Launch of Ad Agency Operation Automation Support Service Utilizing 'Agent Skills'
Sniff Out and Hitokuse have jointly launched a service that automates ad agency tasks such as submission, budget management, and reporting by applying Anthropic's 'Agent Skills' AI standard, reducing task times from hours to minutes.
📋 Article Processing Timeline
- 📰 Published: April 14, 2026 at 20:01
- 🔍 Collected: April 14, 2026 at 11:31
- 🤖 AI Analyzed: April 19, 2026 at 17:48 (126h 16m after Collected)
Sniff Out Co., Ltd. (Headquarters: Shinjuku-ku, Tokyo, Representative Director: Kai Tsumoto, hereinafter "Sniff Out") and Hitokuse Co., Ltd. (Headquarters: Shinjuku-ku, Tokyo, President and Representative Director: Wataru Miyazaki, hereinafter "Hitokuse") will jointly begin initiatives on April 14, 2026, to apply "Agent Skills," an industry-standard specification for AI agents, to advertising agency operations, automating the three major tasks of ad submission, budget management, and reporting. This initiative will be promoted as a joint project between AI agent development vendor "Sniff Out" and "Hitokuse," which has expertise in digital advertising agency operations.
Click here for details
[Background and Challenges]
The ad submission, operation management, and reporting of advertising agency operations still largely depend on manual labor. It has become common for even experienced staff to spend more than 3 to 5 hours per week per project on setting tasks that span multiple ad platforms, with formats and operation rules varying from client to client.
Automation through general-purpose RPA and standard programs has struggled to flexibly adapt to this "diversity by company and project," and the reality was that full-scale automation across the industry had not progressed. As a result, the maximum number of projects a single staff member can handle is considered to be 5 to 10, creating a structural problem where labor costs determine the revenue ceiling.
[What is Agent Skills?]
Agent Skills is an industry-standard framework for designing and executing AI agents, established by Anthropic (developer of Claude). It was open-sourced as a common industry standard at the end of 2025, and major AI vendors including OpenAI and Google have announced their adoption.
In this framework, the tasks that an AI agent should execute are defined as a single "skill," and the agent autonomously executes operations spanning multiple steps based solely on natural language instructions. Because it is an industry-common standard, it is expected to become the foundation for automating all types of operations in the future, similar to MCP (Model Context Protocol). This initiative provides accompanying support to automate ad operations using Agent Skills, and its effectiveness in actual operations has already been proven.
[Overview of the Initiative]
Sniff Out and Hitokuse will jointly build and introduce custom Agent Skills tailored to each company's business processes, judgment rules, and client requirements. This makes it possible to introduce AI agents tailored to each advertising agency, client, and other communication styles, achieving production launch in as short as one week.
[Main Operations Targeted for Automation]
1. Automation of ad submission
Simply by handing the client's instruction document and meeting notes to Agent Skills, it automatically submits ads via API to major platforms such as Google, Meta, and Yahoo!. Tasks that previously took 30 to 60 minutes are reduced to approximately 3 minutes.
2. Automation of daily operations and budget management
It constantly monitors the delivery pace, CPA, and period budget, automatically making daily adjustments based on set thresholds. It also supports alert notifications to Slack and advance warnings of period end and budget depletion.
3. Automation of reporting
It accommodates different reporting formats for each client and fully automates data collection, formatting, consistency checks, and transmission. In parallel operation tests, cases with fewer errors than human-created reports have been confirmed.
[Addressing Security]
In delegating operation privileges for ad accounts to the AI agent, this initiative will implement the following security designs as standard.
- Risk management via threshold control: Setting threshold controls and approval flows for operations such as budget changes and ad suspension. (e.g., automatic suspension when the spend rate is 99% or higher, and mandatory confirmation by the person in charge for changes exceeding ±20%)
- Segregation of duties: Skill creation and editing can only be performed by administrators, limiting general employees to execution-only privileges.
- Context engineering: Explicitly defining rules and judgment criteria that the AI must follow.
- Complete audit trail: All operations are logged.
[Service Overview]
Pricing: From 500,000 yen per month (Initial implementation fee is separate; individual quotes based on project scale)
Target: Advertising agencies, business companies with in-house ad operation teams
Contact: info@sniffout.net, i.yoshida@hitokuse.com
Details: https://ad-a
Click here for details
[Background and Challenges]
The ad submission, operation management, and reporting of advertising agency operations still largely depend on manual labor. It has become common for even experienced staff to spend more than 3 to 5 hours per week per project on setting tasks that span multiple ad platforms, with formats and operation rules varying from client to client.
Automation through general-purpose RPA and standard programs has struggled to flexibly adapt to this "diversity by company and project," and the reality was that full-scale automation across the industry had not progressed. As a result, the maximum number of projects a single staff member can handle is considered to be 5 to 10, creating a structural problem where labor costs determine the revenue ceiling.
[What is Agent Skills?]
Agent Skills is an industry-standard framework for designing and executing AI agents, established by Anthropic (developer of Claude). It was open-sourced as a common industry standard at the end of 2025, and major AI vendors including OpenAI and Google have announced their adoption.
In this framework, the tasks that an AI agent should execute are defined as a single "skill," and the agent autonomously executes operations spanning multiple steps based solely on natural language instructions. Because it is an industry-common standard, it is expected to become the foundation for automating all types of operations in the future, similar to MCP (Model Context Protocol). This initiative provides accompanying support to automate ad operations using Agent Skills, and its effectiveness in actual operations has already been proven.
[Overview of the Initiative]
Sniff Out and Hitokuse will jointly build and introduce custom Agent Skills tailored to each company's business processes, judgment rules, and client requirements. This makes it possible to introduce AI agents tailored to each advertising agency, client, and other communication styles, achieving production launch in as short as one week.
[Main Operations Targeted for Automation]
1. Automation of ad submission
Simply by handing the client's instruction document and meeting notes to Agent Skills, it automatically submits ads via API to major platforms such as Google, Meta, and Yahoo!. Tasks that previously took 30 to 60 minutes are reduced to approximately 3 minutes.
2. Automation of daily operations and budget management
It constantly monitors the delivery pace, CPA, and period budget, automatically making daily adjustments based on set thresholds. It also supports alert notifications to Slack and advance warnings of period end and budget depletion.
3. Automation of reporting
It accommodates different reporting formats for each client and fully automates data collection, formatting, consistency checks, and transmission. In parallel operation tests, cases with fewer errors than human-created reports have been confirmed.
[Addressing Security]
In delegating operation privileges for ad accounts to the AI agent, this initiative will implement the following security designs as standard.
- Risk management via threshold control: Setting threshold controls and approval flows for operations such as budget changes and ad suspension. (e.g., automatic suspension when the spend rate is 99% or higher, and mandatory confirmation by the person in charge for changes exceeding ±20%)
- Segregation of duties: Skill creation and editing can only be performed by administrators, limiting general employees to execution-only privileges.
- Context engineering: Explicitly defining rules and judgment criteria that the AI must follow.
- Complete audit trail: All operations are logged.
[Service Overview]
Pricing: From 500,000 yen per month (Initial implementation fee is separate; individual quotes based on project scale)
Target: Advertising agencies, business companies with in-house ad operation teams
Contact: info@sniffout.net, i.yoshida@hitokuse.com
Details: https://ad-a