Workflow
50%
of knowledge workers are expected to need skills to work with, govern, or create agents by 2029.
Gartner, August 20251If every step still needs a person, you do not have a new workflow. You have a copilot.
Teams say humans stay in the loop and then leave the loop undefined. The person either rubber-stamps everything or redoes the work. Neither creates leverage.
How I see it
Put people on the decisions that matter
A human-plus-AI workflow is a division of labor. Software gathers context, drafts, classifies, and writes back to systems. People take the steps that require authority, taste, relationship, or risk acceptance.
The design work is to name those steps before the software is built. Which fields may be written without review? Which actions need a named approver? What evidence must sit next to the recommendation? What happens when the case is outside the boundary?
Good control points are few and expensive. A person who reviews every sentence is cheaper to replace with a better prompt than to staff forever. A person who accepts financial exposure, customer commitments, or legal risk is not optional.
The operating test is simple. After a month, are people spending time on the judgments you designed for them, or are they babysitting the software? If they are babysitting, the workflow is wrong.
How the engagement runs
What you actually buy.
A service is a sequence with an artifact at each step. It is not a transformation program you purchase as a bundle.
Map the unofficial path
Inbox, spreadsheet, chase, and the person who unsticks the case — not the SOP.
Cost per case and the real handoffs are visible.
A clean process map nobody recognizes.
Decide how far to change the work
Improve the current path, redesign it, or give one outcome to an agent.
The path matches the economics and the risk.
Automating leftover tasks on the old path.
Put people on the control points
Judgment, authority, and risk acceptance stay human. The rest can move.
Named steps are retired after the new path holds.
Every step still needs a person, so nothing got cheaper.
Do not buy this for
These engagements fail for predictable reasons.
Reviewing everything
A 100 percent review path is the old workflow with extra software cost. Sample, threshold, or escalate. Do not inspect every case by default.
Hiding the recommendation
If a person cannot see why the software suggested an action, they will either blindly accept it or ignore it.
Leaving escalation to chat
Exceptions need an owner, a queue, and a time limit. A Slack message is not a control point.
Before we start
Questions that decide whether to engage.
Which actions create irreversible customer, financial, or legal exposure?
Those actions stay human until the evidence and the error rate say otherwise.Which steps are slow only because a person is copying or formatting?
Those steps should not have a human in the loop.What evidence does the reviewer need to decide in minutes, not hours?
If the packet is incomplete, the human step will recreate the old delay.How is a disagreement between the person and the software recorded?
Without that record, you cannot improve the path.Who can widen or narrow the software's authority?
If nobody can change the boundary, the design will rot.
How we size it
Human time that should remaindecisions that require authority or risk acceptance × time to decide with a complete packet
If remaining human time is still most of the case, the software is not doing enough of the work.
This is the wrong engagement if
Do not start this engagement yet.
- Leadership wants full autonomy on a process that creates irreversible exposure.
- Nobody will name who is accountable when the software is wrong.
- The team will not change approval habits, even with a better packet.
- The work is relationship work that should stay human end to end.
A useful next step
Bring one expensive 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