Team Standardisation &
Advanced Workflows
Your team has now learned how to prompt effectively, built a prompt library, and used Claude for deep analysis. This final Efficiency Track tutorial is about locking in those gains at scale. A team where every member uses Claude differently produces inconsistent work. A team with shared standards, shared Projects, clear ownership, and defined quality gates produces consistently excellent work — and improves over time. This tutorial shows you how to build that.
- Design a complete Claude Projects architecture for the whole agency
- Write production-quality Project Instructions for client and agency Projects
- Define which tasks should always use Claude and what the expected output looks like
- Establish quality gates so Claude output is reviewed before it reaches clients
- Run an effective Show & Tell session to embed learning across the team
- Build a process for continuously improving the prompt library over time
1Taking Stock — Where the Team Is Now
After four weeks of training, here's what a well-progressed team should have in place. Use this as a checklist before the Week 4 standardisation work:
If some team members are behind, Week 4 is a good opportunity to catch up — the standardisation work benefits from everyone being involved. Pair ahead-of-schedule colleagues with those still working through earlier tutorials.
2Designing Your Agency's Claude Projects Architecture
Claude for Teams lets everyone share Projects. The question is: how should you structure them? Too few Projects and context gets muddled. Too many and no one remembers which to use. Here's a structure that works well for a technical SEO agency:
Who owns each Project? Assign a named owner responsible for keeping the Project Instructions up to date. Agency Standards and Prompt Library = a senior SEO or the team lead. Client Projects = the account lead for that client. This prevents Projects from going stale.
3Writing Production-Quality Project Instructions
The Project Instructions document is the most important thing in any Project. It's the permanent context Claude always has — so it needs to be comprehensive, current, and well-structured. Here are the two most important templates:
Template 1 — Agency Standards Project
# Agency Identity You are a senior technical SEO specialist working at [Agency Name]. We are a technical SEO agency specialising in [your specialisms, e.g. e-commerce, SaaS, enterprise]. # Our Methodology We follow a "technical first" approach: fix crawlability and indexation before on-page or content work. We prioritise issues by: 1. Crawl and indexation blockers (Critical) 2. Ranking signal issues (High) 3. User experience and conversion (Medium) 4. Future-proofing and best practice (Low) # Output Standards - Internal documents: concise, technical, direct. No padding. - Client documents: professional, jargon-free, solution-oriented. - Issue severity labels: always use Critical / High / Medium / Low. - Recommendations: always state what to do, not just what's wrong. - Never use filler phrases: "it's important to note", "in conclusion", "as an AI language model", "certainly!", or similar. # Tools We Use - Crawling: Screaming Frog SEO Spider - Analytics: Google Analytics 4 - Search data: Google Search Console - Rank tracking: [your tool] - Backlinks: [your tool] # Client Communication Tone Professional, confident, and plain-spoken. We do not over-promise. We explain what we found, why it matters in business terms, and what we're going to do about it. Avoid: "amazing", "fantastic", "game-changing", or other marketing hyperbole. # What You Should Always Do - Flag any assumptions you've made if data is incomplete - State your confidence level on recommendations where relevant - Offer alternative approaches when there is genuine trade-off - Ask a clarifying question if the brief is genuinely ambiguous
Template 2 — Client Project Instructions
# This project is for: [CLIENT NAME] You are a senior technical SEO specialist working on the [Client] account at [Agency Name]. # Client Overview - Website: [domain] - Industry: [industry] - Business type: [B2B/B2C, product/service] - Target audience: [describe their customers] - Primary goal: [what they hired us for] # Technical Context - CMS: [e.g. WordPress 6.x with Yoast SEO] - Hosting: [e.g. WP Engine, self-hosted] - Approx. indexed pages: [number] - Dev resource: [e.g. One in-house dev, slow turnaround] - Known issues: [list any known problems] # Key Pages and URL Structure [Describe the main sections of the site and their URL patterns] e.g. /blog/ = editorial content, /services/ = commercial pages, etc. # Our Key Contacts - Primary: [Name, Role] — [technical/non-technical?] - Secondary: [Name, Role] # Competitors [List 2–4 main organic competitors with their domains] # Current Focus [What are we actively working on right now?] e.g. "Currently in Phase 1: fixing crawlability issues identified in the March 2026 audit. Developer implementation begins April 2026." # Things to Avoid [Any known sensitivities — e.g. "Do not suggest changing the URL structure — the client has rejected this twice."]
Keep it current: A Project Instructions document that's 6 months out of date is worse than none — Claude will confidently act on stale information. Schedule a monthly 10-minute review of each active client Project to update the "Current Focus" section and any resolved issues.
4Standard Operating Procedures — Which Tasks Always Use Claude
Without clear guidance, some team members will use Claude for everything and others for nothing. SOPs define which tasks should always go through Claude, what the input looks like, and what a good output looks like. Here are the core SOPs for a technical SEO agency:
Generate technical SEO brief from intake form
Account lead fills intake form → runs Prompt C1 from Tutorial 5 → edits output → sends to team
Prioritise Screaming Frog findings
SEO analyst exports Issues tab → runs Prompt #2 from Tutorial 2 → reviews and adjusts priorities → builds action plan
Interpret GSC performance data
Pull last month's GSC data → run Prompt #6 from Tutorial 2 → combine insights with PageSpeed and ranking data → write narrative
Translate technical findings into client language
Any technical issue that needs client explanation → run Prompt #4 from Tutorial 2 before including in any client-facing document
Review all schema before implementation
Write or modify JSON-LD → run Prompt #3 from Tutorial 2 → apply corrections → validate in Google's Rich Results Test
Cannibalisation check before new content
Before creating any new page → run Prompt A1 from Tutorial 5 against current GSC data → confirm no existing page already covers this intent
5Quality Gates — What to Check Before Using Claude's Output
Claude is a powerful analyst and writer, but it can be confidently wrong. Quality gates are the checks your team applies before any Claude output is used in client work. They're not about distrust — they're about professional standards.
- Any specific statistic or data point Claude cites
- Technical recommendations against the actual client CMS/setup
- Schema JSON-LD in Google's Rich Results Test before implementing
- Cannibalisation findings by checking actual GSC URL data
- Any claim about competitor activity or industry benchmarks
- Rewritten title tags — check final length in a pixel tool
- Raw, unedited Claude output — always read and lightly edit first
- Analysis that references data you haven't verified
- Strategy recommendations without sense-checking against your expertise
- Schema or code that hasn't been validated in the appropriate tool
- Anything Claude was uncertain about (it will flag this if you asked it to)
The 80% rule: Claude's output is usually 80% ready. The 20% that needs human attention is typically: (1) specifics Claude couldn't know without more context, (2) tone adjustments for a particular client relationship, and (3) factual claims that need verification. Build the habit of reading output critically, not just copying it.
6Keeping the Prompt Library Alive
A prompt library that never gets updated goes stale. Search behaviour changes, clients change, Google's guidelines change, and your team's workflows evolve. Here's a lightweight process for keeping it current:
The monthly prompt review (20 minutes)
Run this review on the first Monday of each month. Open the Prompt Library Project in Claude.
Encouraging team contributions
The best prompts often come from the people doing the day-to-day work, not from the team lead. Create a low-friction way for anyone to suggest a new prompt:
| Method | How it works |
|---|---|
| Slack channel #claude-prompts | Anyone posts a prompt they've found useful. Team lead reviews monthly and adds the best ones to the library. |
| Project instruction note | Add a line at the top of the Prompt Library Project: "To suggest a new prompt, paste it into this conversation with the tag [SUGGEST]." |
| Weekly standup slot | Add a standing 2-minute agenda item: "Any useful Claude prompts to share this week?" |
7Running the Week 4 Show & Tell Session
The Show & Tell at the end of the training programme is one of the most valuable sessions you'll run. Team members on the Build track demo their tools; the rest of the team sees what's possible and often surfaces new tool ideas. Here's a ready-to-use agenda:
Welcome & context (5 min)
Team lead recap: what we set out to achieve, what the four weeks covered, and why today matters.
Demo 1 — URL Status Checker (Tutorial 3)
Live demo on a real client URL list. Show the CSV output and explain where in the workflow this replaces manual checking.
Demo 2 — Internal Linking Finder (Tutorial 4)
Live demo on a real client sitemap. Open the HTML report in the browser and walk through the top opportunities found.
Demo 3 — PageSpeed Batch Checker or GSC Puller (Tutorial 6)
Show the API-powered tool in action — even a pre-run report is fine. Focus on: "here's what used to take 45 minutes manually."
Open floor — team ideas (15 min)
Facilitated discussion: "What other tasks could we build tools for?" Capture every idea, no matter how vague. These become the backlog for Tutorial 8 and beyond.
Standards walkthrough (10 min)
Team lead walks through the SOPs from Section 4, the quality gates from Section 5, and the Projects architecture from Section 2. This is the moment they become official agency practice.
Capture the ideas list: The open floor discussion almost always surfaces 5–10 tool ideas from people who know the pain points best. Photograph or transcribe the list immediately — the most common outcome of not capturing it is forgetting the three best ideas by the following Monday.
8Practice Exercises
Set up all six Projects from Section 2 in Claude for Teams:
- Create the Agency Standards Project and fill in the Instructions template from Section 3
- Create or update the Prompt Library Project with all prompts from Tutorials 1–5
- Create a Client Project for your three most active clients using Template 2
- Assign an owner to each Project and note when the Instructions were last updated
- Invite all relevant team members to each Project
Adapt the SOPs from Section 4 for your agency's specific workflow:
- Review the six SOPs listed — do they match how your team actually works?
- Remove any that don't apply and add any tasks you do regularly that aren't listed
- For each SOP, write out: the trigger, who runs it, the prompt to use, and the expected output
- Add the final SOP list to the Agency Standards Project Instructions
- Share with the team and ask for one round of feedback before finalising
Organise the Week 4 team session:
- Book a 60-minute slot with the full team
- Brief the Build track members: each person has 8–10 minutes to demo their tool
- Run the session using the agenda from Section 7
- Capture the open floor ideas list — turn it into a prioritised backlog
- After the session, send a short follow-up to the team: which tools are now available, where the scripts live, and how to run them
9Summary
Key takeaway: Individual skill with Claude creates individual productivity gains. Shared standards, shared Projects, and defined SOPs multiply those gains across the entire team — and make sure they compound over time rather than plateauing. The Show & Tell is the moment this shifts from a training programme into an agency-wide working practice.