How We Plan, Scope & Sell AI Projects in our 100k+/Mo AI Agency

Key Concepts

AI automation projects, project scoping, system design, development, deployment, optimization, problem definition, outcome definition, triggers, input data, software integration, volume, budget, feasibility, reverse engineering, breaking down projects into phases, diagramming, proposal creation, subscription-based pricing, building and testing, error handling, deployment, client communication.

Project Scoping

• Main Goal: To gain clarity on the problem, objectives, requirements, and metrics for the automation.
• Key Points:
  • Focus on getting necessary information, not all information, to judge feasibility and design a system.
  • Aim to complete scoping within the first two meetings to expedite the proposal process.
  • Smaller time window between the first call and proposal increases conversion likelihood.
• Necessary Details:
  1. Defining the Problem: Understand the underlying business problem or need, not just the client's proposed solution.
     • Example: Client wants a voice agent for inbound leads, but the real problem is low conversion rates.
  2. Defining the Outcome: Get specific examples and requirements from the client to reverse engineer the system and align expectations.
     • Example: Ask for sample personalized emails to understand the desired output of an email automation system.
  3. Defining Triggers and Input Data: Determine when the automation should run and what data is available.
     • Importance of identifying optional data fields to avoid system failures.
  4. Software Integration: Identify necessary software and their APIs to assess feasibility and create onboarding documentation.
  5. Volume and Budget: Define how often the automation will run and the budget for usage costs to ensure ROI.
• Example: Coaching business with overwhelmed sales reps due to LinkedIn ad leads.
  • Problem: Sales reps overwhelmed, slow follow-up times, impacting conversion rates.
  • Client's Solution: AI voice agent.
  • Outcome: Qualified leads booked automatically, unqualified leads receive personalized email with a low-ticket offer.
  • Triggers/Data: LinkedIn lead form submission (name, email, LinkedIn URL, company URL).
  • Software: LinkedIn Ads, Go High Level (CRM), Synflow (optional voice agent).
  • Volume/Budget: 150 leads per day, budget not a major concern due to potential savings from not hiring more sales reps.
• Feasibility:
  • Newcomers often want to scope out everything to ensure feasibility, but it's better to take a calculated risk.
  • Show trust and confidence, figure things out during the process, and offer a refund in the worst-case scenario.
  • Client projects are where you learn the most.

System Design

• Main Goal: Design and establish a process before automating it.
• Key Points:
  • Reverse engineer from the desired outcome.
  • Break down large projects into smaller parts or phases (sub-projects).
  • Focus on the shortest path to deliver value or a "win" to the client (MVP approach).
• Steps:
  1. Reverse Engineering: Start with the desired outcome and work backward to determine the necessary steps.
     • Example: To get qualified leads booked and unqualified leads followed up, you first need a lead qualification/scoring system.
  2. Breaking It Up: Divide the project into smaller, manageable phases.
     • Example: For the coaching business, break it into lead research, lead qualification, and voice agent/email sequences.
  3. Prioritization: If multiple automations are desired, start with the one that's quickest, easiest, and delivers the highest leverage.
  4. Diagramming: Use diagramming software (FigJam, Whimsical, draw.io, Lucidchart) to visualize the system and understand dependencies.
• Example (Coaching Business):
  • Phase 1: Lead Research
    • Outcome: Enriched CRM with relevant data points for sales reps.
    • Trigger: LinkedIn ad submission.
    • Input Data: Name, email, LinkedIn URL, company URL (optional).
    • Process:
      • Use a LinkedIn scraper (e.g., Apify) to get lead summary, current job title, and years of experience.
      • If company URL is available, scrape the website and/or company LinkedIn page for company size and summary.
      • Use a language model (LLM) to extract data points.
  • Phase 2: Lead Qualification
    • Dependency: Requires defining qualification criteria.
    • Outcome: Updated CRM with lead score.
  • Phase 3: Voice Agent/Email Sequence
    • Dependencies: What happens if someone doesn't answer? Where does it need to update in the calendar?
• Proposal Creation:
  • Include a breakdown of phases with estimated timeframes.
  • For each phase, specify the action, expected output, and dependencies.
• Pricing:
  • Subscription-based pricing is recommended for flexibility and easier iteration.
  • Avoids the friction of re-quoting for every change or addition.
• Onboarding:
  • Explain how to integrate the software and set up integrations in the client's accounts.

Development

• Main Goal: Build, test, and debug the automation system.
• Key Points:
  • Build and test step-by-step, ensuring each module works before adding the next.
  • Maintain an MVP mindset and avoid over-optimization.
  • It's okay if the automation fails on a small percentage of cases, as long as it adds significant value overall.
• Steps:
  1. Building and Testing: Ensure each step works before moving on.
  2. Debugging: Test on 10-100 examples and improve the system.
  3. Error Handling: Add retry mechanisms for API modules to handle failures.
     • Use built-in error handlers in no-code automation software (e.g., "break" module in Make.com, "retry on fail" in n8n).
• Error Handling Examples:
  • Make.com: Use "break" modules with a delay to retry API calls.
  • n8n: Enable "retry on fail" in API modules with a maximum number of retries and a wait time.

Deployment and Optimization

• Main Goal: Deploy the automation and continuously improve it based on real-world usage and feedback.
• Key Points:
  • Deployment is often just the beginning.
  • Expect a cycle of improvement and iteration based on client feedback.
  • After initial deployment, clients often have ideas for improvements and additions.
• Client Communication:
  • Open a Slack channel with the client for direct communication.
  • This is crucial for understanding their business and addressing their needs.
  • Limit requests to once a week to avoid overwhelming the client.
• Next Steps:
  • Once the automation is running smoothly, continue with the next automation or phase.
  • Expand on the current automation based on client feedback.

Synthesis/Conclusion

The video outlines a four-step framework for planning and delivering AI automation projects: project scoping, system design, development, and deployment/optimization. The framework emphasizes the importance of understanding the client's problem, defining clear outcomes, and breaking down projects into manageable phases. It also highlights the need for continuous improvement and open communication with clients. By following this framework, individuals and agencies can increase the success rate of their automation projects and improve client satisfaction. The speaker advocates for a proactive, iterative approach, encouraging newcomers to take calculated risks and learn from real-world client projects.