Frontline Workflow Diagnostics
Observe the actual shift: searching, duplicate work, confusing handoffs, unwritten exceptions, and avoidable supervisor pull-ins.
- workflow observation
- decision friction
- handoff mapping
- baseline metrics
I design the operational systems behind customer-facing work: the policies, knowledge, decisions, interfaces, and workflows frontline employees rely on. Less searching. Less guessing. More room for human judgment.
An employee is expected to find a rule buried in five tabs, interpret a policy nobody clarified, and keep a customer waiting, then gets graded on the outcome. Another training session will not repair a missing decision.
Observe real frontline work, identify unnecessary memory and search burden, clarify which decisions belong to the organization, and design support around the moment of action. AI is an option, not the starting point.
Practical frontline systems design: observe real work, organize approved rules, clarify decision authority, prototype better support, and test whether speed, accuracy, or confidence actually improves.
Observe the actual shift: searching, duplicate work, confusing handoffs, unwritten exceptions, and avoidable supervisor pull-ins.
Translate approved client rules into clear paths: what can proceed, what to verify, what to disclose, and what must be escalated.
Structure information by client, location, situation, and moment so finding the answer is no longer its own job.
Measure the current workflow, test a bounded improvement, and learn what changed before funding a rollout.
Design focused decision-support interactions that fit the work already happening instead of adding another dashboard.
Decide what systems should remember, retrieve, check, and coordinate, and where people must keep judgment, relationships, and accountability.
I begin by observing the work, not selling another tool. The people doing the job belong at the center of the design.
The person should stay where context, responsibility, trust, creativity, and judgment actually matter.
The system should absorb the burden of remembering, finding, reconciling, and coordinating information around the work.
Each engagement moves toward concrete support: a decision map, governed knowledge model, prototype, or bounded pilot. We test against real tasks instead of congratulating ourselves on a polished presentation.
Not every team needs new software. Start with one recurring task, one costly decision, or one policy-heavy workflow.
Observe one high-friction workflow and document searches, rule interpretation, interruptions, exception paths, and rework.
Turn approved rules and observed work into a worker-centered interaction model, knowledge structure, and testable prototype.
Test a bounded frontline workflow with baseline measures, human oversight, approved source rules, and clear stop/go criteria.
Agent First Consulting applies the method to frontline operations. Agent First is a separate decision-support product built around the same philosophy. Other experiments are not consulting services.
Start with the frontline worker and redesign the responsibility split between people and systems.
Frontline diagnostics, decision architecture, knowledge design, system prototyping, and validation.
A separate decision-support platform for rules-heavy teams, beginning with multi-client contact centers.
Illustrative exercises that reveal decision friction, supervisor tax, and information burden.
The point is not to comment on every AI headline. I am developing a specific research program around work design, operational burden, human control, decision environments, and the division of responsibility between people and systems.
Redesign the responsibility system before automating it. The model should come after the work is understood.
Context. Rules. Retrieval. Decision. Communication. Action. Human control. Seven layers for finding what the worker is being forced to figure out.
These are not slogans for their own sake. They are working claims I am testing through products, consulting engagements, Labs, and public writing.
Interactive exercises expose context switching, ambiguity, retrieval burden, and escalation cost so the philosophy can be experienced rather than merely described.
“Why are people still being asked to remember things software already knows?”
That question became the beginning of Agent First. I became interested in the enormous amount of human work that is really searching, remembering, switching systems, reconstructing context, interpreting rules, and carrying procedural complexity in working memory.
My work starts there. Not with the newest model. Not with a feature list. With the person trying to do the job, and the question of what the surrounding system should be responsible for instead.
The result is part operating-model design, part product design, part knowledge architecture, and part applied AI. The goal is simple: make difficult work easier to do correctly without designing the humanity out of it.
Good. That's usually where the interesting work starts.
Tell me who does the work, what they must remember or find, which decisions slow them down, and what happens when the answer is unclear. You do not need a technical specification.