Who This Playbook Is For
You run a business — or a department inside one — and you've got a nagging feeling you're haemorrhaging money on manual work that shouldn't still be manual. You've heard the AI hype. You've probably tried ChatGPT a few times. Maybe you've even built a few automations with Zapier or Make.
But you haven't cracked the real thing yet: getting AI to genuinely run parts of your operation, day in, day out, without someone babysitting it.
This playbook is the bridge between "we should probably do something with AI" and actually doing it. No theory. No academic waffle. Just the frameworks, numbers, and step-by-step plans we use with clients who pay us five figures to implement this stuff.
What you'll walk away with:
- ✓ A clear picture of exactly how much manual work is costing you
- ✓ Five automations you can deploy in your first week that deliver measurable ROI
- ✓ The architecture for building an AI agent team that runs your operations
- ✓ A 30-day implementation roadmap you can start Monday
- ✓ A ready-to-use business case template for getting buy-in from leadership
Chapter 1: The AI Revenue Leak
Why You're Losing Money Right Now
Here's an uncomfortable truth: by the time you're talking to us, you've already wasted somewhere between $50K and $500K on work that didn't need a human.
We did a study across 40 of our clients — SMEs and mid-market companies, mostly in the UK, ranging from $2M to $50M in revenue. The average company was spending 23% of total payroll on tasks that fall into one of seven categories. Every single one can be automated today.
The 7 Manual Processes Bleeding You Dry
1. Email Triage & Response $18K–$45K/year wasted
The average knowledge worker spends 2.5 hours per day on email. In a team of 10, that's 25 hours per day of collective email faffing. At $35/hour blended cost, you're looking at $227K per year. Even if only 40% is automatable — that's $91K.
2. Data Entry & Transfer $25K–$80K/year wasted
The silent killer. One client — a recruitment firm with 30 staff — had three full-time people whose jobs were essentially moving data between systems. We automated 85% of their data transfer work in six weeks.
3. Invoice & Expense Processing $15K–$60K/year wasted
Manual invoice processing costs $8–$15 per invoice. At 500 invoices/month, that's $48K–$90K per year. AI brings it down to $1–$3 per invoice. We've seen companies lose $30K+ per year on duplicate payments alone.
4. Report Generation & Data Analysis $20K–$75K/year wasted
A property management client had a finance analyst spending every Friday — the entire day — building a weekly report. We automated it. It now generates at 6am every Friday. Cost: $3K setup, $200/month. Annual saving: ~$25K.
5. Meeting Scheduling, Prep & Follow-up $10K–$35K/year wasted
The average manager spends 4 hours per week on meeting-related admin. For 5 managers, that's effectively a half-time employee doing nothing but meeting admin — $47K–$68K per year.
6. Customer Onboarding & Support Triage $20K–$90K/year wasted
Companies lose 12–18% of new customers during onboarding simply because the process is slow. At $5K average customer value, losing 15 out of 100 costs you $75K.
7. Compliance, Documentation & Audit Prep $15K–$120K/year wasted
A mid-size financial advisory firm was spending $8K/month on a compliance consultant. We automated 60% of the grunt work. The consultant now works two days a month instead of four. Annual saving: $48K.
How to Calculate YOUR Revenue Leak
Grab a spreadsheet. This takes 30 minutes and it'll be the most valuable half-hour you spend this month.
| Category | Automation Rate |
|---|---|
| Email triage | 40–60% |
| Data entry/transfer | 70–90% |
| Invoice processing | 80–95% |
| Report generation | 60–80% |
| Meeting admin | 50–70% |
| Customer onboarding | 60–85% |
| Compliance prep | 40–60% |
Real Example: A $8M Professional Services Firm
Team size
35 people
Total payroll (loaded)
$1.82M
Weekly automatable hours
187 hrs
Annual cost of those hours
$389K
Net Year 1 saving
$172K
Net Year 2+ saving
$196K
Chapter 2: The 5 AI Quick Wins (Week 1 ROI)
Stop planning. Start here. These five automations can be live within days, not months.
Email Triage Automation
Classifies incoming emails by urgency, routes them, drafts responses for routine enquiries.
Setup time
2–4 hours
Weekly saving
10–15 hrs/person
Cost
$50–$150/mo
Setup steps →
- Connect your email to Make or n8n
- Build classification rules using GPT-4o or Claude with 50 example emails
- Set up routing: urgent → Slack + flag; routine → draft response; spam → archive
- Run in "shadow mode" for 2–3 days to verify accuracy
- Go live
Invoice Processing
Extracts data from invoices, matches POs, codes accounts, routes for approval.
Setup time
4–8 hours
Processing time
<1 min/invoice
Cost
$100–$400/mo
Setup steps →
- Choose: Dext, Rossum, or custom pipeline
- Set up chart of accounts and supplier mappings
- Configure approval workflows by threshold
- Connect to accounting software (Xero, QuickBooks, Sage)
- Process first batch and verify
Meeting Prep & CRM Updates
Auto-generates briefing docs, transcribes meetings, extracts actions, updates CRM.
Setup time
3–5 hours
Weekly saving
4–6 hrs/person
Cost
$15–$30/user/mo
Report Generation
Pulls data from your systems, generates formatted reports with narrative summaries, delivers on schedule.
Setup time
4–8 hours
Compile time
<5 minutes
Cost
$100–$300/mo
💡 Pro tip: Start with the report your CEO actually reads. When leadership sees an AI-generated report that's better than the manual version, buy-in for everything else becomes trivially easy.
Customer Onboarding (Self-Service)
Automates the entire journey from signed contract to active customer.
Setup time
6–12 hours
Dropout reduction
50%+
Cost
$200–$500/mo
💰 If your average customer is worth $10K/year and you cut onboarding dropout from 15% to 7%, that's +$96K in annual revenue.
Chapter 3: AI Agent Architecture
What AI Agents Actually Are
A chatbot answers questions when someone asks. An AI agent takes action without being asked. Three characteristics:
Autonomy
Triggered by events, not human prompts
Tool Access
Reads and writes to your business systems
Decision-Making
Follows rules + uses LLM judgement
The 9-Agent Model for Running Operations
Inbox Agent
Email & communications
Finance Agent
Invoices, expenses, payments
Calendar Agent
Scheduling & meetings
CRM Agent
Customer data & pipeline
Content Agent
Documents, reports, proposals
Onboarding Agent
Customer & employee onboarding
Support Agent
Customer support & helpdesk
Compliance Agent
Regulatory & policy
Orchestrator
Cross-agent coordination
How to Structure Your First AI Team
- Start with 2 agents — typically Inbox + Finance or CRM
- Get them stable — run 2–4 weeks, monitor, tune
- Add the Orchestrator — coordinate between agents
- Scale gradually — one new agent every 4–6 weeks
⚠️ Golden rule: Every agent needs a human supervisor who checks in daily and reviews edge cases.
Chapter 4: Implementation Roadmap
The 30-Day Plan: Zero to AI-Automated
Days 1–3: Audit & Assessment
- • Complete the Revenue Leak Calculator
- • Identify top 3 automation opportunities
- • Audit your tool stack — APIs, integrations
- • Document top 5 manual processes step-by-step
Days 4–7: Foundation Setup
- • Choose automation platform + set up LLM access
- • Deploy Email Triage in shadow mode
- • Start Invoice Processing setup
Days 8–14: First Agents Live
- • Take Email Triage live
- • Complete Invoice Processing → shadow mode
- • Deploy Meeting Prep (plug-and-play)
- • Begin documenting results
Days 15–21: Expand & Optimise
- • Take Invoice Processing live
- • Deploy Report Generation
- • Start Customer Onboarding design
- • First optimisation pass on Email Triage
Days 22–30: Consolidate & Plan
- • Deploy Customer Onboarding
- • Compile first monthly AI Operations Report
- • Build business case for Phase 2
- • Plan next three agents
Tool Stack Recommendations
Total monthly cost for a 20-person company: $800–$2,500/month
| Category | Recommended | From |
|---|---|---|
| Automation | Make ★★★★★ | $55/mo |
| LLM | OpenAI + Anthropic | Pay-per-use |
| Documents | Dext ★★★★☆ | $25/mo |
| Meetings | Fireflies ★★★★★ | $15/user/mo |
| CRM | HubSpot ★★★★★ | Free |
| Support | Intercom ★★★★★ | $65/mo |
AI vs. One More Hire
| New Hire | AI Stack | |
|---|---|---|
| Annual cost | $35K–$55K | $12K–$30K |
| Available hours | 1,820/yr | 8,760/yr |
| Ramp-up | 3–6 months | 1–4 weeks |
| Error rate | 2–5% | 0.1–0.5% |
| Scalability | Linear | Near-zero marginal |
Chapter 5: ROI Calculator & Business Case Template
The Board-Ready Business Case
Copy this structure. Fill in your numbers. We've used this to secure buy-in at companies from $3M startups to $200M enterprises.
1. Executive Summary
We are currently spending approximately $[X]/year on manual operational processes that can be automated. A phased implementation over 6 months would deliver a net saving of $[X] in Year 1.
2. Current State
Total automatable hours/week, annual cost, key pain points, error rates, customer impact.
3. Proposed Solution (3 Phases)
Phase 1: Email + invoices + meetings. Phase 2: Reports + onboarding. Phase 3: CRM agent + orchestration.
4. Investment Required
Automation platform, LLM API costs, specialist tools, implementation support, internal time.
5. Expected Returns
Conservative (40% automation), Moderate (60%), Aggressive (80%). Present all three; stake your reputation on conservative.
6. Risk Mitigation
Shadow mode testing, human oversight, phased rollout, enterprise-grade security, monthly reviews.
7. Success Metrics
Hours saved, error reduction, processing time, employee satisfaction, customer satisfaction.
What to Track
📊 Efficiency
- • Hours saved/week
- • Processing time/unit
- • Error rate before vs. after
- • Backlog size
💷 Financial
- • Direct cost saving
- • Revenue impact
- • Cash flow improvement
- • Cost avoidance
👥 People
- • Employee satisfaction
- • Retention rates
- • Capacity created
✅ Quality
- • Output consistency
- • Customer satisfaction
- • Compliance accuracy
Appendix: AI Tool Directory
The 20 tools we recommend most frequently. Pricing current as of early 2026.
Automation Platforms
| Tool | Best For | Price | Rating |
|---|---|---|---|
| Make | Visual workflows | $55/mo+ | ★★★★★ |
| n8n | Technical teams | Free/$20+ | ★★★★☆ |
| Zapier | Simple automations | $55/mo+ | ★★★☆☆ |
LLM Providers
| Tool | Best For | Rating |
|---|---|---|
| OpenAI GPT-4o | General-purpose, fast | ★★★★★ |
| Anthropic Claude | Long docs, analysis | ★★★★★ |
| Google Gemini | Multimodal, large context | ★★★★☆ |
Meeting, CRM & Support
| Tool | Best For | From | Rating |
|---|---|---|---|
| Fireflies.ai | CRM integration | $15/user | ★★★★★ |
| HubSpot | SME all-in-one | Free | ★★★★★ |
| Intercom | AI-first support | $65/mo | ★★★★★ |
| Notion AI | Internal knowledge | $7.50/user | ★★★★☆ |
What Comes Next
This week
Complete the Revenue Leak Calculator. Know your number.
Next week
Deploy Quick Win #1 (Email Triage). Get your first taste.
This month
Work through all five Quick Wins. Build momentum and evidence.
Next month
Present your business case. Start planning Phase 2.