The AI Playbook

Real numbers:

For most quick wins (meeting transcription, research acceleration, first drafts):

• $20-50 per user per month (ChatGPT Plus, Claude Pro, or similar)
• For a 20-person team: $600-1,000/month

ROI calculation is straightforward:

• If each person saves 5 hours per week at $75/hour
• That's $375/week per person = $7,500/week for 20 people
• Annual value: $390,000
• Annual cost: ~$12,000
• ROI: 32.5x

The key: Start with one measurable workflow. Track time saved, quality improvements, and consistency gains. Then you have real data to justify scaling.

Most businesses see payback in 4-8 weeks on their first systematized workflow.

This is the most important question, especially for banks and professional services.

What you need to know:

Enterprise Agreements:

• ChatGPT Enterprise, Claude for Work, and similar enterprise tiers explicitly state: your data is NOT used for training
• You own your inputs and outputs
• SOC 2 Type II compliance, encryption at rest and in transit
• Business Associate Agreements (BAAs) available for HIPAA compliance

What NOT to do:

• Don't use free consumer versions for business data
• Don't feed in customer SSNs, account numbers, or regulated data
• Don't put anything in AI you wouldn't put in an email

Practical approach:

• Start with non-sensitive workflows (meeting notes, research on public companies, marketing drafts)
• Use enterprise agreements before handling customer data
• Build a clear policy: what can/cannot go into AI systems
• Train your team on data handling protocols

For highly regulated industries (banking, accounting, healthcare):

• Look at on-premises or private cloud deployments
• Some vendors offer air-gapped solutions
• Your compliance team should review contracts before implementation

The liability question: Treat it like you would outsourcing to a vendor. Due diligence, contracts, insurance, and clear policies.

This isn't a reason to avoid AI—it's a reason to implement it thoughtfully.

No. You don't need AI specialists. You need good managers.

The truth: Your existing team can handle this. The workflows described in the keynote don't require coding or technical expertise.

What you DO need:

1. A Process Owner:

• Someone who owns the workflow you're systematizing
• They know the current process intimately
• They have the authority to change how work gets done
• This is a manager or department head, not an IT person

2. Your IT team's involvement:

• Security review of tools
• Enterprise account setup
• Integration with existing systems (if needed)
• This is support, not leadership

3. Training approach:

• 2-hour workshop on the specific workflow
• Hands-on practice with real examples
• Ongoing feedback sessions
• Document what works

You don't hire for this. You train for this.

The accounting firm that's been around 60 years? You have people who've managed process improvements for decades. This is the same thing.

If you want outside help: Hire a consultant for the FIRST workflow to help you design the process. Then your team runs with it. Don't outsource your learning.

The most practical answer is to start with one of these three.

1. For most business workflows:

• ChatGPT Plus ($20/month) or ChatGPT Enterprise (custom pricing)
• Best for: Writing, research, analysis, brainstorming
• Most versatile for general business use

2. For detailed analysis and longer documents:

• Claude Pro ($20/month) or Claude for Work
• Best for: Processing long reports, detailed analysis, maintaining context
• Excellent for professional services

3. For Microsoft-heavy shops:

• Microsoft Copilot (included in M365 or standalone)
• Best for: Teams already using Office 365
• Native integration with your existing tools

For specialized needs:

• Meeting transcription: Otter.ai, Fireflies.ai, Fathom
• Workflow automation: Make.com, Zapier (for later, after you have basics down)

Recommendation: Pick ONE. ChatGPT Plus for most businesses. Run your 90-day challenge with that single tool. Don't tool-shop until you've mastered systematic use of one.

Tools matter far less than process. A mediocre tool used systematically beats a great tool used randomly every single time.

That's why human review is non-negotiable in the process.

Think about it like this: Would you let an intern send a client proposal without review? Of course not.

Same principle applies here.

Build review into the workflow:

• AI drafts, human reviews and approves
• AI researches, human verifies key facts
• AI analyzes, human makes the decision

For high-stakes outputs:

• Two-person review process
• Fact-checking protocol
• Clear accountability (a human signs off)

Document your process:

• "All AI-generated client communications must be reviewed by [role]"
• "Financial analysis must be verified against source data"
• "Legal documents require attorney review"

The risk isn't using AI. The risk is using it without process.

Your professional liability insurance should be informed about AI use, and they'll likely require you to demonstrate systematic oversight. Which you should have anyway.

They're probably not. Here's why...

The data shows:

• 78% of businesses have adopted AI somewhere
• Only 28% are using it systematically
• Most are in the same "random experimentation" phase

The real question isn't "Are we behind?" It's "How fast can we move from random to systematic?"

Your advantage: You have decades of process management experience. You know how to implement change. You know how to train teams. You know how to measure results.

The companies that have been playing with AI for two years but haven't systematized it? You'll pass them in 90 days with a disciplined approach.

The window is open. But the ones who move decisively in the next 6-12 months will have a compounding advantage that becomes very hard to catch.

Start with one workflow. Pick the one that's:

• High frequency (happens constantly)
• Painful (everyone hates doing it)
• Measurable (you'll know if it worked)

Then:

1. Map it - document every step
2. Insert AI where it makes sense
3. Train a small group (3-5 people)
4. Run it for 30 days
5. Learn and adjust
6. Scale to the full team

That's it. You don't need a grand strategy on day one. You need one win that proves the model and builds momentum.

Many businesses have the talent in-house to do this. If you want a strategic partner to accelerate that first cycle and ensure you build the muscle correctly, that's where we can help.

Either way—start this week.