How I Structure Every AI Consulting Engagement (The 5-Phase Framework)

The fastest way to undermine a consulting engagement is to start without a framework. You end up doing ad hoc research, making scope commitments you cannot price accurately, and delivering work that lacks a coherent throughline.

I learned this the hard way in my first year of AI consulting. Projects ran long. Scope crept constantly. Clients were not sure what they were getting until they got it.

The fix was building a repeatable 5-phase engagement structure — and then using AI to accelerate every phase. Here is exactly how I run every engagement now.

Why Frameworks Matter for AI Consulting Specifically

AI consulting has a unique credibility problem. Clients have been burned by overpromised automation projects that delivered nothing. They are skeptical. A clear, documented framework signals professionalism and manages expectations before the work begins.

It also lets you quote fixed-price projects with confidence. When you know exactly what each phase costs in time and effort, you stop guessing on proposals.

Phase 1: Discovery (Weeks 1-2)

Goal: Understand the client's current state before proposing anything.

Key activities:

• Stakeholder interviews (using AI-generated question banks tailored to their industry)
• Current-state process mapping (I use Claude to turn interview transcripts into visual workflow descriptions)
• Technology audit (what tools they use, how data flows, where bottlenecks are)
• Pain point prioritization matrix

AI acceleration: I feed all interview transcripts and notes into a structured discovery prompt that outputs a Current State Summary doc in under an hour. What used to take a week of synthesis now takes an afternoon.

Typical duration: 8-12 hours (2 weeks elapsed)
Deliverable: Current State Assessment + Opportunity Inventory

Phase 2: Diagnosis (Week 2-3)

Goal: Quantify the problems and size the opportunity.

Key activities:

• Data analysis (volume metrics, error rates, cycle times — wherever the client has data)
• Gap analysis between current state and industry benchmarks
• Opportunity sizing (conservative/moderate/aggressive scenarios)
• Root cause analysis on the top 3-5 friction points

AI acceleration: I built a Diagnosis Prompt that takes raw data exports and outputs a structured Gap Analysis with opportunity sizing in three scenarios. A task that used to require Excel modeling for 6+ hours now takes 45 minutes.

Typical duration: 6-10 hours
Deliverable: Diagnosis Report with quantified opportunity size

Phase 3: Design (Weeks 3-4)

Goal: Build the solution architecture and get client buy-in before building anything.

Key activities:

• Solution option development (usually 2-3 approaches at different investment levels)
• AI tool/workflow selection and rationale
• ROI model (built on the Diagnosis data)
• Implementation roadmap with milestones
• Risk register

AI acceleration: The ROI model template and solution architecture framework are now AI-assisted. I brief the system on the diagnosis findings and get a draft architecture and financial model in about 90 minutes. I refine it from there.

Typical duration: 10-15 hours
Deliverable: Solution Architecture + ROI Model + Implementation Roadmap

Phase 4: Delivery (Weeks 4-12+)

Goal: Build and deploy the agreed solution.

Key activities:

• Prompt engineering and workflow development
• Integration work (connecting AI outputs to client systems)
• Pilot testing with real client data
• Iteration based on feedback
• Change management (getting the humans to actually use the new system)

AI acceleration: Every deliverable in this phase — training materials, workflow documentation, test scripts, change management emails — gets drafted through AI and refined. Speed-to-draft has roughly doubled.

Typical duration: Varies by scope (40-200+ hours)
Deliverables: Working AI systems + documentation

Phase 5: Documentation & Handoff (Final 1-2 weeks)

Goal: Make sure the client can run what you built without you. And position for the retainer.

Key activities:

• Operations manual creation (AI-generated from implementation notes)
• Training sessions (recorded, with AI-generated transcripts and summaries)
• 30/60/90-day optimization checklist
• Retainer proposal (optional: ongoing optimization support)

AI acceleration: The handoff documentation package — which used to take 2-3 days to write — now takes about 4 hours with AI assistance. I input the project notes and outputs; the system drafts the full ops manual structure.

Typical duration: 8-15 hours
Deliverable: Operations Manual + Training Materials + Retainer Proposal

The Business Case for Repeatable Frameworks

With this framework, I can now:

• Quote fixed-price projects accurately (because I know what each phase costs)
• Set client expectations up front (they know exactly what they are getting)
• Delegate phases to subcontractors without rebuilding from scratch each time
• Identify upsell opportunities naturally (Phase 5 retainer conversion rate: ~60%)

The framework also compounds over time. Every engagement adds examples, prompt refinements, and calibration data that makes the next engagement faster.

If you want the full framework with the actual AI prompts I use in each phase, I have documented everything at thewedgemethodai.com.

---

What does your consulting engagement structure look like? Do you run fixed-price or time-and-materials? I am curious how others handle scope management. Share in the comments.