How to Run an AI Audit That Actually Delivers ROI (Not Just Hype)

Win of the Week

Zohour just closed another $5K audit, turning a free audit into a six-figure opportunity and landing a paid audit that’s now leading to a multi-phase, long-term client. Momentum is real.

Ready to land your first (or next) audit? Book a 1:1 strategy call here

Why 80% of AI projects fail, and how to be the consultant who’s in the winning 20%

December 13, 2025
Read Time: 10 minutes

Let’s get real: 80% of AI projects fail. That’s not my stat, that’s McKinsey. Most businesses are implementing AI backwards and haemorrhaging money. They see a demo, get excited, plug it into their operations, and hope it works. Three months later, the tool is collecting dust and they’re back to manual processes.

But here’s the thing: there’s a 20% that’s getting massive returns, 100%+ ROI, scaling without hiring, and eliminating bottlenecks that were bleeding them dry. What’s the difference? They start with diagnosis, not technology. They find the expensive problem first, quantify what it’s costing, and only then prescribe the solution.

That’s the difference between being the consultant who gets paid for results and the one who gets ghosted after the first project.

Let’s break down the real process, step by step.

Step 1: Define Clear Outcomes (Not Vague Wishes)

Most discovery calls start with, “We need AI,” or “We want to automate our operations.” That’s not an outcome. That’s a wish. If you can’t measure it, you can’t improve it. If you can’t prove ROI, you’re not getting paid for the implementation.

Here’s what a clear outcome actually looks like:

• Cut lead response time from 4 hours to under 5 minutes.

• Increase conversion rate from 20% to 24%.

• Save property managers 12 hours per week on repetitive tasks.

• Respond to 100% of after-hours calls instead of 0%.

Notice the pattern? Each one has a current state, a target state, and it’s measurable.

When I worked with a property management team, they said, “We need better communication with tenants.” That’s not an outcome. After some digging, we translated it to: “Reduce tenant question response time by 68% and increase lead retention from 70% to 90%.” Now I can build a solution, measure results, and prove ROI.

How do you extract these outcomes? Ask outcome-forcing questions:

• What’s the most expensive problem you’re dealing with right now?

• If we could solve one thing that would move the needle on revenue, what would it be?

• Where are you losing money that you know about but haven’t fixed?

Don’t ask what they want to automate. Ask where they’re bleeding money. Every outcome must have a dollar amount attached. “Faster” isn’t enough. “Faster, which means we capture 20% more leads worth $161,000 annually,” that’s the shift.

Step 2: Map How the Business Actually Works

This is where most agencies and consultants fail. They talk to the CEO or founder, get a high-level overview, and start building. But the CEO knows the destination, not the road. The people actually doing the work know every pothole, detour, and breakdown.

On every client engagement, I don’t just talk to the executive sponsor. I get on 10+ calls across the organization, department heads, individual contributors, the people in the trenches. That’s how you uncover the real bottlenecks.

Don’t just ask, “What do you do?” People will give you the sanitized version. Instead, ask, “Walk me through yesterday morning when you got to the office. What did you do first? Then what? Then what after that?” When you ask about yesterday, they tell you what actually happens, not what’s supposed to happen.

I once worked with a sales team where the VP said, “Our reps spend their time selling.” But when I talked to the reps, I found out they spent 9–11 a.m. manually building lead lists, checking LinkedIn, cross-referencing Salesforce, copying data field by field, two full hours before they even started selling. The VP had no idea.

Once you have these interviews, visualize the workflow. Use a whiteboard, Excel, or even a napkin, just boxes and arrows showing how work actually flows. When you see it laid out, the bottlenecks scream at you.

Document everything: time spent, tools used, handoffs, where things get stuck. This becomes your roadmap for where AI can actually help.

Step 3: Break Down Every Task to Its Atomic Parts

Something like “follow up with leads” sounds like one task. It’s actually seven:

1. Open the CRM.

2. Filter for leads from the last 24 hours.

3. Check if they match the ideal customer profile.

4. Read their submission details.

5. Draft a personalized response.

6. Send the email.

7. Set a follow-up reminder.

Now you can evaluate each step individually. Can AI open a CRM? Yes. Can AI check ICP match? Yes. Can AI draft a response? Maybe, if the inputs are structured. This is the level of granularity you need.

I once worked with an heir-hunting company. Their “problem” was researching deceased property owners. Too broad. The atomic breakdown revealed the real workflow: pulling county tax records, filtering by value, searching ancestry databases, cross-referencing obituaries, verifying identities. Steps 5 to 11 were the real bottleneck, 15 minutes per lead, mostly on verifying identities. That’s where AI could help.

On every call, keep asking, “Then what happens?” until you can’t break it down any further. Document time per step, tools, and decisions.

Step 4: Identify What AI Can Actually Do

Not every task is AI-suitable. Here’s the four-question filter I use for every atomic task:

1. Is the input structured? (Forms, emails, database records = yes. Vague verbal requests = no.)

2. Is the output predictable? (Standard responses, data extraction = yes. Creative strategy = no.)

3. Are decisions rule-based? (If/then logic = yes. Complex judgment = no.)

4. Is it repeated often enough? (Daily/weekly = yes. Once a quarter = no.)

If you get yes to all four, you have a prime AI candidate.

Example: For a property management company, we automated 78% of tenant questions (standard categories, predictable answers, rule-based, repeated daily). Result: 15 hours per week dropped to 3 hours per week. The property managers only handled the 22% of complex issues.

AI handles 80% of routine, structured, rule-based work. Humans handle the 20% of edge cases requiring judgment. If you try to automate everything, you’ll run into issues.

Step 5: Prioritize for Impact (The ROI Matrix)

By now, you’ll have 15 to 20 automation opportunities. You can’t build them all at once, and you shouldn’t.

Plot everything on a 2x2 Opportunity matrix:

• X-axis: Implementation difficulty

• Y-axis: Business value

This gives you four quadrants:

• Quick Wins: Low difficulty, high value. Start here.

• Big Swings: High difficulty, high value. Phase 2 or 3, after trust is built.

• Low Priority: Low difficulty, low value. Probably never build.

• Money Pits: High difficulty, low value. Avoid completely.

Example: For a property management client, our quick win was lead auto-collection. Marcus was spending 45 minutes every morning manually checking six lead sources. We built a Zapier integration, basic, no AI needed. Saved $6,656 per year, took two weeks.

Another quick win: Tenant FAQ automation. 30 hours per week across two managers, solved with an AI assistant trained on 12 categories. $42,000 saved per year, three weeks to implement.

Big swing: Predictive maintenance scheduling. High difficulty, high value, but only after quick wins proved ROI.

Always start with quick wins that prove ROI in weeks, not months. Once they see results, the big projects sell themselves.

Step 6: Calculate Real ROI (The Math That Closes Deals)

This is what makes executives cut the check. You need to put dollar amounts on everything.

The formula:
Time wasted per person × number of people × days per year × loaded hourly cost = annual cost of inefficiency.

Example:
B2B software company, SDRs spent 2 hours/day building lead lists. 8 SDRs × 2 hours × 260 days × $40/hour = $166,400 per year wasted. A $60,000 automation paid for itself in 5 months and saved $100K+ every year after.

But cost savings is only half the equation. Quantify lost revenue too. For a property management company, slow lead response was losing them $90,000 per year in revenue. Add up all the waste, and suddenly your $81,000 solution looks like a bargain.

Show the payback period, the ROI, and the total annual impact. When you can show an executive they’re leaving $180,000 on the table and your solution pays for itself in under a year, the decision is obvious.

How to Start Small and Scale Up

If you’re new, don’t start with a four-week, $60K audit for a 30-person company. Start with a two-week opportunity assessment for a smaller business.

Week 1: Discovery, 3 to 5 stakeholder interviews, focus on pain points and bottlenecks.
Week 2: Solution design, plot bottlenecks on your Opportunity matrix, validate with stakeholders, create top three recommendations and a 90-day roadmap.

Charge $5–10K for businesses with 10–20 employees. Build your case studies, refine your process, and scale up as you get better.

Small businesses want quick wins and immediate ROI. Mid-market companies want strategy, education, and a roadmap. Adjust your approach based on who you’re selling to.

Final Thought:
AI isn’t magic. The technology is the easy part. The real skill is diagnosing the expensive problem, quantifying the cost, and prescribing the right solution. That’s what separates the 20% who win from the 80% who waste money.

If you want to be in the 20%, follow this process. Start small, prove the model, and scale up as you build conviction and results.

See you next week,

– Andrew

P.S. Ready to build your own Vibe Consulting business? Book a 1:1 strategy call here to see if you're a good fit for my personal coaching program.