Operations & Systems

Automation Workflows for B2B Operations: Identifying Repeatable Work

Quick Summary

  • What this covers: Practical guidance for building and scaling your online presence.
  • Who it's for: Business operators, consultants, and professionals using AI + search.
  • Key takeaway: Read the first section for the core framework, then apply what fits your situation.

Before building automation, track task-level time allocation. Identify predictable process work separately from creative work and judgment.

Automation candidates should have repeatable logic and clear inputs and outputs. Measure time saved after implementation rather than assuming a result.

The workflows below describe triggers, logic, and tools. Validate each workflow and measure its effect in your own operation.

The Automation Audit: Finding Repeatable Work

Before automating anything, you need to know what to automate. The audit methodology:

Step 1: Track Every Task for Two Weeks

Use a simple time log — I used a Google Sheet with columns for: task name, category, duration, frequency, and a boolean for "did this require judgment?"

Categories I tracked:

  • Content production — research, drafting, editing, publishing
  • CRM maintenance — data entry, deduplication, tag management, reporting
  • Communication — email, Slack, client updates, meeting scheduling
  • Reporting — dashboard updates, client reports, internal metrics
  • Admin — invoicing, file management, tool configuration

Step 2: Score Each Task for Automation Potential

After two weeks of logging, score every recurring task:

Score Criteria
5 Fully automatable — no judgment, predictable inputs/outputs
4 Mostly automatable — needs human review of output
3 Partially automatable — some steps are automatable, others need judgment
2 Minimally automatable — mostly judgment, some steps can be templated
1 Not automatable — pure creativity, relationship, or strategic thinking

Tasks scoring 4-5 are automation candidates. Tasks scoring 3 are semi-automation candidates (automate the predictable steps, human handles the judgment steps). Tasks scoring 1-2 stay manual.

Step 3: Prioritize by Time × Frequency

A task taking 10 minutes but happening 20 times weekly (200 minutes) is a higher priority automation target than a task taking 60 minutes but happening once monthly. Multiply time per occurrence by weekly frequency to rank automation ROI.

Tasks with high automation scores become candidates for the first automation wave. Prioritize them using your recorded time allocation.

Workflow 1: Content Production Pipeline

Manual work: researching keywords, creating briefs, formatting articles, adding frontmatter, deploying files.

Automation design: the AI content workflow handles brief generation, drafting, and formatting. Human time goes to quality review and strategic decisions.

The Automated Steps

Brief generation: A Claude Code or Codex CLI skill reads the topical map, selects the next articles in the priority queue, and generates structured content briefs including target keywords, H2/H3 outlines, internal linking targets, and FAQ questions. Trigger: weekly queue review command.

Drafting: Each brief feeds into a drafting prompt that loads voice specifications, reference articles, and the topical map context. Claude Fable 5 produces the complete article draft. The drafting step runs in batch — 10 articles at once.

Formatting: A script adds frontmatter, standardizes heading levels, validates internal link slugs, and formats entity names in bold. Review the formatting output before publication.

Deployment: A Git commit pushes formatted articles to the repository. The build system generates HTML, sitemap, and schema markup automatically.

The Human Steps (Irreducible)

  • Editorial quality review (3-5 minutes per article)
  • Strategic priority selection (which articles to produce this week)
  • Voice calibration adjustments (updating constraints based on output patterns)

Workflow 2: CRM Data Hygiene

Manual work: manually reviewing new contacts for data completeness, scanning for duplicates, auditing tag compliance, updating stage accuracy.

Automation design: automated scans flag issues; human reviews and resolves flagged items.

The Automated Steps

Daily duplicate scan: A Google Apps Script runs nightly, exporting contacts created in the last 24 hours from a vertical CRM via API. The script normalizes phone formats and checks against existing records for exact and fuzzy matches. Matches get flagged with system:duplicate-review tag.

Weekly completeness audit: Script scans all contacts modified in the past week. Contacts missing critical fields (email, phone, source tag, stage tag) get flagged with system:data-incomplete.

Tag compliance check: Script pulls all tags created in the past week. Any tag not in the approved taxonomy gets flagged for review with the agent who created it and the correct taxonomy alternative.

Stage accuracy automation: Contacts tagged stage:hot-prospect with no logged activity in 60+ days get automatically reclassified to stage:nurture. Contacts in stage:new with a logged call get automatically advanced to stage:contacted.

The Human Steps (Irreducible)

  • Reviewing and executing duplicate merges (automated merging is too risky — data loss potential)
  • Evaluating flagged tag violations and coaching agents
  • Spot-checking stage transitions for edge cases

Workflow 3: Client Reporting

Manual work: pulling data from Google Search Console, Ahrefs, Google Analytics, compiling into client-facing reports, formatting, and sending.

Automation design: automated data pull and compilation, human reviews and adds commentary.

The Automated Steps

Data aggregation: A Google Apps Script pulls weekly metrics from Google Search Console API (impressions, clicks, CTR, position changes), Google Analytics API (sessions, conversions, bounce rate), and writes them to a client-specific Google Sheet.

Report compilation: A template sheet auto-populates with the week's data, calculates week-over-week and month-over-month changes, and highlights metrics that moved beyond threshold (>10% change in either direction).

Chart generation: Google Sheets charts update automatically as data populates. Trend lines, comparison charts, and KPI dashboards regenerate without manual intervention.

The Human Steps (Irreducible)

  • Reading the compiled report and identifying the narrative (what happened, why, what to do next)
  • Writing 2-3 paragraphs of strategic commentary
  • Sending the report with context the numbers don't show

Workflow 4: Email Triage and Response

Manual work: processing 200+ emails, categorizing, routing, responding to routine inquiries.

Automation design: automated categorization and routing, human handles items requiring judgment.

The Automated Steps

Inbox categorization: Gmail filters categorize incoming email by sender domain, subject line keywords, and thread participation. Categories: client communication, prospect inquiry, vendor/tool notification, newsletter, and internal.

Template responses: Routine inquiries (pricing questions, scheduling requests, status updates) trigger template responses via Gmail canned responses. The templates include dynamic fields populated from my contact database.

Calendar management: Meeting requests auto-route to Calendly links based on the type of request. Discovery calls, client check-ins, and internal syncs each have dedicated booking pages with appropriate durations and preparation requirements.

The Human Steps (Irreducible)

  • Reading and responding to substantive client communication
  • Evaluating prospect inquiries for fit before responding
  • Making decisions on requests that don't match templates

Workflow 5: Lead Routing and Follow-Up

Manual work: monitoring inbound leads, assigning to agents, setting follow-up reminders, tracking response times.

Automation design: automated routing handles assignment, escalation handles missed response windows, human reviews exceptions.

The Automated Steps

Instant lead routing: When a lead enters a vertical CRM from any source (web form, lead marketplace, phone call), the CRM auto-classifies by source tag, applies area tags based on property address or stated interest, and routes to the assigned agent based on geographic territory.

Notification cascade: Assigned agent receives push notification, SMS, and email simultaneously. No single notification channel is reliable — the cascade ensures the message reaches them.

Escalation timer: If no contact attempt is logged within 5 minutes, the lead escalates to the team lead. If no attempt within 15 minutes, the lead reassigns to the on-duty agent.

Follow-up sequencing: After initial contact, the contact enters an automated nurture sequence based on their stage and type. The sequence handles touchpoints at predetermined intervals without agent intervention.

The Human Steps (Irreducible)

  • Reviewing exception cases (routing conflicts, unclear geographic matches)
  • Coaching agents on response quality (speed is automated, quality isn't)
  • Adjusting routing rules when territory assignments change

Building vs Buying Automation

When to Build Custom

Build when:

  • Your workflow is unique to your business and no tool covers it
  • Integration between existing tools requires custom logic
  • The automation is simple enough for Google Apps Script, shell scripts, or Claude Code and Codex CLI skills
  • Cost sensitivity matters — custom automation has no per-seat recurring fees

When to Buy Existing Tools

Buy when:

  • The workflow is common across industries (email marketing, CRM routing, reporting)
  • Reliability matters more than flexibility — commercial tools have monitoring, alerting, and failover
  • Time to implement matters — Zapier or Make connects two tools in minutes vs hours for custom code
  • The team includes non-technical operators who need to modify workflows

The Hybrid Approach

Most of my automations are hybrid. a vertical CRM handles CRM-native workflows (routing, action plans). Google Apps Script handles data processing that CRM can't do natively (scoring, enrichment, cross-system reporting). Claude Code and Codex CLI handle complex workflows that require AI judgment (content production, brief generation, quality review).

The stack: commercial tools for infrastructure, custom scripts for integration, AI for judgment-adjacent automation. Each layer handles what it's best at. Forcing any single layer to do everything creates brittleness.

Measuring Automation ROI

Time saved is the obvious metric, but three others matter more:

Error reduction: Track the data quality issues found by automated scans. Review duplicates and their effects on subsequent work.

Consistency: Automated workflows execute the same way every time. The 200th lead gets the same routing speed, notification cascade, and follow-up sequence as the first. Manual processes degrade as operators fatigue, get distracted, or make exceptions.

Capacity: Measure recovered time and decide where to use it. Review delivery quality alongside time saved.

Take Action: Work With Victor

Key Recap

  • The Automation Audit: Finding Repeatable Work: Before automating anything, you need to know what to automate.
  • Workflow 1: Content Production Pipeline: Build when:
  • Your workflow is unique to your business and no tool covers it
  • Integration between existing tools requires custom logic
  • The automation is simp
  • Workflow 2: CRM Data Hygiene: Build when:
  • Your workflow is unique to your business and no tool covers it
  • Integration between existing tools requires custom logic
  • The automation is simp
  • Workflow 3: Client Reporting: Build when:
  • Your workflow is unique to your business and no tool covers it
  • Integration between existing tools requires custom logic
  • The automation is simp
  • Workflow 4: Email Triage and Response: Build when:
  • Your workflow is unique to your business and no tool covers it
  • Integration between existing tools requires custom logic
  • The automation is simp
  • Workflow 5: Lead Routing and Follow-Up: Build when:
  • Your workflow is unique to your business and no tool covers it
  • Integration between existing tools requires custom logic
  • The automation is simp

FAQ

How do I start automating if I have no technical skills?

Start with Zapier or Make — both are visual workflow builders that connect tools without code. Automate one workflow that saves at least 30 minutes weekly. Once you've validated the approach, expand to the next workflow. Technical skills develop through doing. You don't need to become a developer — you need to understand triggers, conditions, and actions.

What's the biggest automation mistake you've made?

Automating a broken process. I built an elaborate lead routing system before fixing the CRM data it routed against. The automation worked perfectly — it just perfectly executed bad logic against dirty data, producing consistently wrong agent assignments. Fix the process manually first. Automate the process that works, not the process that's broken.

How do I maintain automations as tools update their APIs?

Schedule a monthly automation health check. Run each workflow manually and verify the output matches expectations. API changes break integrations silently — the automation stops producing correct output but doesn't error visibly. The monthly check catches drift before it accumulates into business impact.

Can automation replace a virtual assistant?

For structured, repeatable tasks — yes, at lower cost and higher reliability. For tasks requiring judgment, context switching, and human communication — no. The sweet spot is automating the 60-70% of a VA's task list that follows predictable logic, then hiring a VA at reduced hours for the 30-40% that requires human flexibility.



Related Reading:


When This Doesn't Apply

Skip this if your situation is fundamentally different from what's described above. Not every framework fits every business. Use the diagnostic in the first section to determine whether this approach matches your current stage and goals.

Read the method behind the writing.

The Seventeen Principles map how this work is done. The work history gives its background.

Next stepSites and repositoriesThe public sites and code repositories where these methods run.

Topic map: APIs, webhooks, and automation

Articles in this cluster, then the 3 clusters that share the most topics with it.

This cluster: APIs, webhooks, and automation

Neighbor: CRM and pipeline data

Neighbor: Spreadsheets, CSV files, and data handoffs

Neighbor: Records, access, and continuity