pointed at the ads, not at the agent. The FAQ describes continuous multivariate testing and calculations running at machine speed, with actionable insights from placements, keywords, creative items and audiences, and creative rotated on measured performance. An engine that tests the output is not evaluation of the agent that produced it.

What a buyer cannot do is the consequential part here: there is no sandbox to run Albert against historical spend before it touches a live account, no way to score or replay its decisions, no holdout that measures the agent rather than the campaigns, and no regression test after changing objectives or guardrails. The vendor's own answer on measuring Albert's value is a three-month proof of concept comparing KPIs against the same period last year, which is a services engagement rather than a shipped testing surface. For an agent that spends real money autonomously the absence of a dry run compounds the missing guardrail documentation.

Sourcealbert.ai/faqread 2026-09-02

One first party claim points toward computer use, and the FAQ shows why it does not hold. Albert's company description states it combines robotic process automation with AI to manage all aspects of paid digital campaigns.

The FAQ describes the actual mechanism as integration: Albert is integrated with Google's search and programmatic channels, Facebook, Instagram, YouTube and Bing, and its programmatic campaigns are managed through an integration with Google Marketing Platform. Those platforms all publish comprehensive advertising APIs, so driving them through a user interface would be a choice rather than a necessity.

No browser control, headless session or screen operation is documented anywhere. The RPA framing reads as legacy positioning from the platform's pre-language-model era.

Sourcealbert.ai/faqread 2026-09-02