Most AI pilots fail. Not from bad technology, but from the wrong problem, the wrong process, and no plan to reach scale.
I work with business owners and executives to find the handful of processes where agentic AI actually changes the economics of the business, and to avoid burning a year on the rest.
This isn't about adding AI to the process you already have. It's about redesigning the process.
For most of the last decade, software automated pieces of an existing workflow. Agentic AI can take on enough of the work that the workflow itself is worth rethinking from scratch, not just speeding up. That's a different question for most businesses to ask: not "where can we bolt on AI," but "how would we build this process today, if we were building it new."
Fewer steps, less handoff
Processes redesigned around agentic AI often need fewer people and fewer handoffs to produce the same result.
Fewer errors, more consistency
The work gets done the same way every time, with fewer of the errors that come from manual handoffs and fatigue.
More of the work gets done
Work that never got done because there wasn't time or headcount for it becomes worth doing.
Not another chatbot in a workflow nobody asked for.
Most "AI transformation" work optimizes individual productivity: a faster inbox, a code assistant that speeds up one developer, a summarizer bolted onto a tool nobody loved to begin with. It feels like progress and moves nothing on the P&L. I turn that work down. I look for the small number of processes where agentic AI changes customer experience, unit economics, or speed at an organizational scale, and I say so plainly when a business isn't ready for that yet.
Individual productivity tools
Assistants that save one person twenty minutes a day and disappear from the budget by Q3.
Organization-scale processes
Workflows where agentic AI changes a cost line, a cycle time, or a customer outcome the business actually reports on.
I turn work away
If the impact doesn't reach the business's economics, I'll tell you before you spend the budget.
The processes worth scaling are exactly the ones that need a policy governing them. That work isn't separate from this. It's the second half of it.
Finding the process is half the job. Governing it is the other half.
Once agentic AI is running a real process, someone has to define what it's allowed to do, who approves it, and where a human stays in the loop. Most businesses scale first and write the policy later, if at all. I build the policy alongside the process, and audit what's already running against it.
Policy
What the system is allowed to touch, who signs off on changes, and where a human has to stay in the loop.
Audit
A structured review of what's already live: what's controlling it, and where the gaps are, mapped to a named framework, not an internal checklist.
Remediation
A concrete, prioritized list of what to fix, not a slide of concerns with no owner attached.
Assessed and defined against named standards, not an internal checklist.
Two frameworks anchor the work:
NIST AI Risk Management Framework
A risk-based framework for identifying, measuring, and managing risk across the AI lifecycle: what could go wrong, how likely it is, and what to do about it before it does.
ISO 42001
The international standard for AI management systems: how an organization governs the development and use of AI on an ongoing basis, not just at launch.
The assessment measures where you stand against both. The policy defines how you close the gap and stay closed as the system changes.
Most AI pilots fail for one of two reasons.
Before you commit budget to the next one, ask which pattern it matches:
Low value, time-wasting
It's busy work wearing an AI label. Ask yourself:
- Does it save minutes for one person rather than move a number the business actually reports?
- Would the work simply shift to a different tool if this one disappeared tomorrow?
- Could you explain the business case without using the word "AI" at all?
Low impact, hard to scale
It works, but it will never reach the rest of the business. Ask yourself:
- Does scaling it depend on data or integration work nobody has started?
- Is the process it touches too small or too irregular to matter at scale?
- Would this still make sense if the team running it were twice the size?
If your initiative shows several of these signs, the conversation worth having isn't about the tool. It's about which process to work on instead.
Rich Wanden
My expertise is in the work itself: greenfield development, ERP implementations, maintenance and support, modernizing legacy systems, and now building agentic AI systems for enterprise clients. That range is what makes the judgment useful. I've watched software succeed and fail across the entire lifecycle, not just at launch, which is why the work is mostly elimination, not addition: finding the handful of processes where agentic AI actually changes the economics, and cutting the rest before they consume a budget on pilots that were never going to scale. That's the approach I bring to every engagement, most recently at World Tech Enterprises: agentizing processes from the core, not bolting AI onto the outside of what already exists.
I've applied the same discipline to the policy side: assessing and defining AI policy against the NIST AI Risk Management Framework and ISO 42001, the international standard for AI management systems, so agentic AI can run in production without becoming a liability. That judgment is built on thirty years working in transformation programs, first in organizational change management and delivery on SAP programs at Ernst & Young and Deloitte, then in the commercial and delivery side of technology consulting at CSC, and most recently at World Tech Enterprises. Agentic AI is the first technology in those thirty years that makes real process redesign realistic at scale. It's not an incremental step. It's a leap.
If the last AI pilot didn't move anything, the next conversation should start differently.
Run your initiative through the two patterns above, or skip straight to a call.