Here’s the bottom line up front: your company is possibly spending six figures on AI tools while running on workflows designed when fax machines were cutting-edge. The problem isn’t ChatGPT, Notion AI, or your automation platform. It’s that you’re trying to put a smart home system into a house with 1995 wiring. Until you excavate and redesign your foundational workflows, every AI tool you add might as well be just expensive digital wallpaper. Ouch, right?
Over the past six months I’ve spoken to plenty of frustrated business owners. One in particular had gone all out on an AI-powered document processing system and, for a small business, the expense was substantial. They’d been promised a huge reduction in processing time, and the result six months later saw them saving only a few hours per week.
Was the AI working? Yes, perfectly well. But it was still taking three days for a chain of people to approve each invoice. The problem wasn’t the AI. The problem was the approval process. And it had been designed in 1997.
Welcome to behind the curtains of AI implementation and the reality many would prefer to ignore.
The Emperor’s New Workflow
Here’s what’s happening in meeting rooms right now: companies are hemorrhaging money on AI tools that promise transformation. Autonomous agents. Smart assistants. Predictive analytics. The demos are dazzling and the promises profound.
The reality? Disappointing.
Not because the technology doesn’t work. It does. But because it’s being bolted on to organisational processes that were inefficient even before we had email. Think about your expense approval workflow. How many systems does it touch? How many people need to sign off? How many times does someone manually copy data from one tool to another?
Now think about when that workflow was designed. High chance it was before smartphones existed. And it was probably overkill even then.
Why Legacy Workflows Are Kryptonite for AI
I consulted with a business last year that made a significant investment in AI-powered lead generation, incorporating chatbots, predictive enrollment models, and a number of automated email sequences. They dedicated a large part of their budget to implementing this technology and training staff on its operation.
The outcome? Their email list grew substantially, meeting expectations. However, their actual conversion and sales figures remained stagnant.
Why? The advanced lead generation system was undermined by their internal processes. Once a prospective customer expressed interest, they entered a highly manual and disjointed workflow. This involved several separate databases that didn’t communicate efficiently, each managed by a different staff member, causing significant delays and a fragmented customer experience.
Essentially, the AI effectively generated leads as promised, but the inefficient and outdated follow-up workflow acted as a bottleneck, preventing those leads from converting into sales.
The fact of the matter is: AI won’t create new workflows. It accelerates whatever workflows you already have. If your workflows are broken, AI just helps you fail faster.
Think of it like installing a turbocharger on a car with a cracked engine block. You’re not going anywhere faster. You’re just going to blow up more spectacularly.
The Workflow Archaeology Method: Excavating What’s Really There
Many organisations don’t actually understand their processes. They know what the manual says. They know what the org chart implies. But the actual way work moves through the business? That’s buried under years of workarounds, and “that’s just how we’ve always done it.”
Before you invest another cent in AI tools, you need to study what’s already happening in your business. A process that has three distinct phases.
Phase 1: EXCAVATION (Uncovering the Buried Inefficiencies)
You can’t look at this as a theoretical exercise. You need to map how work actually flows, not how you think it flows.
Start with a high-stakes process. Procurement. Hiring. Project approvals. Client onboarding. Pick something that impacts revenue or costs significantly.
Then ask these diagnostic questions:
- How many tools does this workflow touch? If the answer is more than three, you’ve got a problem to solve. A hiring process that touches eleven different systems? No AI tool can save that.
- How many people need to approve this? More than two approvals is usually a red flag. It means decision rights aren’t clear, or there’s a trust deficit, or political territoriality. AI can’t fix organisational politics.
- If this process disappeared tomorrow, what would actually break? You’d be shocked how many workflows exist purely because they existed last year. Legacy inertia masquerading as compliance.
- Can you articulate the success criteria in one sentence? If you can’t clearly define what “good” looks like, how will an AI know? This is where most automation projects derail, vague objectives meet precise technology.
An example of this would be a manufacturing company excited to implement AI-powered inventory forecasting. During excavation, they discovered their inventory data lived in three systems that didn’t talk to each other, updated on different schedules, and used incompatible product codes. The AI would have been forecasting based on fiction.
Spending time first cleaning the foundation? Then the AI should work brilliantly.
Phase 2: CLASSIFICATION (Deciding What to Do With What You Find)
Once you’ve excavated the real workflows, you need to classify them. Not everything deserves to be automated. Some things deserve to be eliminated.
Here are four categories to work with:
RETIRE: Processes that exist due to legacy compliance, outdated regulations, or organisational politics, not actual value.
For example, imagine a financial services firm still requiring three physical signatures on contracts because “that’s what legal said” in 2003. The legal landscape had changed. The process hadn’t. That’s a retirement candidate, not an AI target.
REDESIGN: Workflows with good intent but broken execution. These are processes that should exist but are tangled in unnecessary complexity.
Your invoice approval process possibly falls here. You need financial controls, yes. But do you need five approval layers and manual data entry into four systems? No.
RETAIN: The rare processes that actually work efficiently. These are precious. Learn from them. Don’t fix what isn’t broken.
AI-READY: Workflows that are clean, standardised, and measurable enough to benefit from automation. These are your starting points for AI implementation.
I’m thinking that only about 20% of workflows fall into the “AI-Ready” category initially. The rest need work.
Phase 3: RECONSTRUCTION (Building Workflows Worthy of AI)
This is where the real transformation happens. And it happens before you turn on any AI tool.
Step 1: Simplify First
Remove the barnacles. Every unnecessary approval. Every redundant handoff. Every time someone copies data from one system to another.
Step 2: Standardize
AI thrives on consistency. If your sales team has twelve different ways of logging a customer interaction, no AI can make sense of that chaos. Create consistent inputs. Define standard outputs. Build templates. This isn’t sexy work. But it’s essential.
Step 3: Measure
Define what “good” looks like before you automate.
- What’s the current cycle time?
- What’s the error rate?
- What does success cost in staff hours?
You can’t improve what you don’t measure. And you can’t know if your AI investment worked if you didn’t baseline it.
Step 4: Then—and Only Then—Augment
Now you’re ready for AI. Now you have clean workflows, clear success criteria, and measurable baselines. Now AI becomes transformational instead of just expensive.
The Hidden Cost of Skipping Archaeology
I know what you’re probably thinking: “This sounds like a lot of work before we even get to the AI part.”
You’re right. It is. But here’s what skipping it costs:
Failed implementations. Gartner estimates that 85% of AI projects fail to deliver on their promised value. The primary reason? Poor data and process quality, not technology limitations.
Wasted investment. The average company now uses 110 SaaS tools. How many are actually integrated into efficient workflows? How many are used because they’re supposed to solve a problem, but the problem was never properly diagnosed?
Staff burnout. When you layer AI onto broken workflows, you don’t reduce workload. You just add “AI babysitting” to the list of things your staff has to manage. They’re now copy-pasting between the old system, the new AI tool, and the manual backup process when the AI inevitably hits a workflow snag it can’t navigate.
Competitive vulnerability. While you’re debugging why your expensive AI stack isn’t delivering ROI, your competitor who did the archaeology first is actually transforming their operations.
The Question That Changes Everything
Here’s a diagnostic question I ask every business owner frustrated with their AI:
“If you removed all the AI tools from your organisation tomorrow, would your workflows be efficient?” If the answer is no, stop buying AI tools. Start excavating.
Because the bottleneck in your organization isn’t technological. It’s architectural.
You don’t have an AI problem. You have a workflow problem that AI is exposing.
Where to Start: A Suggested 30-Day Workflow Audit
You don’t need to overhaul your entire business overnight. Start with one high-impact workflow:
Week 1: Excavate
- Map the actual process, not the theoretical one
- Count the tools, people, and handoffs
- Document where it breaks down
Week 2: Classify
- Retire, Redesign, Retain, or AI-Ready?
- Be ruthless, most processes need more help than you think
Week 3: Simplify
- Remove unnecessary steps
- Standardize what remains
- Define success metrics
Week 4: Pilot
- Test the redesigned workflow without AI first
- Measure the improvement
- Then consider where AI adds value
The Truth of the Matter
The future of work isn’t just about having the latest AI tools. It’s about having workflows worthy of them. Others aren’t winning because they have better AI. They’re succeeding because they built better foundations. The question isn’t “What AI should we buy?”
The question is “Are we ready for the AI we already have?” Many companies aren’t. But the ones who do the archaeology, who excavate, classify, and reconstruct, they’re the ones who’ll actually transform.
The rest are just building smart homes on houses with bad wiring. Want help to identify which processes need redesign before AI implementation? The businesses that start here see 3-4x better ROI on their AI investments. Schedule a and let’s build the foundational capabilities required for successful AI adoption.