Agentic workflows
Most of what you're being pitched as an agent should be a workflow.
Mid-market sales teams are drowning in AI agent hype, with nobody to tell them which problems need an agent, which need a plain reliable workflow, and where a human still has to sign off. That judgment is what I sell. Then I build whichever one the work deserves, a true agent or a plain automation, the boring, reliable way, and leave your team running it.
Two engagements. One scores your workflows with your team in the room. One builds the winner in your environment in one to two weeks. Both fixed price, both end with something you own.
The method
Five questions decide what gets built, what gets gated, and what gets left alone.
Every workflow on your team gets scored on the same five dimensions, live, with the people who run it in the room. The scores decide the sequence and the shape of each build, agent or plain workflow. Not the vendor's roadmap, not the demo that impressed the board.
Decomposability
Can the work be broken into steps a system can own, or is it judgment all the way down?
Data and system access
Does the raw material exist where a system can reach it, or does it live in someone's head and inbox?
Error tolerance and reversibility
What happens when it gets one wrong, and can that be undone? This is where the human gates go.
Volume and ROI
Does this run often enough, at enough cost, to be worth building at all?
Ownership and change readiness
Is there a person who will own this thing on Tuesday morning, or does it become orphaned automation?
Some of your workflows will score as leave it alone. That finding costs vendors money, which is why no vendor will give it to you.
The engagements
Score first. Then build the winner.
Facilitated Agentic Workflow Assessment
$9,500 fixed
Not a survey. Not a self-serve quiz. A working session where your workflows get scored in the room, in the argument between the VP who thinks the process is simple and the rep who knows where the bodies are buried.
Credit. Book a Build Sprint within 60 days and the full assessment fee applies against it.
Build Sprint
$24,000 fixed · $15,000 for a one-week scope
One team, one workflow, shipped to production with the human gates exactly where the assessment said they belong. Sometimes that is a true agent with judgment inside it; more often it is a plain automation that just runs. Built in the stack you already have, whether that is Microsoft Power Platform and Copilot Studio or custom code, so nothing new gets licensed to keep it alive. Your people sit with me while it gets built, so training happens as a byproduct of building, not in a classroom.
Outcome standard. Production or it does not ship. No pilots. Pilots are how automation goes to die in a sandbox.
The rules
What I will not build.
No agent where a workflow does the job.
The word agent gets used only when the work actually requires one. Excitement is not a requirement.
No build without an owner.
If nobody on your team will own the thing on Tuesday morning, the sprint waits until somebody will. That is a rule, not a preference.
No automation past its gates.
Anything that can send, spend, or promise on your behalf checks with a human first until the error math says otherwise. Reversibility decides autonomy.
No dependence by design.
The runbook and the pattern are the product. Nothing renews, nothing is licensed, and your team can run it the day I leave.
This is Stage 5 work on the maturity model: intelligence added to an engine that has earned it. If the foundation is not there yet, the honest answer is the RevOps systems line, and I will say so on the first call.
Start here
Thirty minutes. Bring the workflow you're being pitched on.
Tell me what the vendors are promising and I will tell you, on the call, whether it is an agent, a workflow, or a thing to leave alone. If the assessment makes sense after that, we book it.