Workflow Design

What Workflow Design is

Workflow design is the stage between naming an opportunity and building anything. One process is redrawn end to end. Its input, its output and the point where a person reviews the result are all settled before a prototype exists. The AI workflow page defines the thing being drawn, and this page covers the act of drawing it.

The stage has one central idea. Scope is the decision that determines the result, so it is fixed here and not discovered during the build. Everything else on the page follows from that.

Where it sits

The site’s home page shows an adoption sequence: training, an audit, opportunity identification, workflow design, AI automation and business outcomes. Workflow design is the fourth node, after the ranking and before the build.

It is not a sixth step of the audit. The audit runs five steps: map your processes, document how they run, identify the opportunities, test a prototype, and build and scale. Workflow design sits alongside them as the stage where one ranked opportunity is turned into a drawn process. Its start is mapping the exceptions, which the audit article says is where workflow design begins.

Redraw, do not bolt on

The audit article gives the clearest description of the act. Mapping the exceptions is the start of workflow design: redrawing the process around what AI can actually do, and not describing a tool bolted onto the process that was already there. The difference is real. A tool bolted on inherits every awkward step the process already had. A redrawn process removes or reshapes them.

That is why the design comes before a prototype. Testing a prototype of a badly drawn process tells you little. The point of drawing first is that the prototype then tests one clear thing.

The three things settled

The input is what starts the process. The output is what the process produces, in a form somebody can use. The review point is the place where a person checks the result before it goes any further. The workflow design page settles all three before anything is built.

They are also the test of whether a process is ready. A process with a fuzzy input, an output nobody can describe or no place for a person to check is not yet ready to build. Drawing it is how that becomes visible early, while it is still cheap to fix.

Scope is the Decision

The well-scoped workflow article says scope is the decision that determines the result, and that the same technology can produce a strong return in one case and a much weaker one in another depending on how narrowly it was applied. This page uses none of that article’s figures. It uses the argument.

The four principles the article names are a clear objective, a defined scope, built-in guardrails and a business impact that can be shown. Workflow design is where each of the four is written down. A defined single-purpose workflow gives a result that can be measured, and a broad rollout usually does not.

One Process at a time

The definition says one process at a time and not a department-wide rollout, and the one taken first is the one with the clearest input, output and review point. The opportunity identification step decides the order, and workflow design takes the first one and draws it properly.

Starting narrow is not caution for its own sake. A narrow workflow has something specific to compare against, so the result can be judged. The agentic AI page adds that a broad initiative has none of those, so nobody can say whether it worked.

Exceptions, not the Ideal path

The audit’s second step says a process is automated against its exceptions and not against its ideal path. Workflow design carries that into the drawing. The happy path, where the form arrives complete and the approval comes back the same day, is real, but it is rarely most of the work.

The design has to say what happens when the input is wrong, who is asked when a rule does not cover the case and where the work queues while someone waits. A workflow drawn only for the standard case breaks on the first one that is not, and then somebody does the work by hand on top of running the tool.

Every step on the Ladder

The article on automating a task against automating a decision sets out a ladder: Assist, Recommend, Execute and Never delegate. It says every real workflow is a chain of steps, and each step belongs on its own rung. The intake might be Execute, the triage might be Recommend and the final call might be Assist at best or Never delegate.

Sorting the steps is a design job. The article warns that treating the whole chain as one automation decision is how a task-shaped process quietly starts making decision-shaped calls nobody scoped it to make. Its example is candidate shortlisting, which is scoped as Recommend from day one with a named reviewer, so it does not drift towards Execute later.

Guardrails are drawn in

The means-here line on this page says oversight, logging, escalation and the point at which a person takes over are drawn in at this stage and not added once something has gone wrong. The automation page says the same about the build. These things are the difference between an automation the organisation can run and one that only whoever wrote it can run.

The reason is that a workflow with steps carried out on its own also has places where a mistake can travel unseen. Drawing the review point in early means the place where a person looks is decided on purpose, and not found by accident when something has gone out the door.

Design and Governance ask the same questions

The task-and-decision article has a section on where governance and workflow design meet. It says the two are often treated as separate disciplines, one for the compliance file and one for the delivery team, and that separation is where processes drift off the rung they were scoped for.

Its view is that a written AI governance policy is, among other things, a record of which rung each automated process sits on and who signed off on putting it there. Good workflow design already asks the same questions a governance review asks. Keeping them in separate documents, reviewed by separate people on separate schedules, is how a process ends up governed on paper and misclassified in production.

With the people who do the work

Workflow design is not done at a desk. The audit article says an audit checks skill and appetite alongside the process, by sitting with the people who do the work and watching how they actually do it and not how the process document says they do. It asks directly whether they would use the tool being proposed or quietly route around it.

That carries into design. A workflow that fits the process but not the people who run it goes unused, however well it was drawn. The people who run the work are part of the drawing, and their verdict later decides whether the prototype is built properly.

Who does it

Several members of the team list AI workflow designer among their roles. The Arshia Kahani page describes designing the workflow strategies that hold agents, data pipelines and infrastructure together. The Alex Frolo page describes designing the workflows around custom agents. The founder’s page describes leading the workflow design projects that follow an audit, deciding where AI and automation fit into the way a team already works.

The site publishes no other detail about how such a project runs, so this page does not invent any. It does not describe a deliverable, a length or a fee for the stage, and the audit article says building from a ranked opportunity is separate work with its own scope and guardrails.

A worked example

This is an illustration, not a real client. A small firm’s audit ranks its month-end report first. In workflow design the firm draws it. The input is three exports. The output is the finished report in the format the manager already uses. The review point is the person who builds it today, who checks the totals before it goes anywhere.

The firm then writes down the exceptions: an export that arrives late, a figure that looks out of range and a client whose data is laid out differently. It sorts each step onto the ladder. Combining the exports is Execute. Flagging odd figures is Recommend. Signing the report off stays with a person. Only then is a narrow prototype built, so it tests a process that has already been drawn.

What comes after

The next thing is the prototype. It is put in front of the people who would use it, on real work or a realistic sample of it, and it answers one question: does this behave well enough on this organisation’s own work to be trusted with it? The people who do the work judge the result, and their verdict decides whether it is built for production.

What is built for production is then an AI workflow, with an internal owner named for each one that goes live and the people who will run it trained. Scaling happens one proven workflow at a time, each returning to step one of the audit. The business outcomes page covers what the whole sequence is meant to leave behind.

Mix-ups worth avoiding

Four confusions come up often enough to name. The first is treating workflow design as a service with a price, when the site publishes it as a stage and not a product. The second is treating it as a sixth audit step, when the audit has five and this sits between the ranking and the prototype.

The third is treating it as documentation of what already exists. It redraws the process around what AI can do. The fourth is treating design and build as one thing. Scope is fixed in the design so the build does not have to discover it.

Where it goes wrong

The commonest failure is skipping it and going straight to a tool, so the process inherits every awkward step it already had. A second is drawing only the happy path. A third is adding the review point later, once something has gone wrong, and not as part of the drawing.

A fourth is drawing a whole department at once. The site’s rule is one process, the one with the clearest input, output and review point. A fifth is keeping design and governance apart, so the record of which rung each process sits on is kept by someone who never saw the drawing.

Where it appears on the site

Workflow design is named on the home page’s adoption sequence, between opportunity identification and AI automation. The how we audit page sets out the five audit steps either side of it, and the well-scoped workflow article explains why scope decides the result.

Every audit starts with a discovery call, and the drawn process it leads to is an AI workflow.