Ten rocks across Paid Media, Web, and SEO — plus one product-led project. Kickoff is Jul 14; everything starts in July and works through the quarter, with Sep 28 → Oct 1 held free for delays and revisions before rocks close Oct 1.
Google MXs run themselves — automated weekly, surfacing tasks and checks to us rather than the inverse.
Weekly MX checking consumes a large share of the team's time. Using a tool — Claude or, most likely, Viktor — to make Google MXs as automated as humanly possible is a huge lift to the team's week: the tool posts weekly and surfaces what needs attention instead of the team going looking.
Google MXs run fully automated on a weekly cadence, posting output and surfacing tasks/checks — the team no longer manually checks.
Same automation for Meta, as its own rock.
Same problem, second platform: Meta MX checking is recurring manual time the tool should absorb. Runs as its own rock so Meta gets a dedicated build and tuning pass rather than riding on the Google work.
Meta MXs run fully automated on a weekly cadence, posting output and surfacing tasks/checks — no manual checking.
Full upgrade to the copy & creative flow — and to the creative library feeding it.
V1 shipped a lot of new things that now need to be hardened and refined. V2 fully upgrades the copy & creative flow the paid social team lives in, and may extend Spark into PB nurture messaging. It's also a full upgrade to our creative work itself: the Google Drives feeding it get audited so every creative is clear, fully approved, and confidently usable — by humans or by AI pulling from them — with nothing confusing or difficult to understand left in.
Copy & creative flow fully upgraded, all V1 bugs squashed, V1 features refined, Google Drives audited and cleaned — and a decision (plus build, if green-lit) on PB nurture messaging.
Fully AI-enabled request intake — and AI answers the simple asks itself.
Paid media request intake is manual routing work, and simple data requests still pull the team in. A fully AI-enabled intake routes information automatically, and for easy requests AI provides the recommendation directly — the team never gets involved.
Intake is fully AI-enabled with automatic routing; easy data/simple requests receive AI recommendations with zero team involvement.
All web request communication returns to Partner Success — first, by end of July.
Communication on web requests currently runs through a dedicated comms/intake/QA role. Moving communication back to Partner Success first unblocks the rest of the rock set — and the comm layer itself gets an upgrade on the way over: Hiver templates audited to rock-solid, and a common-questions guide for the team.
Full communication on web requests is handled by the PS team — live by end of July — on top of a cleaned-up, rock-solid comm toolkit.
Every web task gets full AI intake — routed to US dev or overseas automatically.
Web intake is more complex than paid media's: beyond capturing the request, the AI has to decide who does the work — hand off to our US development team or our overseas team. Getting this right removes the human router entirely.
All web tasks intake through AI and route automatically to the US development team or the overseas team.
Nothing is certified done without passing the AI QA gate.
Returned work — especially from the overseas team — needs verification that what was requested was actually done. An AI QA bot checks every returning task against its request, for the overseas team and the US team alike, before anything is certified done.
All tasks from both the overseas and US teams pass AI QA before being certified done.
Half the legacy sites on the new framework by Oct 1, with a locked plan for the rest.
Fourteen partner sites still run the old framework. Q3 moves half of them: build the plan fast, start executing on sites in July, and complete net 7 upgrades by Oct 1 — ending the quarter with a concrete plan for the remaining seven.
Net 7 of the 14 legacy sites upgraded to the new framework, and a written, sequenced plan delivered for the remaining sites.
Every on-page deployment feature accounted for, decided, and cleaned up across every site — within 30 days.
On-page deployment features keep pushing things we don't believe in, causing mistakes across sites. We need a full account of every single feature, a distinct keep-or-kill decision on each, and a cleanup applied across the board so nothing deploys that we haven't consciously chosen.
Every on-page deployment feature inventoried with an explicit deploy/don't-deploy decision, and every site fully cleaned up and updated to match — done within 30 days.
The audit ran for half our accounts in Q2 — run it for the rest in Q3, phased, including new partners.
We ran the SEO audit for half of the accounts in Q2. Q3 covers the rest — but before running at scale, James submits his revisions and feedback to Danny, Danny updates the audit, and only then do the remaining accounts run on a phased plan that also covers newly onboarding partners.
Audit revisions incorporated, and every remaining account (plus newly onboarded partners) has its audit run and ready — phased to completion by end of quarter.
Clear the 90-day backlog, then hold a two-week aging standard.
Tasks older than 90 days are sitting incomplete, and the habit is pushing rather than accomplishing. Clear the aged backlog, get fully current, then tighten the definition: nothing sits past two weeks unless its due date is genuinely far out and wasn't manually pushed.
Zero tasks older than 90 days; board fully current; two-week aging standard adopted and holding.
Product-led project — not a James rock, but it needs his involvement throughout.
SearchAtlas as the publishing middleman slows everything down. Removing it and publishing straight to WordPress means faster publishing, true bulk publishing, and it potentially opens up internal-link management handled by us. Owned by the product team as a project; James is the SEO-side partner, not the rock owner.
Publishing goes straight to WordPress with SearchAtlas removed; bulk publishing works; internal-link management feasibility answered.
Product-led project — automated 30-day op cycles. James is the SEO-side partner, not the lead.
Op cycles disappeared and need to come back — but automated, not manual. Every 30 days, Claude/Viktor (or similar) produces the full update so anyone can go see what the SEO team accomplished over the last 30 days. Product leads the build; James supplies the SEO context and validates output.
Op cycles run every 30 days, completely automated, delivering a full team-visible update.