How I Actually Use AI: The 2025 Surton Internal AI Playbook

At Surton, AI isn’t a future initiative—it’s how we work today. Across our team and the 30+ companies we’ve helped implement AI, we’ve developed a specific workflow system that consistently saves 15+ hours per person weekly on writing, analysis, research, and operations.

This guide documents our complete internal AI playbook. It includes the exact prompts we use, our 2025 tool stack, security policies, and real ROI data from implementations.

Quick Take

The fastest AI wins come from internal workflows, not product features. Start with meeting summaries (30 min → 5 min), email drafting (20 min → 5 min), and decision support. Use Claude 3.5 Sonnet for analysis and writing, Perplexity for research, and always provide context—generic prompts produce generic output. Most teams save 10-20 hours weekly per person within 60 days, worth $15k-50k+ annually.

The Surton AI ROI Framework

Before diving into workflows, let’s establish the business case. Here’s real data from our 2024 implementations:

Workflow	Before AI	With AI	Time Saved	Annual Value*
Meeting summaries	30 min	5 min	25 min/meeting	$12,500
Email drafting	20 min	5 min	15 min/email	$9,375
Report writing	4 hours	1.5 hours	2.5 hours/report	$15,625
Research & analysis	3 hours	45 min	2.25 hours	$14,063
Documentation	2 hours	30 min	1.5 hours	$9,375
Decision memos	1.5 hours	30 min	1 hour	$6,250

*Based on $75/hour loaded cost, 5 instances per week, 50 weeks/year

Typical weekly savings: 15-20 hours per knowledge worker
Typical annual value: $30k-75k per person in time recaptured for high-value work

Surton Case Study: Executive Team AI Adoption

A 12-person leadership team at a growth-stage company adopted our AI workflow system in Q1 2024:

• Week 1-2: Setup, training, prompt library deployment
• Week 3-4: 50% adoption on meeting summaries and email drafting
• Month 2: 80% adoption, documented 12 hours weekly savings per person
• Month 3: Expanded to research, analysis, and documentation workflows
• Result: 15 months later, team estimates 18 hours weekly savings per person. Team has not added headcount despite 40% business growth.

Investment: $200/month in AI tools for 12 people ($2,400/month)
Annual return: ~$650k in recaptured executive time

The 2025 Surton AI Tool Stack

Our stack has evolved significantly. Here’s what we use in 2025:

Core Tools (Daily Use)

Tool	Use Case	Cost	Privacy Level
Claude 3.5 Sonnet	Analysis, writing, complex reasoning, code review	$20/month Pro	High (Anthropic enterprise terms)
Perplexity Pro	Research with sources, fact-checking, learning	$20/month	Medium (verify sources)
Otter.ai Team	Meeting transcription, summaries	$20/user/month	High (SOC 2 compliant)
Notion AI	Internal knowledge work, document drafting	$10/user/month	High (Enterprise workspace)

Specialized Tools (As Needed)

Tool	Use Case	Cost
Midjourney/DALL-E 3	Image generation for presentations, marketing	$10-30/month
Descript	Video/audio editing, transcription	$15-30/month
ChatGPT Code Interpreter	Data analysis, visualization, calculations	Included in Plus
Grammarly Premium	Final polish on important communications	$12/month

Tool Selection Logic

Claude vs. GPT-4: We default to Claude 3.5 Sonnet for most work because:

• Better at following complex instructions
• More nuanced reasoning
• Stronger at maintaining context across long conversations
• Superior for analysis and structured output

We use GPT-4o for:

• Creative brainstorming (more divergent ideas)
• Quick questions where speed matters
• Tasks requiring web browsing (though Perplexity is usually better)

The Five Core AI Workflows

Here are the specific workflows we use daily, with exact prompt templates:

Workflow 1: Meeting Intelligence

The Problem: 60-minute meeting, 30 minutes writing up notes, action items scattered

AI Solution: Transcribe → Summarize → Structure → Distribute

Step-by-Step:

1. Record & Transcribe: Use Otter.ai, Zoom AI companion, or similar to get transcript

2. Initial Processing Prompt:

I need a structured summary of this meeting transcript. Extract:

DECISIONS MADE:
- [List each decision with context]

ACTION ITEMS:
- [Task] | Owner | Due Date | Priority (High/Medium/Low)

KEY DISCUSSION POINTS:
- [Main topics covered with key points]

OPEN QUESTIONS:
- [Issues raised but not resolved]

NEXT STEPS:
- [What happens next, by when]

Keep it concise. Use bullet points. Include specific details, not generic summaries.

3. Distribution: Copy into meeting notes doc, share with attendees

Time: 30 minutes manual → 5 minutes with AI
Savings: 25 minutes per meeting
At 10 meetings/week: 4+ hours weekly

Surton Example: Our weekly leadership meeting used to take 30 minutes of my time to document. Now I spend 5 minutes reviewing AI output and adding context. That’s 25 min × 50 weeks = 20+ hours annually on one meeting type.

Workflow 2: Communication Drafting

The Problem: Important email or Slack message takes 20-30 minutes to get right

AI Solution: Bullet points → AI draft → Review → Send

The SIT Prompt Framework:

Every communication prompt should include:

• Situation: Context about your role and the relationship
• Intent: What you want the recipient to do/understand
• Tone: The style (direct, diplomatic, urgent, casual)

Example Prompt:

Situation: I'm the CTO at a 50-person SaaS company. This email goes to our biggest
customer who's been asking about a feature delay.

Intent: Explain that their requested feature will ship 2 weeks later than promised,
but reassure them about quality and our commitment. Ask for a brief call to discuss
priorities.

Tone: Direct, accountable, solution-focused. Not defensive. Show we care about their success.

Here are the key points to include:
- Feature X delayed from March 15 to March 29
- Reason: Additional security review required by enterprise compliance
- We're adding extra QA time to ensure no issues
- Want to understand if other features should be prioritized
- Proposing call Thursday or Friday this week

Draft a 3-paragraph email. Professional but warm.

Time: 20-30 minutes → 5 minutes (provide bullets, review draft)
Savings: 15-25 minutes per communication
At 5 important communications/week: 1-2 hours weekly

Workflow 3: Research & Synthesis

The Problem: Need to understand a new topic, evaluate options, or analyze data—takes hours

AI Solution: Perplexity for research + Claude for analysis

Research Protocol:

1. Initial Exploration (Perplexity):

What are the leading approaches to [topic] in 2025?

Include:
- Major methodologies/frameworks
- Key vendors or tools
- Recent trends or changes
- Common pitfalls
- Sources for deeper reading

Cite specific sources.

2. Deep Dive (Claude with sources):

I need to make a decision about [specific question].

Here's what I know:
[Context about your situation]

Here's the research I've gathered:
[Paste Perplexity output]

Analyze this and help me:
1. Identify the 3 best options for my situation
2. Evaluate pros/cons of each
3. Recommend an approach with rationale
4. Flag risks or questions I should investigate further

Time: 3-4 hours manual → 45-60 minutes with AI
Savings: 2-3 hours per research task
At 2 research tasks/week: 4-6 hours weekly

Workflow 4: Decision Support

The Problem: Complex decision with multiple factors, stakeholders, and uncertainty

AI Solution: Structured analysis → AI critique → Refined decision

The Decision Memo Process:

1. Write your thinking: Document options, criteria, tradeoffs (15 minutes)

2. AI Critique Prompt:

I'm making a decision about [topic]. Here's my current thinking:

[Your analysis]

Please critique this:
1. What factors am I underweighting or ignoring?
2. What alternative framings should I consider?
3. What would someone who disagrees with my leaning say?
4. What additional information would most change this decision?
5. What risks am I underestimating?

Be direct. Challenge my assumptions.

3. Incorporate feedback → Refined decision

Time: 90 minutes wrestling with complexity → 30 minutes structured analysis
Savings: 1 hour per significant decision
At 2-3 significant decisions/week: 2-3 hours weekly

Surton Example: We used this process for a major vendor decision (replace our project management tool). My initial leaning was Option A based on features. AI critique surfaced integration costs and team adoption risk I’d underweighted. We chose Option B and avoided a costly migration mistake.

Workflow 5: Content & Documentation

The Problem: Documentation, blog posts, internal guides take forever to write

AI Solution: Outlines → AI expansion → Human refinement

The Content Process:

1. Create outline: Key points, structure, examples (10 minutes)

2. Expansion Prompt:

I need to write a [document type: blog post / internal guide / documentation] about [topic].

Target audience: [Who's reading and what they know]
Goal: [What should they understand or do after reading]
Tone: [Professional / Casual / Technical / Strategic]
Length: [Target word count or "concise" / "comprehensive"]

Here's my outline:
1. [Point 1]
2. [Point 2]
3. [Point 3]

Expand each section into full prose. Include specific examples where helpful.
Write in a clear, direct style. Avoid fluff.

3. Edit and personalize: Add specific stories, company context, voice (20 minutes)

Time: 3-4 hours writing → 1 hour (outline + AI + edit)
Savings: 2-3 hours per document
At 1-2 documents/week: 2-6 hours weekly

Security & Privacy: The Enterprise Framework

AI adoption without security discipline creates risk. Here’s our 2025 security policy:

Data Classification for AI Tools

Classification	AI Use	Examples
Public	Any AI tool OK	Published blog posts, marketing materials, public docs
Internal	Approved enterprise tools	Meeting notes (sanitized), internal processes, non-sensitive strategy
Confidential	Enterprise tools with enhanced privacy	Customer names (sanitized), financial projections, roadmap
Restricted	Self-hosted or manual only	Customer PII, security credentials, unreleased financials, legal docs

Approved Tool Configurations

Claude (Enterprise):

• Use Claude Pro or Enterprise (not free tier) for work
• Enable “Delete conversations” after 30 days for sensitive work
• Never use for customer PII or credentials

ChatGPT:

• Use ChatGPT Team or Enterprise (not Plus) for work
• Turn off “Improve model for everyone” in settings
• Enable data retention policies

Perplexity:

• Pro plan acceptable for research
• Verify sources independently for important decisions

Notion AI:

• Enterprise workspace required
• Document what internal knowledge is AI-accessible

Red Lines (Never Do)

• ❌ Never put customer PII into consumer AI tools
• ❌ Never share security credentials, API keys, or passwords
• ❌ Never input unreleased financial results or M&A discussions
• ❌ Never use AI for regulated advice (legal, medical, financial) without human review
• ❌ Never blindly trust AI output for important decisions—always verify

Implementation: Your First 30 Days

Week 1: Foundation

• Choose 1-2 AI tools (recommend: Claude Pro + Otter or similar)
• Set up accounts with proper privacy settings
• Document your security policy (even if simple)
• Pick 2-3 workflows to pilot (recommend: meeting summaries + email drafting)

Week 2: Pilot

• Use AI on chosen workflows daily
• Time yourself: manual vs. AI-assisted
• Refine prompts based on what works
• Share learnings with team

Week 3: Expand

• Add 1-2 more workflows
• Build prompt library for common tasks
• Measure time savings
• Identify other team members who should adopt

Week 4: Systematize

• Document your top 10 most-used prompts
• Create team prompt library
• Set up shared AI tool accounts (Team/Enterprise plans)
• Plan expansion to other workflows

Measuring Your AI ROI

Track these metrics monthly:

Time Savings:

• Hours saved per week on AI-assisted workflows
• Multiply by hourly cost (loaded salary) = weekly value
• Annualize for total value

Quality Improvements:

• Faster turnaround on communications
• Better-structured meeting outcomes
• More thorough analysis
• Reduced errors in documentation

Capability Expansion:

• Tasks you can now do that previously required specialists
• Research you can conduct independently
• Decisions made faster with better information

Surton ROI Example:

• Executive team: 12 people × 15 hours/week × $100/hour × 50 weeks = $900k annual value
• Tool cost: $500/month × 12 = $6k/year
• Net ROI: 149x

When to Consider Surton’s AI Implementation Services

If you’re navigating:

• Team-wide AI adoption (not just individual use)
• Security/compliance requirements (healthcare, finance, etc.)
• Integration with existing workflows and tools
• Measuring and optimizing ROI across the organization
• Training and change management at scale

Surton offers AI Implementation services designed for services businesses and technical teams. We handle setup, security configuration, prompt engineering, and team training.

---

Related Resources

• The 2025 Surton Hiring Playbook — AI-assisted hiring workflows
• AI Implementation Guide — Comprehensive guide for services businesses
• How AI Fits Into Day-to-Day Work (Original Newsletter) — The Blueprint edition

---

This is Surton’s definitive 2025 internal AI workflow system. For the original newsletter version, see The Blueprint.