005 · AI Agent Build
One agent workstream for every function: platform + first production agents, three working agents per person by mid-Q3, every agent owned and managed like a product
What this is
The program that makes "agent-augmented" real instead of aspirational: one workstream combining the agent platform/infrastructure, the marketing agents, and the ops & reporting agents — so every function gets working agents without three separate build efforts. Verto's AI engineer supports the build (SOW §3.7: repeatable AI workflows inside the Digital & Growth function, on Redis's existing technology and team); the platform and standards are Redis-owned (Keith + Josh via MOps).
Goal: give every function a working AI agent — with the leverage measurable in hours returned.
Why now
The resourcing model depends on it: every contractor role runs agent-augmented, which only holds if the agents exist, work, and are owned. The two highest-value replacements are already committed — the website content-creation agent (the GrowthX replace) and the reporting agent on the Snowflake MCP — and the coding-agent workflow is the stated capacity answer for the ~100-page refactor. The H2 memo made the cultural bar explicit at the August build workshop: everyone ships.
The operating rules (H2 plan)
- Every team member ships three working agents by mid-Q3 — the build workshop is the first ship; agents are everyone's job, not a specialist's.
- First production agents take over standing tasks across paid media, web, and data analysis — replacements for repeated work, not demos.
- Each agent has a named human owner and is managed like a product — brief, scope, iteration, retirement.
- Hours returned is logged from day one, the inventory is public, and nothing runs without a daily owner — leverage is measured, not vibes.
The build, by wave
- V1 (Month 1) — platform decisions, the repo, and the first pods; the content-creation agent for the website (feeding the AI content workflow and the refactor's page production) and the reporting agent on the Snowflake MCP (answers from the warehouse we own — the Data & tracking stack makes this possible).
- V2 (Month 2) — end-to-end automation and autonomy: the paid-media optimization agent and the campaign-creation agent (all channels).
- V3 (Month 3) — cost controls and governance, the website + conversion-rate agent, marketing agents connected to live campaign data, and a pilot of agent-generated content in one live campaign.
The 79-agent catalog on the Agent List is the reference index — owner-mapped one-liners that become living specs as each is scoped. The committed builds above are the spine; the catalog is the backlog.
Work to be done
- Platform stood up + the first two production agents — decisions, repo, V1 pods, and the content-creation + Snowflake-reporting agents running with named owners and logged hours.
- The workshop bar met — three working agents per team member by mid-Q3, in the public inventory, each with a daily owner.
- V2/V3 shipped with governance — paid optimization and campaign creation live, then cost controls, governance, and the CRO agent — with the hours-returned ledger proving the leverage model.
Current state
Platform decisions underway; the first two builds are committed for Month 1 with Verto's AI engineer onboarding Sep 1. The 6sense configuration and warehouse migration set the data foundations the reporting and optimization agents consume.
Dependencies & risks
Ops & reporting agent ownership is still TBD — assign before V1 ships or the daily-owner rule is broken on day one. Governance and cost controls arrive in V3; watch spend and output quality in the window before then. Agent access rides the contractor security protocol (individual accounts, Okta). [owner to complete]
What this is
The program that makes "agent-augmented" real instead of aspirational: one workstream combining the agent platform/infrastructure, the marketing agents, and the ops & reporting agents — so every function gets working agents without three separate build efforts. Verto's AI engineer supports the build (SOW §3.7: repeatable AI workflows inside the Digital & Growth function, on Redis's existing technology and team); the platform and standards are Redis-owned (Keith + Josh via MOps).
Goal: give every function a working AI agent — with the leverage measurable in hours returned.
Why now
The resourcing model depends on it: every contractor role runs agent-augmented, which only holds if the agents exist, work, and are owned. The two highest-value replacements are already committed — the website content-creation agent (the GrowthX replace) and the reporting agent on the Snowflake MCP — and the coding-agent workflow is the stated capacity answer for the ~100-page refactor. The H2 memo made the cultural bar explicit at the August build workshop: everyone ships.
The operating rules (H2 plan)
- Every team member ships three working agents by mid-Q3 — the build workshop is the first ship; agents are everyone's job, not a specialist's.
- First production agents take over standing tasks across paid media, web, and data analysis — replacements for repeated work, not demos.
- Each agent has a named human owner and is managed like a product — brief, scope, iteration, retirement.
- Hours returned is logged from day one, the inventory is public, and nothing runs without a daily owner — leverage is measured, not vibes.
The build, by wave
- V1 (Month 1) — platform decisions, the repo, and the first pods; the content-creation agent for the website (feeding the AI content workflow and the refactor's page production) and the reporting agent on the Snowflake MCP (answers from the warehouse we own — the Data & tracking stack makes this possible).
- V2 (Month 2) — end-to-end automation and autonomy: the paid-media optimization agent and the campaign-creation agent (all channels).
- V3 (Month 3) — cost controls and governance, the website + conversion-rate agent, marketing agents connected to live campaign data, and a pilot of agent-generated content in one live campaign.
The 79-agent catalog on the Agent List is the reference index — owner-mapped one-liners that become living specs as each is scoped. The committed builds above are the spine; the catalog is the backlog.
Work to be done
- Platform stood up + the first two production agents — decisions, repo, V1 pods, and the content-creation + Snowflake-reporting agents running with named owners and logged hours.
- The workshop bar met — three working agents per team member by mid-Q3, in the public inventory, each with a daily owner.
- V2/V3 shipped with governance — paid optimization and campaign creation live, then cost controls, governance, and the CRO agent — with the hours-returned ledger proving the leverage model.
Current state
Platform decisions underway; the first two builds are committed for Month 1 with Verto's AI engineer onboarding Sep 1. The 6sense configuration and warehouse migration set the data foundations the reporting and optimization agents consume.
Dependencies & risks
Ops & reporting agent ownership is still TBD — assign before V1 ships or the daily-owner rule is broken on day one. Governance and cost controls arrive in V3; watch spend and output quality in the window before then. Agent access rides the contractor security protocol (individual accounts, Okta).