Most organizations draw themselves as a hierarchy of departments. Customers do not experience departments. They experience a sequence of steps, often crossing several teams and several systems, that begins with a need and ends, if things go well, with a result they value. A value stream is that sequence. Naming your value streams is a surprisingly powerful exercise, and one worth doing before you decide where to put agents, how to form teams or how to plan together.
Two kinds of value stream
- Operational value streams deliver value to an external customer: onboarding a new client, resolving a service request, fulfilling an order, releasing a product update.
- Development value streams create or change the systems, products and services that operational streams depend on.
Many organizations have a handful of each. If you list thirty, you are probably naming departments or activities rather than streams.
Who should be in the room
A workshop of a half to a full day works well with a cross-section: leaders who know the strategy, people who do the work at several points along the flow, someone who speaks for the customer and someone who understands the systems and data. Include a facilitator who does not own any of the streams.
A workshop agenda
- Agree on the customer. For each kind of customer, what are they trying to get done? Write it in their words.
- List the triggers and outcomes. What starts the flow, and what does the customer have when it ends? Be concrete: “a request is received” to “the request is resolved and confirmed”.
- Sketch the steps. Walk from trigger to outcome and record the major steps, no more than eight to ten. Keep it high level: this is not yet a detailed process map.
- Mark the people and systems. Who contributes at each step, and which systems and data are used? Note where handoffs occur.
- Name the stream. Give it a short name a customer would recognize. Check that it has a clear start, end and value.
- Identify the people who work in it. Count roughly who contributes. This helps decide how teams might be organized around it.
- Find the pain. Mark where work waits, loops back, loses information or depends on one person. These are candidates for improvement.
Where AI agents fit in a value stream
Once the streams are visible, you can ask better questions about AI. Instead of “where could we use an agent?”, ask “which step in this flow is routine, rule-based and slows the whole stream?” Work through the steps in three passes.
- Mark the menial. Steps that are repetitive, rule-based and low in judgment are candidates for an agent to handle, with a person reviewing exceptions.
- Protect the meaningful. Steps that need judgment, empathy, negotiation or accountability stay with people. Agents may prepare information for them.
- Check the handoffs. Wherever information passes between people, teams or systems, ask whether an agent can carry the context along so it is not rebuilt each time.
Consider each candidate against the risks in your governance playbook: what data is involved, who is affected and who can switch it off.
What the output looks like
At the end of the session you should have: a named list of value streams, a one-page sketch for each, the people and systems involved, a short list of pain points and a first set of candidates for improvement or agent support. Resist the urge to polish. A rough picture that people recognize is worth more than a perfect one nobody has seen.
What to do next
- Pick one stream and run a Kaizen event on its worst step.
- Use the streams to organize multi-team planning, so teams plan around the flow of value, not around reporting lines.
- Attach measures to each stream: speed, quality, customer feedback and learning.
- Review the streams yearly. They change as products and customers do.
Common mistakes
- Describing the org chart. If a stream is named after a department, look again.
- Going too detailed too early. Detailed mapping belongs in the improvement event.
- Leaving out the people who do the work.
- Treating the workshop as a one-off. Streams should be revisited and kept visible.
Our book Organizing for Organizational AI covers how teams, roles and operating models can be designed around this flow. To have a facilitator run the session, see our workshops for enterprises and teams.