Apply
Most companies do not have an agent problem. They have a workflow problem.
48%
of organizations introduced AI without redesigning the workflows or roles it sits in.
Deloitte AI Institute Pulse Check, 20261The work already runs through email, spreadsheets, CRM, ERP, documents, and manual coordination. Adding a model on top of that pile does not create leverage.
How I see it
Agentic Workflows
The case already travels. Email, spreadsheet, CRM, someone who remembers to check ERP. Every handoff starts over. That is why putting a model on the pile does nothing. The queue is still a queue.
An agentic workflow cuts that relay. The agent carries the routine work across systems and stops when a person has to take the risk, make the relationship call, or approve the exception.
I have sat through forty-step maps that never mention the customer is waiting on one result. The steps that only exist to ferry information between systems are the ones to take. Leave the authority where it is.
Do not agent the operating model. One expensive flow is enough. Draw it before and after: cycle time, touches, exceptions, cost per completed case. If that picture needs a program office, the scope is already too large.
Common mistakes
What teams usually get wrong.
Mapping tasks, not the outcome
A 40-step process map can hide the fact that the customer is waiting on one result.
Leaving every approval in place
If the agent still needs a person after every step, you automated typing, not the workflow.
No before/after baseline
Without cycle time, touches, and cost per case, the redesign cannot prove it worked.
A useful diagnostic
Five questions before you fund the work.
How many people touch the case before it is done?
More than four is often a handoff problem, not a talent problem.Where does work wait the longest?
Queues are usually more expensive than the active work.Which steps exist only to move information between systems?
Those are the first agent candidates.Which steps require authority or risk acceptance?
Those become control points, not automation targets.Can you draw the future flow in one sitting?
If the future state needs a program office, the scope is too large.
Economic model
Workflow burdenpeople × loaded cost × percent of time on the workflow = annual economic surface area
A conservative 25% improvement on a $540K surface area is $135K. That is enough to justify a focused engagement.
Three credible paths
How far should you go?
Do not force one solution. Choose the path the economics, the risk, and the organization can support.
Improve the work you already have
Keep the process mostly intact and use AI on the bottlenecks that create delay, rework, or follow-up.
The workflow is already sound and a few steps create most of the friction.
Gains are usually incremental. The operating economics do not change much.
Redesign the workflow around AI
Question every handoff, queue, and duplicate step, then rebuild the process around what AI can now do.
The process grew over years and coordination now costs more than the work itself.
Requires process change, clearer ownership, and a willingness to retire old steps.
Put an agent on a high-value outcome
Give an agent responsibility for one valuable result across systems, with humans at the control points that require judgment, authority, or risk acceptance.
The workflow is high-value, variable, multi-step, and worth engineering for production.
Needs stronger architecture, evaluation, controls, and monitoring. This is not a prompt project.
When this is the wrong next step
Do not fund an agent here.
- The process is already short, owned, and mostly system-driven.
- The delays are caused by external parties you cannot redesign.
- Leadership will not let you retire a step, only add a tool beside it.
A useful next step
Bring one workflow. Get guided into production.
We guide the implementation, go deep on the technical path, and stay hands-on through operations — or tell you when a simpler answer is better.
Discuss an AI opportunity